Loading summary
Transcript770 lines
- [00:01]
A
You're watching Backstop TV.
- [00:03]
B
Today is Thursday, November 6, 2025. We are live from the Backstop Ultra.
- [00:08]
A
Dome, the Backstop Arena. Yes, we have our own backstop right here. Look at this. This is how it works.
- [00:14]
B
It's the infinite money glitch.
- [00:16]
A
Just do it.
- [00:17]
B
If you're.
- [00:18]
A
Just do it, ask for it.
- [00:19]
B
If you're, if you're skiing like this, don't over your skis.
- [00:22]
A
If you get over your skis, just.
- [00:24]
B
Use the infinite money glitch to get out of jail free.
- [00:27]
A
That's right.
- [00:28]
B
Backstop yourself.
- [00:29]
A
That's right. Doesn't hurt to ask.
- [00:32]
B
Now, wild time on the timeline yesterday with comments from OpenAI. OpenAI's finance chief, Sarah Fryer, she was at a Wall Street Journal event on Wednesday and she and the.
- [00:47]
A
She put the timeline in turmoil.
- [00:48]
B
She did put the timeline in turmoil. The headline from the Wall street journal was, OpenAI very difficult, wants federal backstop for new investments. It was a crazy thing. She wound up clarifying it. Let's play the actual Sarah Fryer interview at that Wall Street Journal event so we can just set the table and hear what she actually said. Wall Street Journal, let's go. This is a nice event. Looks like it's OpenAI.
- [01:23]
A
At our core are the model company that needs always to be the state of the art. That's what we've done time and time again. GPT5 is no exception. But even in areas like open source, we're attempting to put the state of the art model always out into the world. And in order to do that, we always want to be on the frontier chip. So the question is, how long does a chip remain on the frontier? Is it three years, four years, five years, or even longer? Now, in a world where we have.
- [01:50]
C
No compute or compute constrained, we are.
- [01:53]
A
Absolutely using chips that have like a 100 equivalents that have been around like maybe six, seven years at this point in time. If that's the case, financing chips gets a lot easier. If the timeline on the chip stays short, that gets harder. And so this is where we're looking for an ecosystem of banks, private equity, maybe even governmental. The ways governments can come to bear.
- [02:21]
D
Meaning like a federal subsidy or something.
- [02:23]
A
Meaning like just first of all, the backstop, the guarantee that allows the financing.
- [02:28]
B
It'S like a ton of bricks that.
- [02:30]
A
Can really drop the cost of the financing, but also increase the loan to value. So the amount of debt that you can take on top of an equity portion for some federal backstop for chip investing.
- [02:42]
C
Exactly.
- [02:43]
A
And I think we're seeing that, I think the US Government in particular has been incredibly forward leaning, has really understood that AI is almost a national strategic asset and that we really need to be thoughtful when we think about competition with, for example, China. Are we doing all the right things to grow our AI ecosystem? We're in a global erotica race.
- [03:07]
B
Are you talking to the White House.
- [03:09]
A
About how to further formalize that kind of backstop?
- [03:12]
C
We're always being brought in by the.
- [03:14]
A
White House to give our point of view as an expert on what's happening in the sector, for sure.
- [03:19]
D
Should we, you know, is there something.
- [03:21]
B
In the work that's so funny? Any announcements?
- [03:23]
A
No. I love you Sarah, but nothing to announce. Nothing that's going on right now.
- [03:27]
D
If you wanted to, we're here to listen.
- [03:29]
B
Yeah, Wild. She wound up clarifying her comments on LinkedIn.
- [03:36]
A
Elsewhere the same interview she did say the market is not exuberant enough.
- [03:41]
E
Yeah.
- [03:42]
B
In the same interview.
- [03:43]
A
Yeah.
- [03:43]
B
And she also said that they're not working towards an IPO yet. There were a whole bunch of scoops that came out of this particular event.
- [03:50]
A
And Ben just texted tvpn. The Backstop Programming Network.
- [03:54]
B
The Backstop Programming network. So on LinkedIn, Sarah Fryer said she wasn't asking for a backstop for OpenAI's investments. She was just making the point that American strength in technology will come from building real industrial capacity, which requires the private sector and government playing their part. Before we move on, let me tell you about ramp.com, time is money Save. Both easy to use, corporate cars, bill payments and accounting and a whole lot more. So obviously OpenAI is so dominant in the zeitgeist that you know every comment, every sentence by uttered by an OpenAI executive will be analyzed to death. It's one of my favorite things to do. So that's exactly what we're going to do on this show. We're also going to be restreaming it. One livestream, 30 plus destinations multi stream reach your audience wherever they are. So the big question here is should a federal backstop exist for large scale industrial projects? Like when there's a lot of money floating around in the economy, should the government be ready to step in if things don't line up, if things don't pencil out? And I would argue that one already exists. And that's what Sarah was referring to. And so Sam Altman actually touched on a similar concept in his conversation with Tyler Cowen that was recorded I think a few days ago, but also went live yesterday, which is just sort of crazy. Timing. But basically Tyler is talking about nuclear energy projects and how the government interfaces with them. And then Sam gives an answer that I think is a little bit more illuminating in this whole backstop of the AI buildout. So I'd love to play the clip of Sam Altman, not this year, talking to Tyler Cowen.
- [05:40]
C
Here's a very difficult question. As you know, both you and I were fans of nuclear power, but we also know the insurance for nuclear power plants is provided by the government. The plants might be quite safe, but people worry. They're nervous nellies. There's a lot of parties involved. So the federal government does the insurance. Do you worry that the future holds the same for AI companies where the feds are your insurer and how do you plan for that? Again, even if AI is pretty safe, as with nuclear power, people are nervous nellies.
- [06:09]
B
This is fascinating because it's how will you insure everything? It's not necessarily a question about the financing, but it winds up being one about it.
- [06:17]
A
Like at some level, when something gets sufficiently huge, whether or not they are on paper, the federal government is kind of the insurer of last resort, as we've seen in various financial crises and.
- [06:30]
B
Insurance companies screwing things up.
- [06:32]
A
So I guess given the magnitude of.
- [06:36]
D
What I expect AI economic impact to.
- [06:38]
C
Look like, I do think the government.
- [06:41]
A
Ends up as the insurer of last resort.
- [06:44]
D
But I think I mean that in.
- [06:46]
A
A different way than you mean that.
- [06:47]
B
And I don't expect them to actually be like writing the policies in the.
- [06:50]
A
Way that maybe they do for nuclear.
- [06:52]
C
And there's a big difference between the government being the insurer of last resort and the insurer of first resort. Last resort's inevitable, but I'm worried they'll become the insurer of first resort. And that I don't want. I don't want that either.
- [07:05]
D
I don't think that's what will happen.
- [07:08]
C
What we're seeing. Why wouldn't you want lithium Rare earths is the government is becoming an equity holder again, not of last resort. It's of second or third resort. And I don't mean this as a comment about the Trump administration. I think this is something we might be seeing in any case or see in the future after Trump is gone. But how do you plan for OpenAI knowing that's now a thing on the table in the American economy?
- [07:32]
A
I put almost no probability on mass onto the world where no one has.
- [07:37]
B
Any meaning in the post AGI world.
- [07:39]
A
Because the AI is doing everything like.
- [07:40]
B
I think we're really great at Finding.
- [07:42]
A
New things to do, new games to.
- [07:43]
B
Play, new ways to be useful to.
- [07:44]
A
Each other, to compete, to get fulfilled, whatever. But I do put a significant probability that the social contract has to change significantly.
- [07:52]
D
I don't know what that will look like.
- [07:53]
B
Yeah. And so he kind of moves past the original backstop question. We can move on from this.
- [07:59]
A
But I also think. Let me give some more context here, because a lot of the dialogue on the timeline started to center around this quote that we just played on the backstop. But earlier in that same day, that same interview, OpenAI CFO Sarah Fryer suggested the market is overly focused about on anxiety about a possible bubble in artificial intelligence and should muster more exuberance about the technology's potential. Quote, I don't think there's enough exuberance about AI when I think about the actual practical implications and what it can do for individuals. We should keep running at it. And so she also said, we're all just building out full infrastructure today that allows for more compute to come into the world. I don't view it as circular at all. A huge body of work in the last year has been to diversify that supply chain.
- [08:49]
E
That's for sure.
- [08:49]
B
True.
- [08:50]
A
So, yeah, that part is true. But again, going supply chain is potentially.
- [08:54]
B
Too diverse if you're in the supply chain because you're like, I have a ton of competitors.
- [08:58]
A
Yeah. And part of this is like, you have to put this in the context of the last week. You know, Friday, everyone saw the interview with Sam and Brad. Brad asks a very simple question that's on everyone's minds. Sam basically doesn't answer the question.
- [09:15]
B
It just kind of says, like, we're going to pay with it, with revenue. We're going to hit our revenue miles.
- [09:18]
A
Sure, sure.
- [09:19]
B
There's a whole bunch of different projects that are.
- [09:21]
A
You put that. But nobody, nobody heard his response and thought, okay, I'm confident.
- [09:27]
B
Totally.
- [09:28]
A
It did not inspire confidence. And so to go into that and then say in this interview, the market is not exuberant enough. And also she can go on LinkedIn and her response. I want to clarify my comments earlier today. OpenAI's not seeking government backstop for infrastructure commitments. I use the word backstop and it muddied the point as the full clip of my answer shows, blah, blah, blah, blah. So she can go out and say that she takes it back or that she didn't mean it to come across in the way that it did. But she, like, very clearly in the moment, agreed with the interviewer's point.
- [10:06]
B
Yeah, it's interesting because I feel like it is true that there is a government backstop. Whenever there's a trillion dollars of spend going on, there is a de facto backstop. This happened in 2008. There was a global financial crisis. Banks were leveraged on housing. The government came in and acted as the lender of last resort. Tons of people hated this and it created Occupy Wall Street. People argued that it should not have happened, but economists would say that the alternative was worse cause the economy would just continue to spiral. You'd see more and more bank failures. And so in that moment it was the rational thing to do. Now the problem is that if you are identifying the backstop before you get into trouble, it creates this moral hazard. And that's why people were so upset with Wall street. Because if you go into it being.
- [10:56]
A
Like the main thing is going out.
- [10:57]
B
And announcing we're too big to fail.
- [10:59]
A
Right, Announcing a trillion dollars plus of, of spending, obviously some on balance sheets, some off balance sheet partners, and then going out and not being able to answer the question convincingly on how you pull it off. And then a few days later saying the market's not exuberant enough. We think it would make sense for the government to get involved in some capacity when you're building functionally. Elsewhere in the interview she talked about monetizing commerce, monetizing discovering products, monetizing the transaction. And so to be going out and suggesting something like that when you're building like OpenAI today, yes, it's a national lab, yes, it's strategic to the United States that OpenAI is being built here. But ultimately you're building a company that's just going to compete with Google and Amazon and a variety of other companies, right?
- [11:55]
B
Yeah. And I would include those in the backstop, like if Google overinvest or if Google, you know, is, is doing a training run on a variety of Neo clouds and they get over leveraged and there's you know, like the buck might not stop with one like, like there's a world where there's an overbuild and the people that get caught is some sort of like real estate investment trust or some sort of, you know, aggregator of private credit or some sort of senior debtor on some data center project. And that, yeah, it doesn't really affect the big clouds. But I don't see, I actually don't see a problem with the government behaving like it always has. I don't have a problem with that. I'm fine with a backstop and I wouldn't change anything about the way the government works. And I'm serious about that. I'm not joking because, because I think that, I think that the alternative is actually worse. I think the alternative, if you go back to 2008 and you say, okay, let's not bail out the banks, I think that you just extend the recession.
- [13:04]
A
Yeah, yeah, yeah. But, but if in 2005, yeah, the banks were saying, and various players were saying, we're going to do a bunch of super risky lending that we're not sure is going to pencil out and we just want to know that if things get really bad, you'll be there to bail us out. But we're going to make a ton of money in the process.
- [13:23]
B
Yes, there is. How would you feel about that? There is the problem of it. So my conclusion to my post today was America has a playbook for backstops. I don't think Sarah Fryer or anyone at OpenAI is asking to rewrite the playbook. But there's something that feels very moral hazardy about the CFO of the most important company at the center of a massive economic wave. Sort of telling everyone how the magic trick works. Why people are latching onto this, I think is not so much the idea that if there's an economic boom and bust cycle and the government steps in at some point, that's not that big of a problem. But it is a problem. If you say, I'm openly acknowledging the way the system works and I'm going to push it to the limit, that's a little bit rough. What do you think, Tyler?
- [14:12]
A
Yeah, I was just going to say like the AB test of like 2008 you can kind of say is like Great Depression where there's this big critique.
- [14:18]
D
Where, you know, people were basically mad.
- [14:22]
A
That there wasn't stimulus earlier.
- [14:23]
B
Totally.
- [14:23]
D
And then basically by not doing stimulus.
- [14:26]
A
You drag out the depression like way longer than it should have.
- [14:28]
B
Totally. I mean, it was the foundation of the Keynesian economic model. Like this idea that like, it is actually advantageous to spend. I mean, like, yeah.
- [14:36]
A
And I guess don't get wrong, there.
- [14:37]
B
Are tons of people who would say, like, don't.
- [14:40]
A
Not against government backed infrastructure spending.
- [14:42]
B
That's not what that is. That's not what this is. That's a separate thing.
- [14:46]
A
I know it's a separate thing. And I'm just saying if the government was saying, we're effectively going to guarantee that a bunch of infrastructure can be built, roads, bridges.
- [14:58]
B
I actually like the, hey, we're a backstop. I like the backstop. More than we're going to try and pick a winner and do a specific project. The specific project. That's how we got Solyndra. So like Solyndra was backstopped. It was this solar project. It was energy infrastructure. We wanted to build out clean energy in America. And the, I think of the Obama administration basically gave loans to this company and the tech just didn't work out. And so the taxpayer just got a write off or not a write off but like, you know, the debt didn't work and didn't, didn't get paid back. And so there was just like a loss. That technology never worked. The company went bankrupt and it was a big like egg on the face moment. Obviously it's not going to bankrupt America because it's just one company, is just one small project. But it was still like, why do we spend, I think hundreds of millions of dollars or whatever it was on this like moonshot technology? Like shouldn't the private markets be the ones allocating capital around that? That's very different than if there's a systemic problem in the economy stepping in to just lower the interest rates. Like look at what was the backstop in dotcom. Do you know what they did as a result of dotcom?
- [16:13]
A
No.
- [16:14]
B
Like they didn't go in and say, oh, you're in pets.com, we're making you whole.
- [16:18]
A
Yeah.
- [16:18]
B
What they did was they said there's a recession right now. We need to lower interest rates, we need to cut taxes a little bit. We need to just stimulate the economy just a little bit to get back on our feet and just start up the economic engine.
- [16:33]
A
Yeah.
- [16:33]
B
So they did tax cuts and lower interest rates and that is a backstop and I'm fine with that. I think that's fine. Like if, if the, if the AI bubble pops and the end result is that we get lower interest rates and some tax cuts and like just, hey, let's.
- [16:50]
A
Yeah, but again, I think that's, that's very different. Very different than, than various players wanting to develop data centers and lenders feeling uncomfortable about not having clarity on. Are these chips going to be worth anything in six years? Are they going to be worth anything in five years? Are they going to be useless after a few years? Right. It's like if that's on the lender.
- [17:11]
B
Like the lender can do the analysis.
- [17:13]
A
No. And when I, when I hear, when I hear these comments from Sarah Fryer.
- [17:18]
B
Yeah.
- [17:19]
A
I'm, I'm, what I'm hearing is we have a bunch of lenders that would, that would happily finance this if there's a government guarantee that they won't lose money. And that's very different.
- [17:32]
B
Yeah, that is odd. I don't know.
- [17:36]
A
So if you get some type of setup where the government does come in and say, not just for OpenAI but for a variety of these players, yeah, we will provide a backstop. The lenders start to feel comfortable, they start to lend. What's the lender's incentive? If a lender makes money, when they lend money and they have some type of structure where they can't lose money, what are they going to do? They're going to lend as much money as they possibly can, of course. And OpenAI will take as much money they will take as every. Everybody involved. The lenders will lend as much money. OpenAI will take as much money as they possibly can or benefit as much as they can from it. And ultimately, I just think it's. I'm not against, you know, looking back through, you know, 0708 and the example you gave around post.com crash. I'm not against government intervention during a crisis totally. But using the government to create what will ultimately just be an even greater crisis and ultimately just socializing the losses from all of this and, you know.
- [18:40]
B
I think it doesn't always plan out that way though.
- [18:42]
A
Yeah.
- [18:42]
B
Tesla, government backstop, right. Gave loans, paid back the loans. Tesla paid back the loans with interest. The taxpayer made money on that. And anyone who says, oh, Elon only, you know, he's just, the government pays for everything. That's. It's like they're wrong. The government made investment in Tesla and they got back money with interest. And so the government participating in private markets, it doesn't always end poorly. It's a little weird. And I think that if you go to the Tyler Cowen clip, like, Sam is not asking for the government to take a position. He's like, I don't want to be intel. I don't want them on my cap table. I don't want any of that. It's like, sure. So honestly, I don't know. I don't know what they're asking for because it is a crazy, crazy piece of comms to say to even talk about it in the current zeitgeist because it makes it feel like, oh, well, why are we talking about how to deal with a crisis? Do you think there's a crisis coming that's stressful? Yeah, Right.
- [19:41]
A
And you know, the general public already hates AI enough. Figuring out a structure that allows, allows, you know, potentially socializing yeah. The downside is going to, you know, I don't know. I would expect that something like this does happen.
- [20:01]
B
Yeah. I mean, that's why David Sachs came out and said like, there will be no federal bailout for AI. The US has at least five major frontier model companies. If one fails, others will take its place. It's also. That's a crazy quote too because it's like no one was talking about a failure of a big foundation model lab.
- [20:19]
A
Yeah. And this, this is so. So if open. OpenAI, if we were in a. If we, if the AI race looked more clearly like a China designing AI weapons.
- [20:33]
B
Yeah.
- [20:33]
A
And OpenAI is purely a defense tech player that is also developing AI weapons. And in order to be competitive, in order to not have the CCP take full control over the world with their sort of dark superintelligence, we needed a national champion national lab. We needed to give them as much infrastructure, as much energy, all these things. That scenario I'd be looking at this a lot differently. OpenAI is, in my view, I've been saying this for at least a month now. I just view them as a large tech company. Right. They want to make money from subscriptions, they want to make money from ads, they want to make money from commerce. They have a social media app, they have an erotica product.
- [21:12]
B
I don't have a product yet. Have not launched that.
- [21:14]
A
Okay, well, I'm sure, you know, they've announced the launch of it.
- [21:18]
B
Right.
- [21:19]
A
And people have used it in that way in the past. And so I don't understand the, you know, we need a trillion dollars or.
- [21:30]
B
I think literally me, you, Sarah Fryer, David Sachs. We're all in agreement here that no one's saying OpenAI gets a bailout if the company doesn't do well.
- [21:41]
A
But no one's saying that. But a bailout is different than a bailout is different. I think it's important to like break these out. A bailout is like, in my view is like OpenAI is going to default and they need a capital injection in order to stay solvent. And government.
- [22:00]
B
No one's talking about that.
- [22:01]
A
No one's talking about that. And David Sacks was clear. We're not going to have a bailout but a backstop of loans that labs and player. Like that seems to. I would expect that to happen.
- [22:15]
B
What do you mean?
- [22:16]
A
I expect some type of backstop scenario to like despite me being play out.
- [22:21]
B
Really? You think that there's going to be like major defaults across the.
- [22:25]
A
No, no, no. I'm saying the backstop I'm not saying the major default piece. I'm saying I expect a program, special program, some type of program, like usually.
- [22:35]
B
Only gets set up if there's a crisis. Like the, like TARP, like the Trouble Asset Relief Program in 2000.
- [22:42]
A
Yeah, but there's other, I mean, there's other programs.
- [22:45]
B
Like you don't set up the price until, until there's a crisis.
- [22:49]
A
Okay, but look at, look at housing. There's programs to help people buy homes that are government guaranteed. Look at the SBA loan program.
- [22:59]
E
Right.
- [22:59]
B
Everyone needs a slice of the top tranche of data center debt.
- [23:03]
A
Yeah. So why do banks lend to small businesses through the SBA program?
- [23:07]
B
Yeah, yeah, of course. The economy.
- [23:10]
A
Yeah.
- [23:10]
D
And.
- [23:13]
A
It'S backed by the U.S. government.
- [23:15]
B
Yeah. And just like it's what taxpayers want. Like you want to be able to deduct your mortgage. The American dream, the social contract is that you can buy a house. And so if we can incentivize or make it easier for people to buy houses, that's a good thing. That's what everyone wants. And so everyone kind of socializes the risk of mortgage losses by creating these government backed loans. The question is, I guess the bigger question is like, should America vote for, you know, actually socializing the losses of the AI buildout? Like if you don't like AI, then you're like, no, there's no point. But if you think it's like the railroad, or you think it's like the Internet infrastructure build out, or you think it's like the computing revolution, then you should be all in on that. You should be thumbs up on like, yes, put the economic, put the weight of the US government behind this project.
- [24:11]
A
Yeah. And I just think there's no public appetite for this. Look at the state, look at the state of the average American.
- [24:21]
B
I think you're right.
- [24:22]
A
They can't afford a home. They can't even afford McDonald's.
- [24:25]
B
Yeah.
- [24:26]
A
And they're scared of losing, are they? They're scared of get, they're, they're, they're already worried I might get a job and then I might lose it to AI.
- [24:33]
B
Yep.
- [24:34]
A
And you have people working in AI that are already talking about job displacement. We had it. We, we had a VC on our show very recently. He was saying, you know, talking about what a large opportunity investing in AI is because, because this is labor displacement.
- [24:49]
B
We have some breaking news. Sama has tweeted, addressing the backstop. First, before we read that, let me tell you about privy wallet Infrastructure for every bank. Privy makes it easy to Build on crypto rails securely. Spin up white label wallets, sign transactions, integrate on chain infrastructure all through one simple API. Let's go to Sam Altman. His post. I would like to clarify a few things. Been a lot of clarifying lately. I feel like I've read this post before. There was a new post.
- [25:16]
A
I appreciated that erotica thing. Gavin Baker posted earlier. We need a moratorium on OpenAI public appearances and press releases to stabilize the market.
- [25:26]
C
No.
- [25:26]
B
What are you doing, Gavin? Okay. Gavin has to come on the show if OpenAI people won't come on the show anymore. Okay. Let's read through the same Altman post. I would like to clarify a few things. First, the obvious one. We do not, do not have or want government guarantees for OpenAI data centers. We believe that governments should not pick winners or losers and that taxpayers should not bail out companies that make bad business decisions or otherwise lose in the market. If one company fails, other companies will do good work. What we do think might make sense is government building and owning their own AI infrastructure go code. Then the upside of that should flow to the government as well. We can imagine a world where governments decide to take to off take a lot of computing power and get to decide how to use it. And it may make sense to provide lower cost of capital to do so. Building a strategic national reserve of computing power makes a lot of sense. This is if you, if you believe in the lore of AI 2027. This is trust the plan. Tyler, can you get me up to speed on what, what AI 2027 predicted on the nationalization?
- [26:34]
A
I think this is actually more situational awareness than 2027.
- [26:37]
B
Yeah.
- [26:37]
F
Yeah.
- [26:38]
A
But yeah, basically Leopold says we're going to have the project.
- [26:40]
B
The project.
- [26:41]
A
The project is basically nationalization of the big labs.
- [26:43]
B
Yeah.
- [26:44]
A
All the big build outs are basically on the government's like cap table or essentially.
- [26:49]
B
Yeah.
- [26:50]
A
And it's basically because we see China's.
- [26:53]
D
Basically doing the same thing.
- [26:54]
A
They start doing it and then we need to retaliate.
- [26:57]
B
We want to do it. Yeah. Jensen Huang was in the Financial Times talking about the. The China will win a race with America, says Nvidia chief. We can go through that a minute, but let's keep reading.
- [27:08]
A
He's like, he's like China is going to win the AI race. We need to sell them as many chips as we possibly can right now.
- [27:16]
D
Please.
- [27:17]
A
They're gonna win if we don't give them even more chips.
- [27:20]
E
Remarkable.
- [27:20]
A
Obviously his position is that we want to get China dependent.
- [27:24]
B
Yep.
- [27:24]
A
On. On American AI infrastructure. So that we can pull it back at any time. But saying they're about to win the AI race when he's also advocating for selling them as many, as many chips as as they can produce doesn't really sit that I'm honestly the other thing is like how, you know, let's talk about. We need to better define what the AI race is like. Gurley earlier this year I think made some good points like what is the AI race? What is it? How does one win? What does that look like? What does it look like in the.
- [28:02]
B
Near term if your AGI pellets the race to a super intelligence, a system that can do any economic work. So you can tell it, go fight this lawsuit. And it'll do that. You can say go buy this company. It'll just buy that company. Create a roll up in pharmaceutical industry.
- [28:18]
A
And the reason that people don't take that narrative super seriously right now.
- [28:22]
B
Yeah.
- [28:23]
A
Is that companies like OpenAI make short form social media apps.
- [28:28]
B
People aren't feeling accelerating, they're not feeling the acceleration. They're not feeling the acceleration.
- [28:32]
A
And that's correct when you're out there saying yeah, we want to basically be Google search, we have our own YouTube.
- [28:39]
B
And then you have multiple, multiple experts kind of talking about how they're not feeling the acceleration. You have Andre Karpathy and a number of other folks coming up on and saying like, yeah, I think it might be more like a decade. I don't really know if it's going to be this like fast takeoff. Well then the question is if we're not getting 10x the results, do we need to be spending 10x as much next year? Maybe, maybe those, some of those exponential curves should be sigmoids.
- [29:05]
A
We will, we will find out. Keep, keep reading.
- [29:07]
B
Okay, so what we do think might make sense is the government building their own AI infrastructure. The project is going to happen. Building a strategic national reserve computing power makes a lot of sense. But this should be for the government's benefit, not the benefit of private companies. The one area where we have discussed loan guarantees as part of supporting the build out of semiconductor fabs in the United States. So like I completely agree with this. Maybe it's a crazy hot take. Brandon was saying I'm going to get canceled for being pro pro bao. I'm pro build out, pro backstop, pro bailout. I'm pro all of it when it comes to building semiconductor fabs in the United States.
- [29:48]
A
And I'll say, well, I've been a bit more negative. I'm pro build out I think the power side is an area that could make a lot more sense, that could benefit the average American a lot more, which is like, hey, the government is going to help finance, bring a lot more power online and we're going to be able to stay at the leading edge in AI. Lead in AI and your power bill. We're actually gonna make your power bill cheaper over time. That's something that people could get behind. But federal backstop so you can get more videos of your cat dressed up like a doctor. I don't think people are gonna be on board with that.
- [30:33]
B
Yeah, it's tough.
- [30:34]
A
Let's keep reading this and then, and then give a proper reaction. So I guess before we move on.
- [30:38]
B
Let me tell you about Cognition, the makers of Devin. Devin's the AI software engineer. Crush your backlog with your personal AI engineering team. Do you want me to continue?
- [30:46]
A
So real quick on this concept of the government building and owning their own AI infrastructure. So this concept has existed for a very long time in defense, where there's things called Go Cos. Which are government owned, contractor operated. So this model is popular in weapons manufacturing where the government will own land and they'll own a facility and then a contractor will actually basically run the facility, produce and sell the finished product to the government. But, but that, that is like worked fine over time.
- [31:24]
B
Yep.
- [31:25]
A
And, and so there's some precedent here for that. I think. It's very unclear to me to how building just a national reserve of computing power makes a lot of sense.
- [31:39]
E
I don't know.
- [31:40]
B
I could do a crazy steel man on it, but.
- [31:42]
A
Of course you can.
- [31:43]
B
I can steel man anything. But yeah, let's stick to the semiconductor fabs because that's where Sam is saying that he's actually advocating loan guarantees for semiconductor fabs. This is seven layers away from his actual core business. You should. It is. It's like he's not, he's not. I mean maybe he will be able to fab. But realistically, it's like he's gonna have a partnership with Nvidia, who's gonna have a partnership with intel and Intel's gonna do f. There's going to be a government loan guarantee for what? The loan on the building that actually turns into the fab. It's several steps away to the point where it doesn't feel. It feels like when you have a founder like Palmer Luckey talk about policy in a completely different area or something like that. If Palmer's like, we should make land cheaper, energy cheaper. You're not like, oh, well, this is tied directly to Anduril. You're just like, yeah, this is just a rational policy by itself anyway. The basic idea there has been ensuring that the sourcing of the chip supply chain is as American as possible in order to bring jobs and industrialization back to the US and to enhance the strategic position of the US with an independent supply chain for the benefit of all American companies. This is of course different from governments guaranteeing private benefit data center buildouts. There are at least three questions behind the question here that are understandably causing concern. First, how is OpenAI going to pay for all this infrastructure it is signing up for? We expect to end this year above 20 billion with annualized revenue run rate and grow to hundreds of billions by 2030. We are looking at commitments of about 1.4 trillion over the next eight years. That's longer than hundreds of billions of billion. Yes.
- [33:32]
A
So these are all new numbers.
- [33:33]
B
Because these are all new numbers.
- [33:34]
D
There's 1.4 trillion. The previous quote was over the next many years.
- [33:37]
A
It was kind of unclear.
- [33:38]
B
Yep.
- [33:39]
A
I think also the, the revenue over 20 billion is new.
- [33:42]
D
I think the most recent number was something like 14.
- [33:45]
B
It was like 13.
- [33:46]
A
14.
- [33:46]
B
Yeah, yeah. And so they've clearly grown a ton since that. Obviously if they exit, I mean they won't exit the year at 20, but if they exit, this is phenomenal growth, like remarkable. I mean, they did like four last year or something. What do you think? Not good. Not big enough?
- [34:03]
D
I mean.
- [34:03]
A
No, it's like straight lines on logarithmic curves.
- [34:06]
B
Yeah, for now. For now we'll see. Okay, so let's see what else.
- [34:14]
A
So obviously this requires continued revenue growth. Each doubling is a lot of work, but we are feeling good about our prospects there. We are quite excited about our upcoming enterprise offering.
- [34:23]
B
For example, new categories like consumer devices and robotics that we also expect to be very significant. But there are new categories we have a hard time putting specifics on, like AI. The can do scientific discovery which we will touch on later. We are looking at ways to more directly sell compute capacity to other companies and people. And we are pretty sure the world is going to need a lot of AI cloud and we are excited to offer this. We may also raise more equity or debt capital in the future. But everything we currently see suggests that the world is going to need a great deal more computing power than what we are already planning for. Second, is OpenAI trying to become too big to fail or should the government pick winners and losers? Our answer on this is an unequivocal no. If we screw up and can't fix it, we should fail and other companies will continue on doing good work and servicing customers. That's how capitalism works and the ecosystem and economy would be fine. We plan to be a wildly successful company. But if we get wrong, if we get it wrong, that's on us. Our CFO talked about government financing yesterday and then later clarified her point, underscoring that she should have phrased things more clearly. As mentioned above, we think that US government should have a national strategy for its own AI infrastructure. Tyler Cowen asked me a few weeks ago about the federal government becoming an insurer of last resort for AI in the sense of risks like nuclear power, not about overbuild. I said I do think the government ends up as insurer of last resort, but I think I mean that in a different way than you mean that. And I don't expect them to actually be writing the policies in the way that they do for nuclear. Of course, this is just the soft backstop that exists generally in our in our economy and in our democracy. Again, this was in a totally different context than the data center bailed out and not about bailing out a company. What we were talking about is something going catastrophically wrong, say a rogue actor using AI to coordinate a large scale cyber attack that disrupts critical infrastructure and how intentional misuse of AI could harm could cause harm at scale that only the government could deal with. I do think that the government should be writing insurance policy I do not think that the government should be writing insurance policies for AI companies. I agree with that. Third, why do you need to spend so much now instead of growing more slowly? We are trying to build the infrastructure for a future powered future economy powered by AI. And given everything we see on the horizon in our research program, this is the time to invest, to be really scaling up our technology. Massive infrastructure projects take quite a while to build, so we have to start now. Based on the trends we are seeing and of how people are using AI and how much of it that they would like to use. We believe the risk to OpenAI of not having enough computing power is significantly more and more likely than the risk of having too much. Even today, we and others have to rate limit our products and not offer new features and models because we face such a severe compute constraint in a world where AI can make important scientific breakthroughs but at the cost of tremendous amounts of computing power. We want to be ready to meet that moment and we are, and we no longer think it's in the distant future. Our Mission requires us to do what we can to not wait many more years to apply AI to hard problems like contributing to curing deadly diseases and to bring the benefits of AGI to people as soon as possible. Also, we want a world of abundant and cheap AI. We expect massive demand for this technology and for it to improve people's lives. In many ways, it is a great privilege to be in the arena and to have the conviction to take a run at building infrastructure at such scale for something so important. This is a bet we are making and given our vantage point, we feel good about it. But we, of course, could be wrong and the market, not the government, will deal with it if we are. I like that. Before we get Jordy Hayes reaction, let me tell you about figma.com, think bigger, real faster. Figma helps design and development teams build great products together. Jordy, give me another reaction. What are you thinking? Are you convinced? Does this update you more positively or are you more bearish based on this?
- [38:29]
A
I think it's a good response.
- [38:31]
G
Okay.
- [38:32]
A
I think leading with more information, always good. Yeah. New information, new information, new information, new information.
- [38:39]
B
Not even, not even really. It's sort of buried, but, you know, it's right there. It's right there.
- [38:43]
A
Yeah. So saying, boom, here's our enterprise offering. Great. They have a solid enterprise business already. I'm sure it'll continue to grow quickly. New consumer devices and robotics excited about that. I think it's hard for them to lean on those. Right. Because when you look at the history of big tech companies launching ambitious products, oftentimes they either flop entirely or lose money for a long time. And OpenAI has a bunch of brilliant builders on their team. But again, you look at Sora, even though Sora was innovative. Right. I don't, I don't think it's going to be. I don't think, I don't think Sora will be a successful, you know, consumption platform. I still remain pretty bearish there. And so I think it's hard for them to. When I, when I see that, you know, they're so.
- [39:41]
B
Has fallen. It's now 4th ChatGPT, then Threads, then Google Gemini.
- [39:46]
A
That's for Connor Hayes over at Threads.
- [39:48]
B
Threads is crazy. I mean, we were talking. Was it anish yesterday at a 16Z, he was like, have you posted on Threads with that? Am I crazy?
- [39:54]
A
Yeah, yeah, yeah.
- [39:55]
B
No, it's him, right?
- [39:56]
H
And.
- [39:56]
B
And I was like, yeah, no, I take your point. Like tech. My community is not on Threads right now, but I still think it's Actually worked pretty well. And it's, I mean certainly by the charts it looks like it's working.
- [40:09]
G
But.
- [40:09]
B
But look, but Sora is holding on too.
- [40:10]
A
So. So look at the history of big tech companies which.
- [40:15]
B
Wait, hold on, you gotta see number five, neon. Money talks. Get paid for your calls. It has 914 ratings, a 2.7 average. People don't love this app, but it happens to be the 5th biggest free app in the store. Sora on the other hand has 88,000 ratings. 4.7 stars. Yeah, people.
- [40:38]
A
Okay, so Reds has 1.5 million going back to saying the categories he's excited about. So new consumer devices.
- [40:48]
B
If Zuck's really hard.
- [40:49]
A
So Zuck has been doing this right and just been literally just lighting money on fire for years now, right?
- [40:58]
B
Enterprise offerings.
- [40:59]
A
If you're saying we are looking at about 1.4 trillion of commitments over the next eight years and you're saying that that part of our solution to delivering on this is new consumer electronics. I personally I don't lean on that too hard. Robotics too. I wouldn't be surprised to see various forms of robotics from OpenAI. Right. There's also Merge, which is separate company that OpenAI is invested in, BCI. But none of these things scream cash flow to me, scream high margin revenue that we're going to be able to reinvest against these commitments. Right. These to me seem very likely like businesses, new business lines that also lose money. You have a money, a company that's losing. I guess the sense is that they're losing about $10 billion a quarter today. That's probably, let's assume that's accelerating on they're going to finish the year at $20 billion. Run rate losses are also accelerating. They want to launch new products that are probably also going to lose money. And so again I'm excited about these categories. I think they're going to make cool novel products. But it doesn't give me more confidence that they can meet those commitments.
- [42:21]
B
Let me pitch you a recapsulation of everything we've been talking about. Keep the government out of the application layer, focus on the energy layer, Just work on driving down energy prices and let all the tech companies duke it out and they will be beneficiaries of cheaper energy. But I want tons of American jobs building everything from natural gas to solar. That solar plant, I want 10 of those and I want it staffed by American. How about that?
- [42:57]
A
Yeah, yeah, yeah, I'm on board with that.
- [42:58]
B
We could agree with that.
- [42:59]
A
I'm on board for the government Financing the build out of every possible energy source that we can.
- [43:07]
E
Right.
- [43:07]
A
I hate looking. So that's something we can agree on.
- [43:10]
B
I have another pitch for you.
- [43:11]
A
Super on board with that. The other thing is we're also looking for ways to more directly sell compute capacity to other companies and people. And so when I look at this, when you're simultaneously saying we're so compute constrained that they're making the AI lazy and they're not launching new features.
- [43:29]
B
Yeah.
- [43:30]
A
And then also we're looking at ways to sell compute to other companies. To me, this just, again, this just says OpenAI is a new hyperscaler. And it's so hard to compete with the hyperscalers that have huge amounts of cash flow that they need every advantage that they possibly can. Right. And one way they could get an advantage is through programs like Sarah suggested and then walked back.
- [43:56]
B
I don't think she was suggesting that at all. I don't think she was suggesting a backstop for OpenAI specifically.
- [44:02]
A
Not OpenAI specifically, but there would be certain companies that would. That would get a part of these programs and could benefit.
- [44:12]
B
Yeah. I have another idea. Universal Basic vanta. Every American entrusted to vanta, they could automate compliance, they could manage risk, they could accelerate trust with AI. Because VANTA helps the Americans get compliant fast. And VANTA doesn't stop there. Their AI and automation powers everything from evidence collection to continuous monitoring to security reviews and vendor risk. Whether you're starting up or scaling, the timeline continues to be in turmoil. Jason Calacanis chimed in and said if OpenAI were to fail, it would have zero impact on the future of AI.0. He says they have a dozen viable competitors and 10 folks who would buy the assets, let it rise or fall based on merit. Like, this is like, the fact that this is the reaction to that quote is like, so crazy. Just like, you talk about this a lot with brand where it's like, you can get a lot of attention, but you need to zoom out and think about what are the two words that are going out in people's mind. Right. And so you might have some funny, rage baity viral marketing campaign, but if it's like, oh, we're going to have our CEO dress up like the garbage. Who's the garbage man from Sesame Street? You know, the Oscar, the Grouch.
- [45:35]
A
Oh, yeah.
- [45:35]
B
Like, it's like at a certain point, like, even if it goes viral for whatever reason, like, you're just associating your brand with trash and like, you zoom out and like, that's rough. And like, the idea of like federal backstop just immediately conjures massive failure and it just puts your company and failure right next to each other so very, very rough. Very rough. Let's read what Gabriel said. He said, this is why traditionally CFOs are only allowed to speak fake accounting identities in public. Backstop is what the Federal Reserve calls it when they have to save 20 banks from exploding in a global financial crisis. The word you were looking for, Sarah, was partnership. Yeah, that makes a ton of sense. It's interesting. I'm actually more pro backstop than partnership because I don't love the idea of the government picking winners in any particular layer of the stack. But partnership just sounds so much more positive than backstop which conjures up the. It just conjures up financial crisis. It just feels like major layoffs, recession, stock market sell off as soon as you start talking about backstops. So very rough. Doug over at Semianalysis has another take. I think he's pro backstop too. I think he's AGI pilled. He says, I'm on vacation, which I'm desperately trying not to check work stuff. But I've been reading some books on railroads and like the final piece is the government needs to just start passing pro AI legislation. We are in the Gilded Age. Tyler, did you read it the same way?
- [47:14]
A
Yeah. I mean, I don't even think that.
- [47:17]
D
The government needs to like spend a.
- [47:19]
A
Ton of money on building data centers.
- [47:20]
B
Yeah.
- [47:20]
A
Basically all you need to do is just like remove a lot of the environmental regulations regime. Like we're just not going to get.
- [47:25]
D
Natural gas with the current like regulatory regime.
- [47:28]
B
Yeah.
- [47:29]
D
Or, or we're not going to get nuclear within 15 years.
- [47:32]
A
We're not going to get all these things.
- [47:33]
H
Yeah.
- [47:34]
A
So it's basically like, I don't even, I don't think that you need to have some massive injection of like capital into, you know, the data center or the fabs or any of this stuff.
- [47:43]
B
You just need to remove a lot.
- [47:44]
A
Of like the bloat and regulation.
- [47:46]
B
Yeah, it's, it's wild, wild time. There was a bunch of other stuff that Sarah Fryer said in her, in her Wall Street Journal appearance. They need a huge amount of money to stay on the cutting edge. Compute grew 20x in two years to do that again. It will take a trillion dollars at current compute costs to pay for that. They have revenue, they have VC investment, but thirdly, they need financial innovation on financing their chips. I was thinking to myself, I think, you know that meme that's like what's an opinion that will get you canceled or something like that. And then you like quote tweet it with your opinion or you do like the man standing like I believe I feel like GPU depreciation schedules should be like 30 years. This is a hot take that I'm working on the defense of. But I feel like I don't know exactly about how long they last, but I feel like the viable work that will be done by a one hundreds will stick around for a very long time. And I think that we will be surprised looking back on how like yes, we got better chips, but the old chips actually stuck around and were doing GPT4 level workloads forever. And I don't know how much profit they will make. But I think just in terms of will you have it still running or will you be throwing it out? Because the narrative right now is like, oh yeah, we're going to be throwing out all these chips in two years. What do you think, Tyler? You push back on that?
- [49:19]
A
I feel like it's not like it's kind of an objective question. Like new chips come out, old chips are worth less. They depreciate after like five years. Like you gotta switch out for the new ones.
- [49:30]
B
No, you don't. You don't. Because like you would say the same thing about like an Oracle database, right? Like you'd be like, oh yeah, like you sold someone a physical mainframe computer. Like now there's a better computer. But it's like, no, like it is serving its goal, it's serving its purpose, which is acting as a mainframe computer. And that's why mainframe computers stuck around for so long were better computers. But you already have it. You already have it and you're just depreciating it's not, it's a non cash expense. This is.
- [49:58]
A
So I guess it makes sense if you think that we're still going to be serving GPT4. Yeah, but it's like why, why would.
- [50:04]
D
Anyone want GPT4 if you can have.
- [50:07]
A
A cheaper, better model?
- [50:08]
B
Why would anyone want Fortran when we have Python?
- [50:11]
A
Well, like those are like bloated old companies that like aren't important.
- [50:15]
B
Oh, so we're post bloat now? There's no more bloated companies.
- [50:18]
A
The new companies that are using AI a lot are not going to be the same ones that, that are gonna.
- [50:22]
D
Want to use the old models.
- [50:23]
B
I think there might be companies that are just getting off of on prem data centers, going into the cloud and they're gonna be adopting GPT4 in like 2032 and they're gonna be like, this is sick. And then from 2032 to 2050 they're gonna be like, man, GPT4 is amazing. That's the level of adoption for some of these companies. It's gonna be slow. And so you're just gonna be like, yeah, I got the A1 hundreds for you. They're the best ever. Yeah. But I think, no, I agree with this. This is a truly nuclear take. Ben Thompson wrote about it a little bit that was interesting. He said start with the physical. So I mean, you know, we're in fun times. When Ben Thompson drops on sirtechery the benefits of bubbles. That was the title of his article.
- [51:10]
A
Bubtalk.
- [51:10]
B
Bubtalk. This is like something that I think will be very memorable. Before I read from this, let me tell you about graphite.dev code review for the age of AI. Graphite helps teams on GitHub ship higher quality software. F so start with the physical. A huge amount of the money being spent on AI has gone to GPUs, particularly Nvidia rocketing the fabless design company to a nearly $5 trillion valuation and the title of most valuable company in the world. The problem from the the Carlota Perez perspective is that all of this spending on chips is relative to the sort of infrastructure she wrote about. Railroads, factories, fiber. Short lived chips break down and get superseded by better ones. Most hyperscalers depreciate them over five years and that may be generous. Whatever the correct number is, chips don't live on as fully depreciated assets that can be used cheaply for years. Which means that to the extent speculative spending goes towards GPUs is the extent to which this bubble might turn out to be a disappointing one. So if there's a bubble in railroads and we're like, we think railroads are going to unlock 5% GDP growth. It's amazing. And then you build all the railroads. The worst case scenario is you end up with a bunch of rails and you can still put cars on them, they still work, but with a bunch of GPUs that depreciate and get useless and they break. They don't stick around forever. But I don't necessarily think about it that way. I think that some of these data centers might be useful for a very long time. Even if they're not not on the margin, I don't know, I think they get fully depreciated. I think they still do useful things.
- [52:47]
A
The good thing, John, is we're going to find out.
- [52:49]
B
We're going to find out right now.
- [52:50]
A
We're effing around right now.
- [52:51]
B
We're going to find out.
- [52:53]
A
Ethan Frost in the chat says the A100 and my 8 sleep pod 9 got me dreaming. Slop. You cannot even imagine.
- [53:00]
B
Exactly.
- [53:00]
A
Let's go. Dean Ball says Fryer is describing a worse form of regulatory capture than anything we have seen proposed in any US Legislation. I am aware of a firm lobbying for this outcome is literally rather than impressionistically lobbying for regulatory capture. This is going back to what I was saying before. If there is some type of program where the government is guaranteeing loans made for certain data center projects, like anybody will just get to be in that program. Right. You're going to have to. I'm sure the various players involved in a deal would have to be qualified in some way. Probably going to need a lot of lawyers and things like that. And you're going to need a trusted brand. You're going to need to be able to show that you have demand all these kind of things. And I'm with Dean Ball on this one and Sarah and the team have walked it back. But they said it.
- [54:04]
B
Well, yeah, I mean the administration basically said no immediately too. So even if that was what she meant, she was testing the waters. Who knows? David Sack said there will be no federal bailout for AI. But he did address your comment, Tyler. He said that said we do want to make permitting and power generation easier. The goal is rapid infrastructure buildout without increasing residential rates for electricity. Let's hear for that. Finally, to give the benefit of the doubt, I don't think anyone was actually asking for a bailout. That would be ridiculous. But company executives can clarify their own. Their own comments. Good point.
- [54:38]
D
Yeah. I mean, also, it's like obviously the.
- [54:40]
A
Administration would never say that. Oh yeah, we are going to backstab everything because that just like massively increases.
- [54:46]
B
Like, you know, rampant, totally speculation. That's like totally risky. Yeah.
- [54:49]
D
So it's like this basically does not.
- [54:51]
A
Mean anything of what the administration will.
- [54:53]
D
Actually do because we've seen time and.
- [54:55]
A
Time again like they want to see the stock market be going up.
- [54:58]
B
Oh yeah.
- [54:58]
D
If it starts going down too much.
- [55:00]
A
You know, it's only. I mean, the White House dropped a reminder yesterday. They said midweek reminder. Don't be a panic in. And they put up this pretty cool picture of Trump.
- [55:12]
B
Don't be a panicking. We figured it out by the way.
- [55:14]
A
People are not panicking American.
- [55:17]
B
It's panicking American.
- [55:18]
A
People aren't quite listening yet. Coreweave is down 19%.
- [55:24]
B
It's so crazy that this news cycle led to Zero Hedge saying there will be no federal bailout for AI like Trump AI official on X. Like that's where David Sacks like message goes. Which is crazy because like a few weeks ago no one was talking about needing a bailout. Now it has the aesthetics of like oh you need a bailout. Like what's going wrong? Like that it was going so well just a couple weeks ago. Like it's just such a. It is, it is such a, such a, such a, such an odd vibe. Very, very rough.
- [56:00]
A
Nvidia down almost 8% over the last five days. So anyways, a bunch of other names are down a ridiculous amount. Duolingo is down 30%.
- [56:15]
B
Snap is up right.
- [56:17]
A
Carmax down 18%. HubSpot down 18%.
- [56:20]
B
Wow.
- [56:21]
A
DoorDash down 15%.
- [56:23]
B
Crazy sell off.
- [56:26]
A
And a lot of these. You know what's been notable this week? Companies are selling off after an earnings beat. Well, Palantir down 10.7% in the past five days.
- [56:39]
B
Whether you're going long or you're going short, do it on public.com that's right. Investing for those that take it seriously. They got multi asset investing industry leading yields. They're trusted by millions. Snap shares surged after the company announced a $400 million partnership with Perplexity to incorporate AI powered search engine into Snapchat. The deal gives Snap a new business line and puts the social media company squarely into the mix around AI chatbots.
- [57:07]
A
I said something about this Saturday, not this specific deal, but I was like, why is Snap not monetizing their almost half a billion daus of young people and this is a great way for them to do it. I don't know how long Perplexity can. Oh, they're using cash and equity. That's helpful. We'll see how long they can afford this kind of deal. But in the short term I think it's smart way for them to at least a way to do a single deal that can really drive hopefully some meaningful usage for them.
- [57:45]
B
Yeah. The partnership was made along the third quarter earnings. It seems like Snap jumped. Well, the shares jumped as much as 18% after markets opened in New York on Thursday. Stock was down about 32% so far this year through Wednesday is close. 477 million DAUs in the third quarter for Snap, an 8% jump from a year ago. So they're still growing. And the company also reported 943 million monthly active users. So they're almost at a billion MAUs and half a billion DAUs, which is remarkable. So, yeah, I mean, if you can kind of put those two companies together, that could be a really good outcome.
- [58:29]
A
Maybe, I don't know, wild that Snap has almost a billion monthly actives and is valued substantially less than Perplexity.
- [58:40]
B
That is wild. Yeah, but maybe they'll, maybe they'll tie the knot somehow, do some sort of merger. I don't know. Perplexity has been kind of on the hunt for a dance partner for a while, right?
- [58:53]
A
Yeah.
- [58:53]
B
And we've heard that Apple was maybe kicking the tires and they're deals all over the place. And I mean, we've also said it was kind of odd that Snapchat didn't have a dance partner on the AI side, since every other major social media platform is kind of bolted onto a lab at this point. XAI and X are obviously the most prominent example. But even like LinkedIn and Microsoft, and obviously what Meta's doing and what Google's doing at YouTube, there's so many different ways for that, for every other social network to be a beneficiary of whatever their foundation lab is doing. And Snap just doesn't have that. And you don't necessarily want to stand that up from scratch. Well, let me tell you about Julius AI, the AI data analyst. Connect your data, ask questions in plain English and get insights in seconds. No coding required. Rune also chimed in on Backstop Gate. He says, I don't think the US government should backstop data center loans or funnel money to Nvidia's 90% gross margin business. Instead, they should make it really easy to produce energy with subsidies and better rules. Infrastructure that's beneficial for all and puts us at parity with China. Build your own. Wow. This is an example of a person who's not into AI. Danny, with 142 followers here saying, build your own effing energy sources. Stop being parasites on the grid. And Rune says, yes, all major data centers today compact with power sources. And so it's interesting because the way the grid is set up, the way the US economy is set up, is not like we don't build separate grids for each use case. Like we didn't build a separate grid for Netflix Ultradome. Yeah, exactly. Some people might say that. That hospitals deserve better access to electricity than Netflix, right? Like, do we really need Netflix? Like, is that life or death? No, like we should route the energy resources accordingly. That's not how it works in our economy. Like, we have one grid and we all draw from it and we all pay whatever price that is. But this idea of the government putting the thumb on the scale of and basically doing away with like what is effectively net neutrality or grid neutrality is super, super alluring to people right now. I think you're right that that's going to be like a major political issue that people are going to say, why are my rates going up for the thing I don't like? When it's kind of always been that way. You know, it's been like, you know, why are airline ticket prices going up? Because people are ordering things that are shipping by air. Like that happens. There's all sorts of different ways that these services flow back and forth, but people are not into the AI one.
- [61:56]
A
Joe Wiesenthal says what is the case against bailouts if winning the AI race against China is existential? I'm not actually making this argument or have this view necessarily, but public money backstops bailouts seem like the natural endpoint of how AI is discussed at these levels.
- [62:14]
B
Yeah, I mean the case, yeah, there is no case against bailouts if winning the AI race against China is existential. America in every company should do everything possible. It should be the sole focus. The question is just like this idea of AI being existential and the AI race in China being existential is certainly decelerating because we're moving away from AI superintelligence. The prompt is invade Taiwan and it just happens. That doesn't feel like it's coming. And so people are like, well let the Chinese have some slop. Who cares? It's not that big of a deal because it's more about diffuse little gains in the economy here and there. It's not. It's not.
- [63:08]
A
I feel like I should read through a little bit of this FT article about Jensen saying China will win the AI race with the us.
- [63:16]
B
Yeah. Before you do, let me tell you about fall generative media platform for developers. The world's best generative image, video and audio models all in one place. Develop and fine tune models with serverless GPUs and on demand clusters.
- [63:26]
A
Well said. In the starkest comments yet from the head of the world's most valuable company, Huang told the Financial Times. China is going to win the AI race. Huang's remarks come after the Trump administration maintained a ban on California based Nvidia selling its most advanced chips to Beijing following a meeting between US President Donald Trump and Chinese Leader Xi last week. The Nvidia chief said that the west, including the US and the uk was being held back by cynicism and we need more optimism, huang said on Wednesday and on the sidelines of the Financial Times Future of AI Summit, Huang singled out new rules on AI by US states that that could result in 50 new regulations. He contrasted that approach with Chinese energy subsidies that made it more affordable for local tech companies to run Chinese alternatives to Nvidia's AI chips. Power is free, he said. The FT reported this week that China has boosted energy subsidies for several large data centers run by Chinese tech giants, including ByteDance, Alibaba and Tencent. Local governments have beefed up power incentives after Chinese tech groups complained to regulators about the increased costs of using domestic semiconductors from companies such as Huawei and Cameracon. People familiar with the matter said most such systems are less energy efficient than those made by Nvidia. Huang had previously warned that the latest American AI models were not far ahead of their Chinese rivals, urging the US Government to open up the market to its chips to keep the rest of the world dependent on its technology. But following his meeting with Xi, Trump said last week that he did not want to let China use Nvidia's cutting edge Blackwell chips, the most advanced. We will not let anybody have them other than the United States, trump told cbs. We will let them deal with Nvidia, but not in terms of the most advanced.
- [65:03]
B
Yeah, so there's an argument here for selling chips to China and still competing and winning the AI race, right? It's like giving both teams, I don't know, like pumped up footballs and then turning them loose on the gridiron. Like you're saying, like we still want, we want a fair fight, we want a free market. We want everyone who wants to buy chips can buy them and then it's still a race. Everyone gets a pair of cleats and then you have to actually race. And so what are you racing on? You're racing on ingenuity. You're racing on ability to actually build things effectively, find valuable uses for the technology and then scale those up. It's not the best argument, given that he is like the whole question of like trying to sell chips to China while making this argument is there is a little bit of a conflict there. But he's been very clear about this whole idea of the way to win is to have China standardize on American technology.
- [66:09]
A
And we know from history that even if we give China chips, they will still spend as much as they need to in order to catch up themselves.
- [66:26]
B
I don't know. I do think that there's. I do think that you can put supply pressure on the Chinese.
- [66:31]
A
Sure, maybe.
- [66:32]
B
Yeah, a little bit.
- [66:33]
A
Maybe you could slow it down a little.
- [66:35]
B
I mean, they've been investing in semiconductors for years. Right.
- [66:40]
A
How many five year plans?
- [66:41]
B
Yeah, they're going to continue to do that with or without. But I still think that there is a reasonable argument. It's not the best argument, but there is some level of an acceptable argument to this idea that if you continue to push Nvidia chips there, it just makes the Huawei ascend just a little bit less exciting. Because if you're in a hurry and you're high flyer, what did High flyer do with Deepseek? They trained on Nvidia chips because it was easier than waiting around and figuring out how to do it with Huawei.
- [67:12]
A
For the most part.
- [67:14]
B
Yeah.
- [67:14]
A
The hard. I mean, you know, one of the reasons that people, the, the people that are loudest about the AI race are currently building and enabling products that just look and feel like traditional tech products. Right. And so it makes, it makes. You know, this post, I think is pretty reflective of kind of general sentiment, specifically outside of the tech industry from Elliot. He says, if OpenAI blows up and investors are saved by the government, I would vote for Fidel Castro. Again, I don't think. This is not what Sarah Fryer was proposing. Right. But it is a funny.
- [68:03]
B
Yeah, people are not into it. There's also a little bit of cognitive dissonance there too, because if the products are just tech products, no one was really upset that Uber and Facebook and Google were trying to sell their products into China. No one was really upset about that. The whole reason why it became like, don't sell AI to China was because it was like, AI so valuable, it's so powerful. And so if you don't believe AI is super powerful, then you should maybe be okay with selling it to China because you're just like, well, we'll sell steel and we'll sell corn and we'll sell software and why not sell some chips? Because is it super intelligence? If it is, then yes, you want to hoard all the chips. This is the whole Tyler says.
- [68:56]
A
I just look at the.
- [68:57]
B
Jensen's not AGI because he would hoard the chips for himself.
- [68:59]
A
The timing of, of Jensen coming out and saying China's going to win the AI race and not even clarifying what that means. Right. Are they going to build more, better data centers faster? Are they going to have better models? Are they going to unleash superintelligence at the same time? Because I assume Trump will See this quote, and he's going to say, well, I don't want to lose to China on anything. And that is the kind of like getting that reaction is what can create.
- [69:26]
B
Yeah, I think it goes back to the original AGI pilling definition of what is the scale of compute that you need to win. And if you have an order of magnitude more compute scaling, AI models all over the place, you are in the lead of the race and you want to be in the lead. So what's the risk? This is why when Dwarkesh mapped out the path to AGI and he had the probability distribution, China was like, if AGI happens in 15 years, China is more likely to get it because they're on a steeper energy ramp than we are. But if AGI happens sooner and it is just sort of an algorithmic breakthrough, we do have an advantage in America because we have more data centers right now. And so the idea of the energy race. Jensen is making a good point here. I think he is 100% correct about, like, we need more optimism. We need to focus on these rules on AI that could result in 50 new regulations in every different state. And contrasted that approach with energy subsidies.
- [70:38]
A
I think we're all very against state by state AI regulation. A lot of these laws are just being created and the various companies are not going to be able to even comply with them. And you could have instances where companies just say, oh, start, we're not going to, we're not going to sell into Colorado at all. But Jensen, at least in that bit, was not advocating for government investment on the energy side.
- [71:06]
B
More deregulation.
- [71:08]
A
Yeah.
- [71:08]
B
Which is, I think, a fine point to make. And I think it's a fine, I think it's fine to make that point as a point of contrast with China. Like, if you want to get President Donald Trump's attention around energy regulation. This is a pretty good headline. This is a pretty good headline, right? I don't see the problem. Like you want to say, hey, let's drive down energy costs in America. Make it about geopolitics.
- [71:38]
C
Why not?
- [71:40]
B
Do you have a problem with that?
- [71:42]
A
Making it about geopolitics? It's pretty effective.
- [71:47]
B
It's effective, right? You said you want to get Trump's attention. This is the way to do it. It's good. Let me tell you about turbo puffer search. Every byte, serverless vector and full text search built from first principles and object storage. Fast 10x cheaper and extremely scalable. Where do you want to go next?
- [72:08]
A
This post from High Yield. Harry, please Bro, just subsidize our losses and then we're good. Please. We're a national strategic asset.
- [72:15]
B
Wall street gossip just put the too big to fail thing. People. People are going wild.
- [72:21]
A
I mean, this is. This has been.
- [72:23]
B
Oh, this play by play from Aaron Francis is pretty wild.
- [72:26]
A
Sam Altman snaps at Brad Friday Michael burry shorts.
- [72:29]
B
Did he actually. Did Michael burry actually short? I thought he was back and forth.
- [72:32]
A
But no, he did short. It was just exact. The size was exaggerated.
- [72:36]
E
Oh, okay.
- [72:37]
B
That's what it was.
- [72:37]
A
Ghana. Alex Karp crashes out on tv. I didn't think that was that much. It just seemed like a pretty normal carp interview.
- [72:47]
B
Oh, yeah, I saw a little bit of CNBC.
- [72:49]
A
He was like, maybe slightly more annoyed. DoorDash Mrs. OpenAI CFO floats a bailout. Not quite accurate. Floats a backstop maybe, and then takes it back. McDonald's is for the upper class. Did you see this?
- [73:04]
B
No.
- [73:05]
G
Like.
- [73:08]
A
The lower income consumption of McDonald's is declining, which is concerning.
- [73:16]
B
That's not great. Well, we will have to continue digging into that later. For now, we have our first guest of the show, Brett Taylor from Sierra. Brett, how are you doing? Good to see you.
- [73:28]
E
Doing great.
- [73:28]
B
Congratulations on all the progress.
- [73:31]
E
I appreciate it.
- [73:33]
B
We have to hear your reaction to Backstop Gate. We could definitely go over Sierra product updates first, but do you have any immediate reaction to this? The timeline? Because it's crazy right now.
- [73:50]
E
The timeline are you talking about?
- [73:52]
B
I'm talking about the X timeline. This idea of a federal backstop for AI buildout. The build out that's going on. Everyone's going back and forth. Do you have a take?
- [74:01]
G
Take?
- [74:01]
E
I do not have a take.
- [74:03]
B
Okay.
- [74:03]
E
I'm sorry for no hot takes on this one. It's like, as you know, we just got off of Sierra Summit. I just been waiting. And in the timeline today, no hot take right now. Sorry for that.
- [74:15]
B
Okay, in the trenches with Sierra. Let's break it down. What are the biggest updates in your world?
- [74:21]
E
So, yeah, yesterday we had our first customer conference, Sierra Summit. It was awesome. I mean, you know, as an entrepreneur, it was sort of funny. I had this like, before and after post on X and it was like, like, you know, just a few years ago, it was like Clay and me and a whiteboard. And now we've got 400 customers in person. But three big product announcements. First was some new channels. So if you think about customer experience, you often think about things like phone and chat. We announced that you could publish your agent directly to ChatGPT. We think ChatGPT is the new front door for the Internet. And if you think about building your customer experience, it's on your first party properties like your mobile app or your website. But a third properties like ChatGPT and eventually Gemini and others as well.
- [75:00]
A
So the flow there is somebody is looking for, they're having an issue with a product, they're talking with ChatGPT about it and then theoretically there could be some type of pop up that would say like, do you want to talk with an agent about this specific issue? And then the individual company would have more control over that interaction. Is that right?
- [75:24]
E
Well, it's sort of, it's built on ChatGPT apps which was announced at developer day earlier this month, late last month.
- [75:31]
H
Month.
- [75:31]
E
And the idea is this, let's say you know, you're a home seller and you have a first party agent which you can go to your website and chat with it, maybe browse through homes. Well, a lot of your customers are starting on ChatGPT as well. So you've done all this investment to make this agent, you have to build it twice which parts of your experience you want to chatgpt. Now with a click of a button, you can just publish your agent directly via ChatGPT Apple and it really means you can have the same experience in your mobile app and your website as you have on ChatGPT with this new extensibility model they developed. Here's the big picture. I mean, if you just think of any consumer brand, every new platform brings distribution channels, right? The mobile had app store, the web had portals and then search. In the age of AI, it's agents, it's chatgpt, it's Gemini. A lot of our customers, customers are thinking about what do we want our presence to be on our own properties, what do we want our presence to be? When people are starting from ChatGPT, we think just the world of channels are changing. Now it's going from chat and phone and your website to chat and your phone and your website. ChatGPT Gemini agent to agent communication. So when we help our, our customers build customer experiences, we're trying to, you know, help them reimagine what that might be. In a world where a person's personal agent might be doing the shopping for them, how do they set themselves up for success there? The complex part is there's sort of two sides to that coin. It's like where do you maybe commoditize your brand? What do you want to hold back on your first party experience? What do you want to be present on these third party sites. So we've really tried to build a platform that's flexible, but we're really excited about it because I think it is just a brave new world. People are trying to figure out where demand generation demands fulfillment comes from in this age of AI and really proud to be one of the first companies to enable it.
- [77:23]
B
Yeah, help me understand a little bit more about the brand side. I feel like as a consumer there's so many times, I'm sure so many people do this in ChatGPT. They're asking about a product and they're just asking the underlying model and they're just assuming that if I'm asking about Diet Coke, that chatgpt has hoovered up all of Reddit on Diet Coke and the actual Diet Coke website and the FAQ page. That Diet Coke has been put up for years. And so if I ask it how many calories are in there, it's going to get it roughly right. But at a certain point, the other side of the equation wants to have a little bit more control. And it feels like you can publish that as an app in the ChatGPT app store. But what is the funnel from? I just went into a new chat and I don't know that it's in the App Store, but. And I'm just talking about Diet Coke over here. In just a chat, is at some point ChatGPT gonna say, hey, you should actually just be talking to the real app?
- [78:21]
E
I think that is the big question. You know, I don't think the discovery model is completely known yet. And you know, our, it's sort of funny, but our reductive form is like no one really knows where consumer behavior is gonna go or even how these discovery models work. But step one is make an agent. Because if the only way to use your service is clicking around your website, there's no way to be present in that world because it's just going to be, you know, essentially the index content within the large language model. So our philosophy now is it's like the early days of the App Store or the early days of the Internet. I mean, I remember when I was at Google in 2003, the index would update once a month. We called the Google Dance because people's rankings would change. You fast forward 20 years. SEO and SEM is a super mature industry. Right. It's like something that is quantified. We're kind of in the 2003 of this, so no one really knows. But our whole philosophy is like, step one, take your customer experience that Your website, your mobile app today, turn it into an agent. And what that means is enable it to be conversational, enable it to work over the phone, enable it to work on your own website, your own mobile app, then experiment with publishing parts of it to ChatGPT as a ChatGPT app. And you know, the discovery model. How do people activate it with an ention? Does it recognize your intent and light it up? I think the form factor here, just like newsfeed, has changed a lot over the years. I don't think anyone could completely knows yet. But you can't wait until your competitors figure it out first, right? So this is an interesting thing about these new technologies and one of the things we try to do at Sarah is say, hey, we're going to be a partner to you to help navigate this world even though it's uncertain. And I think a lot of people also, it's a question of commoditization. If an agent is doing the purchasing on behalf of a consumer. What's the role of my brand? What's the role of my first party property? It's almost like if you look at the retail world in mobile, which retailers had the brand equity to have their own mobile app which were dependent solely on Instagram for the traffic to their retailer. It's a very nuanced question and everyone's kind of figuring out in real time. So I don't pretend we have all the answers. But what's neat about what we've done is we're saying, hey, step one, make an agent and prepare yourself for this future and make it super low cost to experiment. Which I think if, if we're in the 2003 era of search, being able to experiment quickly is worth a lot. And that's kind of the value proposition for our customers.
- [80:37]
A
Have you ever seen enterprises adopt a product like they're adopting? Sierra, I'm thinking about the post you did. Maybe it was a couple weeks ago at this point or 10 days ago or something where it was like excited to welcome and then you just listed off like 40 companies in the Fortune 500. I've never seen a company as ever make a post like that. It was pretty shocking. But yeah, I'm curious if there's any other moment in your career that you've seen this kind of velocity of adoption.
- [81:13]
E
I really haven't. And man, I can tell you somebody who started two companies before I joke with Clay and my co founder. There's two kinds of stress as an entrepreneur. The existential kind, because you're not growing or the holy shit, everything's on fire kind, because everything is growing and we're experiencing the latter and it's fucking awesome. Excuse my language. You know, it's like just because, you know. Oh, sorry, like getting bleeped out here, you know.
- [81:36]
A
No, no, no. I was just an air horn.
- [81:40]
E
Yeah, okay, good. You know, it's, it is fun. And actually, here's the thing, actually, I'll say one other announcement we had yesterday, but just to answer your question, like, the world is changing a lot and right now companies are looking for a trusted advisor to help them, guide them on this new path, whether it's what demand is going to come from ChatGPT, what does a conversational version of my interface look like? Are my competitors moving faster than I am? Just think about if you run a large telco right now, every basis point of attrition is worth millions of dollars. And if you can get even a small edge, these are the moments in technology where you can actually change market share, even in markets that are sort of ossified. And so, you know, I appreciated that post on X as well. It made me smile when I wrote it. But on stage yesterday, you know, we had the Rocket Mortgage executive team talking about how the Rocket Assist, which they built on our platform, has improved the conversion rate for refinancing by 4x. We had the CTO of Wayfair on stage talking about how it's impacted the returns and warranty claims. We had the Chief Product Officer of Sirius XM talking about the same. And we launched this product called Agent Data Platform, which I'm happy to talk about more, which is saying, hey, can we move the world of agents from conversational customer service agents into agents that can have long term relationships, long term memory and actually drive sales? And I think why were we able to have all these people show up yesterday is we're looking towards the future, I think, with a bolder vision than a lot of our competitors in this space. We have a really, I think, consultative approach with our customers. To say, we're not going to just throw software over the wall and hope you're successful, we're going to help be successful as technology. And just like our conversation about ChatGPT is, we're not going to predict the future perfectly, but if you attach your cart to our horse, so to speak, we're at least going to be marched in the right direction. And I think a lot of companies right now are trying to figure out what they want to be when they grow up as it relates to AI agents And I'm really proud that we're the partner for some of the best brands in the world.
- [83:41]
A
Yeah, I'm sure you saw or saw clips from the Karpathy interview on Dwarkesh saying that it wasn't, it wasn't the year of agents, it's the decade of agents. What was your reaction to that? Because I'm sure internally at Sierra, it certainly feels like agents are having a moment. But I think broadly agents as a category at least, consumers have felt a little bit let down on some of the product experiences today. On the consumer side, we've been kind of promised these product experiences that aren't really good enough to use reliably yet. But on your side, but you're obviously seeing companies roll these out, get tremendous results. And so I feel like it's here, it's just not so evenly distributed.
- [84:24]
E
I think that's right. I think both are kind of true at the same time. There's basically AI agents and AGI. And I think as you think about consumer agents, for those to really work, it maybe isn't complete AGI, but it's close because the amount you need to generalize for an individual like the three of us, have you looked at the home screens of our phones? There's probably very little overlap, Right. To actually have an AI agent that can automate all of our lives, you sort of get the union of all the complexity of everyone's personal lives. In contrast, if you think about sort of, you know, Rocket assist on, on Rocket.com, you're helping somebody go through the refinancing process. It's a relatively structured agent in the grand scheme of things. You know, it's. And so I think where the value right now in the AI market is where you can turn science into engineering. So it's companies like Sierra helping with customer engagement, sales, customer service, or Harvey working in legal. And then the moment you say we're going to make a computer using agent that does everything for all people, that's a science problem. And so Andrej Karathi is probably one of the smartest people I've ever met. So I think he's probably right in a lot of ways. But I think it will be uneven. I think we'll end up with AI agents that make really great math discoveries. But at the same time, you're going to have an agent that can barely use a computer effectively and you're like, it's like somehow a genius and an idiot at the same time. And that's the complexity of this technology. But that's why I'm so bullish on the applied AI market. You know, I think I like to think of it as like, we're taking the science and turning it into engineering. And, you know, I'm really proud that we work in healthcare with, like, Cigna and R1, the revenue cycle management firm. We work with multinational banks, we work in the mortgage industry. And it's because we're taking all this science and we're whittling it down to here's what's shovel ready right now, today. And I think the reason why we've had such traction in the market is that's really valuable. You know, there's a lot of science projects out there. We have a term for it at Sierra called AI Tourism. And we're like, no, AI tourism is not useful. It's just, you know, peacocking for your board that you have an AI strategy. What you want are the things that can actually drive sales and cost savings today. And that's what we're trying to deliver at Cira.
- [86:33]
A
How is the role of and responsibility set of customer experience teams changing already from the companies that you guys are working with? I mean, historically, it'd be like, maybe you start out as like an individual, like CX rep or agent. You're, like, handling tickets. Eventually you're managing a team of reps, and then you're managing a team of managers that are managing reps, and you just go from there. Do you see it evolving already.
- [87:02]
B
Or.
- [87:02]
A
Is it more an individual rep is managing tickets themselves that get elevated and then kind of overseeing the work. How is that kind of evolving today?
- [87:12]
E
One of the most fun and interesting things is with a couple of our customers, the team that was managing their contact centers changed their job titles to AI architects. So you go from teams who essentially have operating roles managing teams of people, to teams who are responsible both for those teams of people, but also the agent that works with them in concert. So some of the agents answering questions automatically, you know, some is actually managing the team that when the agent transfers, goes to them. But it's a real organic system. One of the other features we launched yesterday was called Live Assist, and it's saying, can you take the agent you made for your consumer and put it on the desktop of the people in your call center as well, but have a really seamless conversation, you know, because there's not going to be things like AI agents are great, they're not great at everything, but can you have, like, a single system of intelligence that helps with that whole process? So it's dramatically changing the role, particularly of the teams managing customer experience, because it's gone from just managing people and processes to managing this piece of software. And the thing, I mean, you all will understand this intuitively is when you take something analog, like a phone call to a person and you make it digital. With an AI agent, all the past two decades of digital technology are available to you. You can run AB tests, you have real time analytics, it can self improve. Those are all things. You can't run an AB. I mean, you could if you have a 12,000 person call center. Running an AB test is like a huge operational burden, but now you can do it with the click of a button. And the interesting thing is it's stuff we've been doing on web pages for the past 20 years, but now you can do it on, on phone calls. And so it's actually for our product managers, like a lot of our inspiration is actually saying, like, what have we been doing in the world of digital technology for the past two decades that we can just bring to this previously analog channel? Which I think is a really exciting and oddly simplistic way of looking at the technology. But it's kind of how we think about it.
- [89:09]
B
It's awesome. Yeah.
- [89:10]
A
Do you care? Would it upset you if the models didn't improve at all for the next five years?
- [89:19]
E
It would upset me, particularly as a chairman of OpenAI, but I think I get where you're going with the question, which is how much breakthrough technology do we need for Sierra to make it applicable? I think right now actually the agents work exceptionally well. I'll tell you two areas that I think could really improve a lot. Native voice to voice models can get a lot better. They sound really interesting today, but a lot of them are prone to hallucination and, you know, aren't great at rules adherence. And I think right now a lot of applied AI, you might be in the world of speech to text and text to speech because you need better reasoning, better adherence. And I think that will change. And I'm really excited about in general, video and voice. Feel like we're still in the early ish innings relative to the core reasoning models. And then the other thing is, I think, I think I always bring this up. We're like in the 1997 era of making agents. Like if you look, I found this article for Sierra Summit about the creating websites in 1997 and there was this Wired article, I'll send it to y'.
- [90:25]
B
All.
- [90:25]
E
And it was basically about banks spending $23 million to add transactional support to their website. It's like adding a login form basically. And then you fast forward to like the late 2010s and you know, Kylie Jenner starts a multi billion dollar cosmetics line with as I read, had seven full time staff. And you're like, that's kind of. So I think we're kind of in the 1997 era of building agents where it's still like way too hard and you end up putting a lot of engineering around what is a very intelligent set of models just to make it work well. And I think I always imagine like what do you need to have a seven person team create a multi billion dollar business on agents? And I think we just have a lot of product and technology work left to do. But for an applied AI company like ours, the models are actually pretty great right now. I would be disappointed if they didn't get better, but I don't see like a major technical barrier to progress right now, which I think is really exciting, you know, as a business person.
- [91:21]
A
Yep, makes a lot of sense. Amazing to get the update. Congrats on the first summit and come back on again soon.
- [91:27]
B
We'll talk to you soon.
- [91:28]
E
Thanks.
- [91:29]
B
Have a great rest of your day. Before we bring in our next guest, let me tell you about Google AI Studio. The creator create an AI powered app faster than ever. Gemini understands the capabilities you need and automatically wires up the right models and APIs for you. Get started at AI Studio Build. Our next guest is the CEO of Roblox, Dave Buzzucki. How you doing, Dave?
- [91:50]
C
Good to see you guys.
- [91:51]
B
Good to see you again.
- [91:52]
A
Good to see you again. Twice in one week.
- [91:54]
B
Look at one week. We didn't stream our last conversation. I'm very excited that we are live for this one. But since it is your first time on the show, I would love for you to just give us a little bit of an update on where the business is, what's the newest, what's the latest and greatest in your world.
- [92:13]
C
Hey, so first off, thanks for having me on. I am a huge fan. You're reinventing really the future of business news. So I love your multimodal streaming strategy and it was great seeing you guys two days ago. Every day about 150 million people now come to Roblox and they play, they communicate, they hang out not in games made by Roblox, but literally in experiences made by a huge creator community. So everything on Roblox is built by the community and we've shared about a year ago. We believe Roblox will get to 10% of all gaming running on the platform. We're at about 3% right now. Just came off a great earnings call last week. Our year on year bookings is growing 70% and our year on year DAUs are growing 70%.
- [93:07]
B
I asked you this before, I'm going to ask you again because it's a funny question. It feels like you have built such a powerful platform, the business is running so smoothly. You have this amazing cohort of particularly young gamers who are going to grow up and probably stay with the platform forever. Can you just do nothing and win? Like if you just keep the servers online and keep the platform flexible, it's not like you have to come up with the next greatest piece of intellectual property because you are the platform. So if I misunderstand that, what do you have to do day to day to actually move the ball down the field to get to that 10% that your goal is?
- [93:52]
C
I like the way you think of that. For those of us that are mathematicians out there, we sometimes think of the second and third derivative, not just the first derivative. What you shared is exactly right. If we walked away, just kept the lights on. There's enormous native growth in the platform. This is a highly technical platform. Creators make experiences, they auto translate, they run anywhere in the world world. We've now seen peak concurrency records for gaming all around the world. In August, grow a garden. One of the more recent big games Ron Roblox hit 25 million concurrence. So I would say we cannot walk away. We need to support infra and support scale. But if we did walk away, I think there may be some native growth that keeps going. What we think about when we're running the company though is the exact opposite of that. We think about are we constantly layering in systems, are we building a machine that can keep growing? And are we building a machine that as we hit 10% of the gaming space, we're growing rate by that. And the gaming market's pretty interesting. It's about 180 to 200 billion and it's arguably really sitting right there for technical innovation, for new ways of distributing, for games that run really well on low end Android phones and on high end PCs for games that incorporate AI within the experience. So we just think there's enormous headroom in this market.
- [95:32]
A
What's exciting to you about AI as a creative tool or, or a tool that enables people to build games faster? And what's most exciting to you about AI that can be integrated into a gaming experience? Because I feel like they're kind of two distinct buckets.
- [95:50]
C
I think there's several buckets behind the scenes. Over the last four to five years, we've built out a real substantial offering of models. We've built ourselves. We have over 400 models, team. Teams that are building everything from cutting edge voice and tech safety models, teams that are building models that auto translate, teams that are powering our search and discovery models. We are starting and very shortly we're going to offer what we call inexperience 4D generation, which is literally the ability to make a game that you or I are playing in. And we can prompt and ask for some kind of type of vehicle, for example, or some type of dwelling and literally build it as we're playing. So it's a little bit of a wide open field here as far as what types of games are people going to build when they have unlimited AI, not just for the creators, but for everyone who's playing in that game. And I do think the future spec, which we think about, which is an enormous technical spec, is can we host one to 100,000 people together? Can they be either photorealistic or cartoony, whatever the developer wants? Can they all be talking, listening, interacting with each other? And can they dynamically change the environment? Can one or two or many of them essentially be dungeon masters? And changing the environment dynamically with AI, that's a huge technical lift, but it's what we're working on.
- [97:30]
A
So when I think of the Roblox business and people are very excited about AI and gaming, I view not to kind of gas you and the team up too much, but I view you guys as having the kind of communication layer, the user base, the distribution. It gives you such an insane advantage. Because if I'm somebody who wants to create a game, I can go to. I can create a standalone game. I could go to another platform or things like that, or I could go to Roblox, where I know there's hundreds of millions of people that want to play games. I. How yet talk about. Yeah, do you think there's a. What percentage of game, when you talk about like owning 10% of the gaming market, do you think. Do you think game developers are of the future, will just be kind of defaulting to Roblox? Is that a goal of, like, if I want to build a game, it's like, why. Why go through the effort of like, trying to build this, build my own kind of like community and player base when I can go to where an existing player base is and have A much higher likelihood of success.
- [98:38]
C
Yeah, I think we don't rest on our laurels, obviously, but I do think the gaming market is a bit different than, for example, the video market. And in the video market, whether it's sh, short, UGC, TikTok, YouTube, Netflix, you name it, that's relatively mature technology. The cameras are relatively mature. Video distribution, playback is relatively mature. But the spec I shared with you, 100,000 people in a multiplayer immersive world with AI is far from technically figured out. And that kind of goes to our vision that the next future of gaming is much more client server, cloud integrated. It's why Roblox has many, many data centers supporting this. It's why behind the scenes, it's not just the game engine, it's the persistence engine and the analytics system and the economy system. And as you correctly noted, the social graph, a friend graph. Where are you, your friends, can you find them? As well as the client system, which is, does a game run on iOS, Android, PC, Mac, Xbox, PlayStation, Quest, and more to come? I would say there's one other angle that we really focus on, which is the economics. The more efficiently we can build this whole platform and the more money that can flow back to the creators, the more that combination of technology and discovery starts to make a lot of sense. And so we do focus on running a very efficient company that pushes money back to the creator community. Well over a billion dollars will flow back to Roblox creators this year.
- [100:25]
B
It reminds me of Substack. There was a time when if you wanted to write online, you started by spinning up a blog website, like writing the code or at least forking the code, or managing your own server and making sure your website cap have stayed up and now you can just build an actual game in Roblox. And I understand that it's early, but that does seem like where it's, what.
- [100:47]
A
Are the keys to managing an economy? Because I feel like you have the unique challenge of managing. You're running and building Roblox the business and the platform and the infrastructure. And then there's the economy.
- [101:01]
B
Do you have a chief economist? Like a whole economics team? Like, how serious is that discipline?
- [101:07]
C
It's really serious. We have a lot of PhDs, we have great econ brains. A lot of times kids will say, hey, why can't you just give me a trillion Robux? And it's a great opportunity to tell kids that if we give you a trillion Roebucks, the average Roebuck will be worth about a billionth of what it used to Be worth?
- [101:29]
E
Yep.
- [101:30]
C
So we actually know about money supply. We know we can't print currency.
- [101:35]
B
Are you the Jerome Powell of the Roblox ecosystem? You get DMs directly, people ask you to print, to turn on the money printer.
- [101:44]
C
Well, imagine I'm a young kid. Why can't you just make a trillion Robux? And so there's a lot of learning there that the value of the ecosystem doesn't change when you print more currency. The currency is just worth less. So we're very careful with the value of our currency. We've done something really right I think from the start. And there's a real temptation to do nonlinear functions accelerators. If a dev gets really big, should we pay them more per hour, per minute, per Roebuck than others? We've really stayed away from all those nonlinear systems. And I think that's added a fairness to it. We, I do believe we have aligned incentives of creators and users in a good way. So that if the system is working, it keeps working year after year.
- [102:39]
B
Is there any world where there's an independent Roblox Fed either through crypto or some sort of spinning out or some sort of nonprofit or some sort of, you know, advisory group where like the monetary policy of, of the Roblox ecosystem is not set by anyone with Roblox stock?
- [102:59]
A
Sounds like a nightmare.
- [103:00]
B
It sounds like a nightmare. I'm not advocating for this. I'm just wondering, have you done the thought experiment?
- [103:04]
C
Yeah, absolutely. There are some fun thought experiments around. If I'm holding Roblox currency, is it backed by a digital vehicle and is that digital vehicle stable and backed by a dollar? So there are, you know, as we get bigger and people start to hold larger amounts of money and we, we have both Robux and Fiat currency, it is worth thinking what gives confidence to consumers.
- [103:29]
B
Yeah, yeah, that makes a lot of sense.
- [103:30]
A
How, how early can you tell that a game is a, is a home run, like a hit?
- [103:35]
B
Oh, interesting.
- [103:36]
A
Like, is there like within, like let's say like 10,000 people try a new game? Is that enough? Can you get enough of a read from that to understand that that game could someday have a million concurrents?
- [103:49]
C
I have both two answers to that. One answer I would say is a more of a. Some companies work intuitively and they design great stuff. Other companies work hyper analytically and they have systems and whatever. On the intuitive front, a couple years ago I was just playing on Roblox and the. There were about 800 people in an experience called Dress to Impress, which was, it was like a Fashion game with their own avatars, their own clothing style, competitively put on your clothing and compete in a Runway show. And it felt really good and I forwarded it to the team just because there's a lot of attempts to really get fashion good simultaneously. We have a great team of people who work on search and discovery and our general philosophy is let's make search and discovery as predictable as possible, as transparent as possible, and as aligned as possible with both the users trying to find cool new stuff as well as the creators, which is bubbling up great new stuff on the platform. Dress to Impress got picked up by that algorithm and got forwarded a bit and that helped drive the speed of the viral success. So both, I would say the users themselves share with word of mouth as well as our algorithms surface grade content. It's why I believe in the last year we had seven experiences hitting 10 million daily active users at some point in Q3, and I believe five of them were new in the last 12 months.
- [105:35]
B
How do you think about the layers of the Roblox economy? And I'd love for you to explain just a cut deeper, maybe on the grow a garden example, but are there creators or do you envision in the future creators that would identify not as a Roblox creator, but as a grow a garden creator? And you're basically creating like a sub economy within the overall economy because the tools of game design are so robust that there's exchange rates. How do you think about the layers of the Onion on the platform that you're building?
- [106:11]
C
I like how you think about this layer thing. In a way, gaming is very different than video in that on a platform like Roblox you can create an experience on Roblox, so you're a creator and that experience can have creators within the experience. And so there are two layers there. There's a little bit of a long term future proofing in that there are more and more experiences. Where they started initially as experiences like Lumberjack Tycoon where you would build a lumber mill and you would be a creator in Lumberjack Tycoon and compete as a creator within that experience. I think where this goes is there are quote games on Roblox that have individual creators within them that may actually make money. Being a creator within an environment made by someone on Roblox, that's a really good long term, essentially dual layer strategy that gives us a lot of capability to support a wide range of creation styles. We've seen that with some experiences like Clipit, for example, was an experience on Roblox where you make short little 3D vignettes. And there are actually creators in that experience itself.
- [107:32]
B
Sorry. Before we move on, I have two wild card questions. First, everyone in the chat wants to know because we saw a lot of people yesterday with lots of fish in the background. What's the biggest fish you've ever caught?
- [107:44]
C
Oh, my gosh. You may be talking. I think it's. There's an experience that's gone really big in Indonesia and we're really happy about the notion of more experiences being different in different countries rather than one experience being hot anywhere. So fish it. Literally, we were thinking about why is it big in Indonesia? We were talking in our executive meeting on, you know, Tuesday. Maybe to go big in North America, you need like a giant bass boat.
- [108:13]
B
With a big Motorola.
- [108:16]
C
Yeah. Whereas Indonesia, it's a little bit more traditional fishing.
- [108:20]
B
What about personally, have you ever been fishing? Have you ever caught a big fish?
- [108:23]
C
Personally, I have been in fish it, but I've never caught a big fish.
- [108:27]
B
What about in the real world?
- [108:29]
C
In a big world?
- [108:30]
B
No, no, no. Like. Like not in Roblox at all, but in the. Like on a vacation. Have you ever, in the physical world, caught a fish in the physical world?
- [108:40]
C
I grew up in Minnesota and I caught one day a bunch of small, little sunfish sitting in a small little boat on a lake.
- [108:48]
A
There we go.
- [108:48]
B
Okay, that's something. Second question. How have you engaged with. Do you believe in the simulation hypothesis? Are you familiar with this?
- [108:58]
C
So are we in a simulation? We are in a simulation. If we are in a simulation, are we in a simulation within a simulation? I have two different views of it. View one, I would sure love the answer. Can you give me the answer, please? Because knowing the answer to this is probably like that is essentially the answer to the creation of the universe, depending on what religion you are. What is God like? That is an answer to about a million things. So I would love the answer. The other one is it's a little unfathomable to me. But we're thinking about. Because a computer that could simulate the universe we are in starts to be a computer that is 100 to the hundred to the hundredth power of processing and all of those types of things. And that just blows my mind. Anyway, so. So if we are in a simulation, the next layer universe that can support this simulation is pretty crazy.
- [110:02]
B
Yeah, yeah. It's a fascinating thought experiment. Absolutely wild.
- [110:06]
A
I wouldn't be shocked if this conversation that we're having right now is actually happening in Roblox.
- [110:12]
B
Yes, and some dimensions in Roblox.
- [110:15]
A
And it's just a simulation itself on.
- [110:17]
B
Something much more tractable.
- [110:19]
C
Actually in the next layer universe, Roblox has been very successful and we don't even know it, but we are in the next layers. Roblox hanging out right now.
- [110:29]
B
I believe it. I believe it. Something way more tractable. Vibe coding for Roblox games, is that a separate company? Is that features that you'll build? Does this already exist? The ability to generate a website or the ability to Vibe code generally is just remarkable. It's lowering the barrier to entry on. On designing mobile apps and designing websites. I imagine that the floor is going to go even lower on what it takes to build a game in Roblox. What are you seeing?
- [111:03]
C
I think Vibe coding is maybe an intermediate step to what we would call real time 3D or 4 dimensional dreaming with other players, which is literally you and I are walking around an environment talking or doing things. And as we're chatting over there, a 3D experience or game or object appears. So I think there's something even further than Vibe creating a Roblox experience, which is how do we support in a Roblox experience a command line where you can basically ask for anything with a friend and you'll see it pop up, whether it's a game or an object or a piece of clothing.
- [111:46]
B
Jordy, we can talk about AI build out. I'm always fascinated by that.
- [111:51]
A
Yeah, I wonder. Yeah, I guess.
- [111:53]
B
Is that affecting you at all? I mean we talked a little bit about.
- [111:55]
A
Yeah, as someone who's developed and manages data centers, how have you been processing the broader AI build out and some of the timelines that some of these projects are happening?
- [112:07]
C
Yeah, I would say the good news is I feel for Roblox very early on we started with a build our own data center philosophy, both core and edge data centers. We have data centers all over the world, in Brazil and Singapore, in Poland. And when we hit these big numbers like 45 million concurrency, we're being able to do that at a reasonable cost. Because we've done that, we are building out our GPU fleet as well. We already, as I said, have 400 models running. And we'll do that in the exact same thing we do with our corned edge data centers. Run the majority of the time on our own bare metal, complement it with burst at peak times to give essentially the optimal cost and reliability.
- [112:54]
B
Okay, maybe take me through like the first big I'm going to buy a data center moment. What was that like emotionally? Were you Using debt for the first time in the company. Were you not nervous or was it just like, yeah, of course I'm financing it. It's like I'm buying a house. Like psychologically, as a CEO of a steward of capital, there's now someone coming in potentially higher in the capital stack. How'd you finance it? What were you thinking?
- [113:19]
C
The good news is we generate a fair amount of cash and we've always generated cash.
- [113:25]
B
So pay for it in cash. He bought his house in cash.
- [113:31]
C
So we actually tend to buy a lot of things with cash. I think if you look at our balance sheet, you'll see billions and billions of cash and a billion of debt we raised a few years ago. So so far we haven't gone heavily down that financing route. We build data centers and they last for three, four, five years. Whatever the depreciation cycle is. We're going to continue to do that. And I don't think we're getting your having to borrow a lot of the data centers we build. We buy capacity and then put our own hardware and bandwidth in together. We'll continue that. We tend to build in many, many locations rather than all at the same place. So so far so good.
- [114:18]
A
When we've seen CEOs out there planning for six or seven year depreciation schedule, you said 3, 4, 5. Do you think they're potentially being a little too generous? Not singling out any specific company, but I'm curious how you're planning with your GPU fleet around depreciation.
- [114:40]
C
Yeah, I know we optimize the length and don't quote me, we should look up exactly what our depreciation is for CPU fleets. I do think CPUs are generally aging better than GPUs. So I think it's worth taking a look at the speed of GPU development versus CPU development and we may see those things end up in slightly different places.
- [115:06]
B
Yeah, a lot of that just comes from if CPUs are increasing 2% a year and GPUs are increasing 20%.
- [115:12]
C
That's exactly right.
- [115:13]
B
You're just getting be way, way, way.
- [115:15]
C
It's worth thinking about for a thousand dollars what kind of GPU I'm going to buy in seven years, there's a chance it's faster.
- [115:23]
B
Yep. Like.
- [115:26]
A
Just a small chance. Wow.
- [115:28]
E
Wow.
- [115:28]
B
Hot take of the century. I love that. Thank you.
- [115:32]
A
This is so much fun internally with the team. Like what are like, how would you kind of describe some of the messaging? I mean we're in such an insane market right now. Roblox went public in 2021 during the peak of ZIRP, had to navigate through quite a lot over the last few years. And now it feels like an even more volatile, unpredictable market. With the team, are the conversations like, hey, let's kind of zoom out and kind of look at you started the company in 2004, so you've seen a few cycles by now, but how have you been leading through this year?
- [116:17]
C
I think earlier what I said is we couldn't predict if all we did was keep the lights on what the growth curve is. I do feel we treat the business as if we have to be innovating at maximum velocity constantly. And we can imagine this spec and we can imagine what our product would be like with some of the things I mentioned earlier. And we do also think there's a future for massive innovation in 3D communication relative to the gaming market and beyond. So we have our heads down building stuff, shipping stuff all the time. We focus a lot on how we've organized the company. It's almost as if Roblox is eight or nine companies within a company, given the breadth of our tech stack. I mean, we are an economy, an AI company, an infrastructure company, a game engine in the cloud company, search and discovery user social graph company. So we have great leaders across that and in each one we have enormous innovations we're trying to bring to market.
- [117:23]
B
Fantastic. Well, thank you so much for coming on the show.
- [117:27]
A
Great to hang out, great to be. Congratulations to the whole team.
- [117:29]
B
Congratulations and congratulations.
- [117:32]
C
Congratulations to you and the model and I love what you're doing.
- [117:35]
B
Yeah, we'll talk to you soon. Have a great rest of your day.
- [117:37]
A
Talk soon.
- [117:38]
B
Goodbye.
- [117:38]
A
Bye.
- [117:39]
B
Before we bring in our next guest, let me tell you about Profound. Get your brand mentioned in ChatGPT. Reach millions of consumers who are using AI to discover new products and brands. We have Vlad Tenef from Robinhood in the Restream waiting room. Let's bring him in to the TVPN ultradome. Vlad, how are you doing today? We'll bring him in in just a minute. What else is going on in the timeline? Andreessen Horowitz announced an A16Z new media fellowship. Eric Torenberg has the same Neo media as many. No, new, not Neo. These are different things. We describe this, we work through this. Or as many people are saying, the Thiel Fellowship for the terminally online. Finding highly functional people who are also very online is hard. So we're creating a fellowship for them. Tyler had a hot take on it. What did you say? What was your what was your New Media Fellowship question?
- [118:29]
A
It was like the riff on the teal question.
- [118:31]
B
Yeah, it was.
- [118:31]
E
Who are.
- [118:32]
A
Who are your mutuals that are like.
- [118:34]
B
That no one else has mutuals with or something like that. Anyway, congratulations to everyone over at A16Z on the new Media Fellowship. And let's bring in Vlad Tenev from Robinhood. How you doing, Vlad? Good to see you.
- [118:49]
A
What's up, John?
- [118:49]
D
How you doing, Jordy?
- [118:50]
E
This is amazing.
- [118:51]
A
Good to see you. You always. You always. Impressed with the camera quality is fantastic.
- [118:56]
B
I also love it because last time you were a little bit further away from the camera and you're getting closer and this is good. And in 10 years, when we're interviewing you for the 50th time, you're gonna be right up. You're gonna be in the studio, actually.
- [119:07]
A
Yeah, yeah.
- [119:07]
B
Maybe just be in person. Just come to LA the next time you're in Hollywood, please. You're welcome. But anyway, let's get to what's new in your world. Take us through what's top of mind. There's a lot going on, but break it down.
- [119:20]
D
Let's see. We did our earnings yesterday and earnings was incredibly strong. Basically just relentless. Product velocity continuing, leading to just increased market share across all assets. You have all the blue chip assets that we've had for a long time, equities, options, crypto. But then prediction markets have just been ripping. You know, we've been doubling every quarter, quarter over quarter since we launched just about a year ago. And so far in October, October has been bigger than all of Q3 put together.
- [119:59]
B
Wow.
- [120:00]
D
And we started at the beginning of Q3. We had nine business lines that generated 100 million plus of annual revenue.
- [120:07]
A
Get the gong now.
- [120:08]
D
We have 11.
- [120:09]
A
They have 11 now right here. We're hitting the gong for 11, nine figures. Revenue business lines.
- [120:19]
D
I love that you guys also have a gong. Please do have one.
- [120:23]
B
But take me through, like, what are the second order readings on how your business is doing? Like, there's one read where I could say, okay, if Robinhood's doing well, that means the market's booming. Maybe there's higher volatility, so people are trading in and out. There's a general trend of people trading, maybe more on their app or more on their phone. There's younger people getting into the market. There's all sorts of different reads. Did you learn anything this quarter about the trend of your. How Americans are participating in the economy broadly? Are there any lessons that stuck out?
- [120:59]
D
I mean, I think that you're right. I think we also get this question of how much of your success is driven by the broader market.
- [121:09]
B
Right.
- [121:10]
D
That's why we try not to dilute ourselves and get too confident when volumes are going up. And we also try not to beat ourselves too bad if volumes are going down in the broader market. We really look at market share and we publish these numbers. It's in the slide deck on earnings market share growing across pretty much all the assets.
- [121:33]
B
Because there is just a very old legacy infrastructure of, you know, we've all seen Wolf of Wall Street. Like there's the stockbroker that calls you, that's probably going out of fashion and you're kind of eating that dinosaur one bite at a time, probably. But that's the part that people don't necessarily see because they're familiar with you and the company and the product, but you are taking market share every single quarter. Probably something like that.
- [121:56]
D
And we're still very small actually in the grand scheme of things. Like we have a third of a trillion of assets on the platform, which sounds like a big number, but there's going to be 120 plus trillion in assets handed down from older generations to younger over the next few decades.
- [122:14]
A
How many?
- [122:16]
D
Wealth transfer.
- [122:16]
A
What was the number?
- [122:18]
D
120,000,000,000, plus in the great wealth transfer. So Robinhood currently is like less than a third of a percent of that. I think we've got a lot of room to run. And if you're a young person in this country, millennial, Gen Z, soon to be Gen Alpha, you don't really want to do your finances at a brick and mortar store. You don't want to call someone on the phone. I mean, you'd like to have someone available if you do want that. But these are digitally savvy, digitally native people that are comfortable doing everything on the smartphone. So what they care most about is having great technology. And I think Robinhood has a real shot at being not just your primary, but also your secondary financial account. We want to be both. So we're just marching along on that vision.
- [123:07]
B
Talk to me about how research is changing. I had this interesting Jordy and I have been joking about we want to get into buying a lot of land. I went to ChatGPT and I said, how can I buy land whether I have $100 or a million dollars? What are the options? What's the most most underlying asset? That's maybe in an ETF or a reit. And it gave me some options. And I feel like this idea of people coming up with investment theses from listening to stuff, but not necessarily getting a specific ticker or a specific portfolio. And then they want to go and construct that and actually build something that expresses their belief. Maybe they believe in America and they want to express it that way. How do you think research is changing and actually servicing the assets in more of a conversational way? Maybe there's AI in there, but how are you thinking about research changing?
- [124:06]
D
I think one of the things we're seeing is it's becoming more granular. And a lot of people thought that it wouldn't happen necessarily in this way. For a while. The trend was really toward indexing and ETFs as a, as a new product, got adoption. They started getting tons of assets. When we were first starting Robinhood, we were kind of moving in a different direction and there was some skepticism from investors who thought that everything was just going to be indexed and indirect and it would just be people investing in broad market indices and diversified funds. And we made a bet in the other direction that in fact, no people want to move away from indexes. They actually want to express their viewpoints more precisely. And so you're seeing more granular investments, individual equities, but not just individual equities. You're seeing options trading, the rise of zero day options, same day expiry. And I think prediction markets are a continuation of that because you're starting to see markets for stocks and whether they're going to outperform earnings. For example, a lot of, a lot of people used to play the presidential election and have a thesis there that they would play by trading stocks or crypto, but now you can do that directly. So I think, I think it's getting even more granular. People will have very specific theses that they understand and they have conviction on and we're giving products for them to express. And on the land side that's super interesting because I would bucket that under alternatives. Right? Like people want access to alternatives. The only way to have access to alternatives right now is through REITs or these diversified indices. But what people really want is to invest in an individual private company as easily as they can invest in a public stock. And maybe they want an individual piece of land or a property in a neighborhood that they think is going to do well rather than just some like commercial real estate fund that has, that has a portfolio everywhere because they can, they can better understand whether that's going to go up or down. I think we're moving in that direction. Tokenization is, is solving a lot of the technical problems with that. And of Course there's regulatory hurdles that once we get through, I think we'll will open up lots of alternative assets to retail investors.
- [126:39]
A
What companies and CEOs do you think do the best job at shareholder engagement and what do you think are best practices around shareholder engagement today? I think I would point out you guys as a company that does this incredibly well. I think part of your strategy is maybe let's do what we want other companies to do. Right. Because. Because the better that management teams get and companies get at having and building relationships with shareholders, the better experience it is to be a shareholder. But what have been your kind of inspirations that have guided your guys strategy and who else do you admire in the market?
- [127:21]
D
Yeah, I love talking about this because our IR team, Chris Cagle, who leads IR from us, has been getting a lot of calls. I think he's become a little bit of a luminary in the how to engage retail shareholders using traditionally kind of like boring forums like earnings. And we've been on a journey ourselves. I remember watching our earnings call a year ago and you know, like many people I go on YouTube and I see let's see who's talking about our earnings. And there were actually groups of, of influencers that watch our earnings live and kind of talk about it. And I sat through the experience and they were looking at for a long time just a slide deck with our logo in there. It wasn't even the new brand of our logo, it was the old one. And then it's just like, you know, we were on a Polycom, the the analysts were on the call, nobody knew whose turn it was to talk. It was a bad experience and we started just by wanting to clean that up. So we did our first video earnings call 3/4 ago and every quarter we've been thinking about how can we do a little bit more. And if you view it from the perspective of a retail shareholder, it can't just be we're reading scripts of information. It has to be engaging. And so it's video. It's getting more stakeholders that can ask good questions live on video if possible, asking those questions in real time. And you know, you guys have figured it out in your forum. I think that what earnings is, is a community event and you should think about it as such. And I think the best CEOs, the best companies who really care about engaging retail have to move from thinking about it as a chore to actually one more event where you can get your brand and what you stand for out to as many of your stakeholders and community members as possible.
- [129:23]
A
It's like thinking about it, like thinking about it like a product. Like if you think about like the earnings call is like a product launch where the customer is the shareholder. Like, you know, most companies today, their earnings calls and the way they treat earnings does not reflect that at all. And it's, and it's such a relatively small investment compared to just operating the company at large for potentially a much better experience. But you were talking about Palantir.
- [129:51]
D
Yeah. I mean, what I'll tell you is we're helping companies. Whenever someone comes up to us and asks us how can we improve earnings, we always want to help. Like I win if more and more retail access the markets. And more companies treat retail as, as first class citizens. So we're more aligned there and we have products. We've got SEI Technologies which offers shareholder Q and A to retail. And we're actually piloting a program where we live stream our own earnings in our app and we'd like to make that available to other companies. So open door after the close today is the first company outside of Robinhood that's streaming their earnings live on video via our app. So I think, and we've gotten a lot of, a lot of companies asking us if they can do that. And we can put together a whole bunch of things that really center around how to make the experience really engaging.
- [130:54]
A
On the topic of retail generally, how are conversations going around IPO allocations? I know in the past you've been very loud about getting more and more allocation for, you know, not, not even just Robinhood users, but retail broadly. What, what's the update there?
- [131:14]
D
Yeah, so when we launched IPO Access, which is our product for getting retail exposure to, to IPOs, it was 2021 and we did it in advance of our own IPO, but we always wanted to make it more broadly available. At first people were skeptical. You know, CEOs and their bankers would try to talk them out of it. More often than not, they would ask, why am I doing this new thing? I should just focus on running my business and get my bankers to, you know, tell me what to do for my ipo. And there was not very much incentive to do things differently. So we really had to like claw and scratch and ask for favors to get retail allocations and IPOs back then. Then the IPO window shut for many years, but when it reopened, it's been a huge change. Now basically every single company comes and asks us about their retail strategy because they've seen what happens in the Public markets to companies that engage retail effectively. And so allocations have been generally going up. Retail is more engaged. CEOs are talking about how, you know, they're giving bigger and bigger allocations. You guys probably saw Bullish, which was after Figma. You know, bullish CEO came out, gave retail a huge allocation, 20% of his deal, 10% of that went to Robinhood. And that used to be considered a crazy number. I mean when we gave retail allocation of 20, 25% in our IPO, people kind of understood because it's your Robinhood. But the idea of any company doing that seemed outlandish. So we're doubling down on that. I think it's clear to companies what we bring to the table, which in many cases the demand on the Robinhood platform for retail shareholders is larger than the total amount they're raising, which has been a huge thing. We can give them all sorts of information about the demand curve and how much retail wants at any given price, which is very valuable. So we're building out that business. We're also going earlier stage. We want to give retail access to companies before they go public. And in the US we've got Robinhood Ventures, which is our big initiative towards that vision.
- [133:45]
A
How much can you share on Robinhood Ventures specifically and how that will enable more retail access to privates?
- [133:53]
D
Well, I could tell you it's a closed end fund and it's going to be a diversified vehicle for non accredited and it'll be like closed end funds traded on an exchange.
- [134:06]
B
Got it.
- [134:06]
D
So when we looked at what kind of vehicle retail would want to invest in, we knew that it had to be liquid. Most people don't like their money being tied up. We knew they wanted high quality names and you know, ideally people would want it to be individual names, but that's not, that's not easy to do for unaccredited and with liquidity. But I think we're moving in that direction. So we're in the quiet period where we're going to. We're moving towards an IPO of this fund and I think it's just a big step towards our vision of being the go to place for investing, whether it be public, company or privates.
- [134:50]
A
Yeah, I've been thinking a lot lately about the challenge right now that the big labs are going through where there's such a relatively small number of people, individuals and funds that are. Even these cap tables have swollen, but they don't have a lot of public support. We've seen when they go through kind of rocky Periods in the same way that if Tesla, for example, has a bad quarter, there's an army of people that are saying this is just a blip. We're riding, we're riding, et cetera.
- [135:23]
B
Everyone wants to know about prediction markets. How are you thinking about partnering on other products? I mean, I remember when Robinhood launched Bitcoin, it was, I don't believe it was in partnership with anyone under the hood, but actually I don't remember. So how do you think about partnerships, prediction markets, Kalshi, where everything is going there and how you, how you got place even.
- [135:46]
A
Did you, did you predict how quickly prediction markets would grow? I mean, that's a great question. In a year they've gone from being something that felt like it was just something that online people talked about and believed in the potential in.
- [136:02]
B
But it was a presidential tracker.
- [136:04]
A
Exactly.
- [136:04]
B
That's what it was.
- [136:05]
A
Exactly. We went from presidential tracker to mainstream in a year.
- [136:09]
B
So yeah, let's talk through this.
- [136:12]
D
Yeah, I mean, I think that initially, before the presidential election last year, prediction markets were, were part of our 2026 plan. And we were planning on offering futures outrights like oil and gold and all the standard futures contracts. And we thought, well, prediction markets could be a way to make it a little bit more mainstream, but they haven't really worked in the past. But then there was something that happened that changed everything and it was the Supreme Court decision allowing presidential prediction markets. And we jumped on that as, as soon as that result came out. We were, about a month before the presidential election, we like reorganized our roadmap. We launched the presidential election market in advance of futures outrights and that did about half a billion contracts in one week, which is a big business, particularly week one in. And then that's followed with the super bowl, which we had to pull back. But then we relaunched sports with March Madness and it's just been on a tear ever since. And I think the story is people thought of it as an election only thing, but prediction markets are much broader. We now have over a thousand contracts on the platform and, and in the same way that options trades in zero days allow traders to express a more granular viewpoint, prediction markets do the same. But they have this other added benefit that they're a source of news and information. So even if you don't want to trade, if you want to know about what's happening or what's likely to happen with a sports event or a political event or an election, you're going to those prediction markets as a more accurate source of information. In many cases more accurate than polls.
- [138:09]
B
Yeah. How do you think about the shape of the team at Robinhood that will work on prediction markets? Is it fundamentally a different legal structure or financial team? Do you have a desk market making or different market making partners or is it a different technology? Because I imagine on all of the crypto stuff you need developers who understand how to write smart contracts at a certain level. Are prediction markets any different from other products? Are they fundamentally different or is it somewhat similar to what you've already been doing?
- [138:43]
D
It's pretty similar to what we're doing. I mean, there are CFTC regulated event contracts and so it's under the same license that we use to offer futures. If you think about it, you've got market data coming in and the trades go on an exchange. Right now we partner with two Kalshi and ForecastX and market makers plug into those exchanges to provide liquidity. From a risk management perspective, in some ways the contracts are even simpler instruments than options and futures. There's no leverage, for instance, that poses more complicated risk management challenges.
- [139:25]
B
Got it.
- [139:26]
D
But we are investing a lot. I mean, from a product and design standpoint, I know there's a lot more that we can do. And if you're paying attention to the Robinhood product, it's changing on a weekly basis.
- [139:38]
B
Yeah, yeah, that makes a lot sense.
- [139:39]
A
How do you, how do you. What's the framework for prioritizing new product lines you've now gone from? Very rarely do companies create a $100 million revenue business and create a second $100 million revenue business. And even more rare is to go 3, 4, 5, 6, 7, 8, 9, 10, 11. Now you guys have built the muscle to take new products from 0 to 1. I'm sure that. But at the same time, story of Robinhood is like focus and delivering great customer client experiences. So I'm curious what the framework is. What justifies net new investment when you're already at this massive scale and it's hard to. When you have these products that have insane product market fit, you have this incoming wealth transfer that I'm sure a lot of that will flow to Robinhood and other new players. What qualifies something as worthy of real investment?
- [140:36]
D
I think the North Star is really how do we become your financial super app? How can we get all of your finances done on Robinhood? Make it so that you don't have to leave. And the reason that's our goal, by the way. I think it's not obvious because a lot of companies have found great success Just going super deep and finding a narrow niche. And we don't aim to just be the best prediction markets platform. We want to be the home for all of your money, your primary and secondary financial account. And the reason is we found that fundamentally, customers love the simplicity of having all of their money and all their financial accounts in one place. And there's advantages to that too, because if we can handle all of your finances and offer you all these products, then we can actually take a cut. We can take a hit on the economics for each one, knowing that in aggregate, the relationship and the average revenue per user is still going to be a big number. So we can operate each one of them at lower margins by virtue of offering more of these. And you ask our GMs, you ask Johan what the best. Johan's our crypto leader. What's the best Robinhood feature? What's his favorite one? He'd point to the brokerage business because one of the main advantages of having your crypto on Robinhood is you can easily sell your crypto buy an equity even over the weekend. And we're the only platform that allows you to trade these things in one place. You get prediction markets, you get crypto options, futures, equities, all using uniform kyc, great funding, Rails that make it easy to move money in and out. We've got a great credit card and banking's rolling out. And because we have all these other products and sources of revenue, we can be super competitive on our banking offering too. Early reads from customers that are testing it out is that they love earning a competitive rate on checking. For example, you know, most banks don't give you anything on checking, and they make it hard for you. They say, okay, well, you know, checking is just for moving money in and out. Savings is where you put money if you want to earn interest. Of course, they don't pay much interest in savings either, but that's beside the point. The reason is they have to earn money somewhere. And checking is like the most simple, straightforward way. So I think it allows us to break that model a little bit more and offer more differentiated value across all products.
- [143:20]
A
Makes total sense. Recently we've seen a number of both public and private companies doing layoffs. And a lot of the messaging they use around that is around how much efficiency they're getting out of AI. Our theory, or my theory personally, is that a lot of these management teams need a good answer for why they're doing a layoff. And saying we're getting efficiency out of AI is one of those how much. Do you think that's real, or do you think a lot of these companies are getting so much efficiency that layoffs are justified purely on the capabilities of AI today?
- [144:02]
D
Yeah, I'm sure you guys have seen that chart on social media that it has. ChatGPT launched in September of 2022, and then s and P goes up and job openings go down, right?
- [144:15]
A
Yeah, totally. And. And I, yeah, ChatGPT is a great product, but it is not replace. You know, we run a company that does research and creates content, which ChatGPT is great at. And we have not. We have not been able to replace anybody. There's that we're still, like, hungry to hire people. And so in our view, even though we're getting leverage out of it, I.
- [144:37]
B
It's been an incredible tool.
- [144:38]
A
Like an incredible tool, but it's not causing us to layoff and it's not causing us to hire any less.
- [144:45]
B
It's been SaaS. Yeah.
- [144:47]
D
I mean, what I tell you is that chart is interesting, right, because the time that ChatGPT launched also coincides with the time in, like, startup history where companies were starting to get fit, right?
- [145:00]
G
Yeah.
- [145:01]
C
And interest.
- [145:01]
A
And interest rates were going through the roof.
- [145:04]
D
Interest rates were going up. But you had Elon taking over Twitter, and then you started seeing lots and lots of companies doing layoffs. You know, Meta started doing them. Even Alphabet, Google did one, which it went from being this taboo topic where almost, if you were doing it, it was a sign of your company failing into everyone's doing it. So it had like this mimetic quality to it. So I think there's an element of that. I don't think it's just AI, but I do think AI has significant impact. I mean, we really felt it ourselves. So. So we focus on AI across the business, but with a particular emphasis on engineering and customer support. So we track, for example, on customer support, one of the top metrics is AI deflection rate. What percentage of tickets that would have come in anyway are now being handled by AI agents. And on the engineering side, we track what percentage of code is generated by AI and the average amount of code that engineers contribute on a monthly basis. And what we've seen is there's real progress. I mean, we had an infrastructure issue that affected our AI chatbots a couple of weeks ago, and we saw an immediate influx. The phone started ringing like crazy during that time period. So it felt very theoretical at first, like, maybe you're just seeing these numbers go up, but do you really know the impact?
- [146:33]
A
But yeah, that Saying, saying that we're getting value in engineering and customer experience is like the, that's like pragmatic and, and very real. That, that to me is the places that companies are getting leverage at scale that's like undeniable. It's when people, when people do a layoff and they're laying off, you know, 10,000 people like, and they're not necessarily engineers or customer support reps, that's when I go like, okay, what were they doing and how is AI kind of replacing their work? Or maybe they just weren't providing that much value and that's why the layoffs.
- [147:06]
D
I completely agree that it's a very effective smokescreen because it makes you seem, you could argue that it makes the company seem better, right? Oh, they're really front foot on AI. So anytime there's an opportunity like that, you should assume there's an incentive for companies to over embellish what's happening.
- [147:27]
A
Absolutely.
- [147:28]
B
At the end of every interview we ask you about Harmonic. Now I'd love to know if there's any sort of change in your thinking. We were just talking to Brett Taylor and he was saying that one of his predictions is that we might have AI that can do advanced science and novel math, but still not really requires a whole bunch of harnessing to just, you know, answer a service ticket effectively and you know, his business will continue while AI is doing advanced math. He kind of laid out a bull case for both Harmonic and Sierra coexisting which, which you know, in the, in the singularity didn't really make sense for a while, but now I think people are wrapping their minds around it. Have there been any updates to. You're thinking about AGI superintelligence, just how tractable attacking science with AI is. Are we on a decade long fight for this or do we think we're going to be making breakthroughs sooner? Just how's your thinking about mathematical AI?
- [148:30]
D
Yeah, I think that it's for sure happening and I'm sure your guys X feeds are a little bit different than mine. I follow all the math researchers and what you've seen happening over the past couple of months in particular is mathematicians who are maybe professors at top universities. There's a UCLA professor, Ernest Ryu for example, who published a thread, it went somewhat viral where he's like, hey, I used AI to actually prove this new result that didn't exist before. And it was a little clunky. I had to go back and forth with the AI. Some of the ideas didn't make sense, but there was this one idea that it had that actually ended up being true. And then we followed up with our Harmonic API and formalized his result.
- [149:26]
B
Interesting.
- [149:27]
D
We formalized it so that the machine actually validated its correctness. But you're starting to see this happening. And I think at first it's going to be in these kind of like obscure small results, like a Erdos problem or a little challenge here and there. But you could see that just like the length of autonomous tasks increases over time, the complexity of mathematical proofs is going to increase over time as well. So we're on the path, and a lot of people predicted it, but there was skepticism about can it do anything new? We're 100% going to get novel mathematical results that are significant. And what Harmonic can do with our API is make it so that when you have 1000 plus pages of AI generated mathematical proofs, it doesn't take a team of 10 mathematicians four years to validate the result. Verification is going to be the bottleneck and you're just going to want the machine to take care of that as well.
- [150:30]
B
You mentioned the API. Yeah, you mentioned the API. Is there a world where mathematicians need a cursor for math on top? Is there an opportunity for you to do sort of a copilot or an application layer product on top of the API, just to increase and get into the centaur era of the mathematician's day to day? What does enhancing their productivity look like if you're not just one shotting every math problem, but you're actually working collaboratively alongside a mathematician?
- [151:04]
D
Yeah, 100%. I mean, what we're building right now is very close to a Claude code for math. And if you think about a lot of mathematicians that are using Lean, which is the formal theorem prover, it's a programming language that is useful for math. They're not working on a chalkboard and chalk anymore. They're in VS code, which cursor is basically a fork of VS code. So mathematics and computer programming are converging into one. And it's pretty amazing for me, as a former mathematician, because people thought that math was like, nowhere close to being computerized in any real way. Like, for a while, the big challenge was how do I, like, type my math formulas so that I can actually, like, save them to a PDF?
- [151:56]
B
Sure.
- [151:57]
D
And we figured that out. But now real mathematics work is being done on computer for, I think, the first time in history.
- [152:04]
A
Everything is computer.
- [152:05]
B
Everything is computer.
- [152:06]
E
Yeah.
- [152:07]
B
You know what's not computer? Our last question, which is a very silly question. But everyone wants us to ask what is the biggest fish you've ever caught? Have you been fishing? We want to know if you've caught any fish and if so, what's the biggest fish that you've caught?
- [152:23]
D
Oh, let's see. My brother in law and I love going fly fishing in Montana. We've done some fly fishing on the Blackfoot river down there. And I'd have to say I'm not as good as him. I've caught trouts of all kinds of rainbow trout, bull trout.
- [152:47]
B
This big. Yeah.
- [152:48]
D
Trout maybe a little bit bigger.
- [152:50]
A
Okay, okay, we're off the screen. We're off the screen.
- [152:53]
C
We're off the screen.
- [152:54]
A
That's good.
- [152:54]
D
Well, I had some difficulty curling the trout, so it must have been at least 50, 60 pounds.
- [152:59]
B
Okay.
- [153:00]
A
Wow. There you go.
- [153:01]
B
Pretty big. There we go.
- [153:02]
A
Air horn.
- [153:03]
B
Well, thank you for for playing along with the silly fishing question. But we somehow got on fishing yesterday and we promised that we would ask all the questions what the biggest fish they've ever caught was. But thank you so much for taking the time to come chat.
- [153:17]
A
Great to get the update.
- [153:18]
B
Always a great time, Always a pleasure.
- [153:20]
D
Gentlemen. Fun hanging out with you.
- [153:21]
B
Have a great one, have a great rest of your day. Quickly, let me tell you about linear. Linear is a purpose built tool for planning and building products. Meet the system for modern software development, streamline issues, projects and product roadmaps.
- [153:33]
A
Up next we have Yuri from Overwolf. Overwolf is the leading platform for in game game creators with over 1500 supported games, 1 178,000 creators and 113 monthly active users. Overwolf enables creators to build, distribute and monetize in game apps, mods and private game servers.
- [153:54]
B
Waiting room. Let's bring him in the TVP at Ultradome. Welcome to the stream. How are you doing? What is your background? What is this beautiful office you're in? I'm fascinated by this.
- [154:04]
H
No, it's something like halfway through Covid I just could not work anymore from the house. So in our backyard we built this wooden shed.
- [154:13]
B
This is it.
- [154:14]
H
This is like my small man cave.
- [154:15]
B
This is amazing. And you have power and air conditioning in there, I assume?
- [154:20]
A
Yeah, yeah.
- [154:21]
H
I mean you have to have air conditioning for sure.
- [154:24]
B
I need this yesterday. I'm going to email you right after the show and get you to tell me exactly what happened because I truly am in the market anyway. We're not here to talk about cool backyard offices. We're here to talk about your business. Please give us a state of the union and give us any updates on how you're explaining the business, where it fits into the video game economy, and then how the business has grown recently because it's an absolute monster.
- [154:53]
H
Sure. So we're the guild for in game creators. And the way we think about our businesses, our responsibility in the world is we build a new profession. We call this profession in game creators. In game creators are developers, software developers who are building apps or mods or private servers or websites around existing games and making a living building those products for game studios that want to enhance their games with UGC, but don't want to go through the whole kind of length of building something like a Roblox or like a UEFA. On Fortnite, we're an engine that allows them to do that. So, you know, if you don't need.
- [155:26]
B
An engine, sorry, on Fortnite, uefn.
- [155:29]
H
Sure. So Roblox, I'm assuming folks understand how Roblox works in the ecosystem and everything and maybe have heard Dave earlier today, UEFN is like the creator program for Fortnite. So everyone knows Fortnite, right? But if you want to build your own Fortnite experience, you could go to UEFA. This is the engine, and you can build your own Fortnite experience. Makes sense.
- [155:55]
B
Yeah, makes a lot of sense.
- [155:57]
H
So what we do is we do that, that kind of same thing, but horizontally. I'm assuming most people know about Unity or Unreal, the game engine. So if you want to build a game cross platform, you can go use Unreal or Unity to build a game. You could use us as a platform to horizontally power these games with ugc. So this is what we're about.
- [156:19]
B
Where are the walls of the garden? I feel like every, every big platform has somewhat of an incentive to throw up walls. And this came up during the NFT boom, where a lot of people were pitching like, oh, well, you'll be able to take your custom skin in Roblox and bring it over to Fortnite. And whenever somebody was pitching that, I was just like, no, they won't. That makes no sense. They have no incentives. But I feel like you've been able to build a business that is somewhat cross platform. But there's probably places where the big game engine developers and the big platforms kind of don't want you playing. So how do you think about, do you actually call people up and just say, hey, can we do this? Or do you understand the economics? How do you think about where the. Okay, don't dip your toe in that particular Walled garden. Because where Company X wants to own that experience from start to finish.
- [157:22]
H
Right. So the gist is we respect the rights of the IP owners and we want to make sure that we're communicating with game studios, that we're sharing with them what's done in their ecosystem. If they want to go deep and they want to partner and they want to integrate our code within their game, they could do that. This would make modding available on Xbox and PlayStation, not just PC.
- [157:41]
B
Sure.
- [157:42]
H
But if they don't and they prefer to kind of keep an arm's length and just, just be aware of what's happening, that works for us as well. So I think the key is indeed respect the walled gardens. And what we're seeing, I think with time is this evolution of, you know, a decade ago or 15 years ago it was like not too many studios were open to folks creating content around their games. And now it sort of gradually becomes the new standard and you have massive IPs like Harry Potter who are all of a sudden open to UGC with all its risks.
- [158:11]
B
Right.
- [158:11]
H
You would think one of the biggest IPs in the world, probably pretty risky to allow creators to do whatever they want. But if you provide an environment that's safe, that does moderation, curation, make sure that the content is consistent with what the IP owner had in mind, you can have wonderful things happen. So the gist is we stick with the game developers, have conversations and make sure that the community supports them with whatever goals they have and their perception of the wall.
- [158:37]
B
How are you measuring impact of the business? Obviously run sort of a two sided economy. What are the key metrics these days?
- [158:46]
H
So we have one North Star, which is greater payoffs. The reason we chose that North Star is we're building a new profession. So it kind of makes sense. We want to make this whole thing sustainable, but because we're UGC and we're freemium, so it's all like free to play pretty much. If there's no money, there's no quality, there's no product market, there's no retention, there's really nothing. So our one core KPI is creative payouts. And yeah, that's the key one. Obviously we look at a bunch of other things, but this is the main one.
- [159:13]
A
How many people globally have a full time job contributing to video game economy? As in independent developers or small teams that do this full time?
- [159:29]
H
You mean in the UGC space or just in general? Do you want to count Roblox and uefn or games?
- [159:35]
A
Yeah, I guess Globally. And then if you can kind of separate it out between all the different kind of categories that you just laid out.
- [159:45]
H
All right, so let's do a bottoms up analysis with the different categories. So obviously we have studios and publishers, put them in one category. And this includes the Blizzards and ea, but also the smaller indie studios that would then go ahead and distribute their game on Steam or on one of the mobile platforms. So I'd say this is one category of folks making a living in the video games industry. I'd say the second category is called UGC Creators and this includes Roblox developers, UEFN Developers UEFN the Fortnite example that we've about talked just given. And also folks that are creating with us. I'd say this is the second category. People who are not necessarily building the fundamentals of the game, but they're enhancing an existing game or an existing ecosystem. Now, obviously there are a lot of service providers like, you know, studios that would do work for hire or would help with QA or would help with localization. So this would fall into the service providers category and then I would go to infrastructure. So you have folks like again, Unity or Unreal or an engine like, I don't know, Speed Tree that is doing planting and shadows and I don't know, a bunch of tooling in the ecosystem. And then you have the go to market. Right. We have influencers. You have a lot of folks just doing, say Minecraft content on YouTube and they make a living off of that. So I don't know off the top.
- [161:15]
A
Of my head with the number of.
- [161:16]
H
People who are doing that and make.
- [161:18]
A
A living, but it feels like kind of best possible time. If you love video games, you can just live anywhere in the world and probably if you try hard enough, you can make a living kind of contributing to these different ecosystems.
- [161:34]
H
Yeah, but I do want to emphasize it's, you know, just like on YouTube, there's a small percentage of creators who are actually making a living building content. Yeah, same dynamic, same thing across the board. Yeah, like, you know, it's very easy to enter like a Roblox and start creating, but there's only a small percentage of folks that are actually successful and have recurring success and can make a living doing that. And we're seeing that within our own ecosystem.
- [162:00]
B
How much have you actually paid out to creators?
- [162:05]
H
So today it's in the neighborhood of 800 million. With the last 12 months of 25, it's 300 million.
- [162:13]
B
Wow.
- [162:13]
A
Wow. You got a gong to hit for you. Those Are. That's an insane number one, by the way.
- [162:23]
H
Another thing I want to emphasize is we started back in 2010, so I'd appreciate an overnight success.
- [162:30]
B
Oh, yeah?
- [162:33]
A
Yes.
- [162:33]
B
Remarkable. Yeah. And 150 million.
- [162:37]
A
How big is the team?
- [162:39]
H
It's like 240 now.
- [162:41]
A
240?
- [162:42]
E
Yeah.
- [162:43]
B
Wow. Well, congratulations. I mean, what a remarkable. What a remarkable company. Very exciting.
- [162:49]
A
Yeah. What is, you know, you've been at it now for what, over 15 years? Something like that. What is? People ask us all the time, what's the future of tvpn? And we're like, more of this. But I'm curious what your answer is.
- [163:13]
H
I started this company when I graduated from computer science and I started as a modder. Our strategy was not platform when we started. We were basically modders building both the engine and the content ourselves and we weren't successful. Mid 2013, we ran out of money and we had to figure out what to do. And only at that point we pivoted and we've committed to just building the engine and becoming a tools and services company versus building both the mods and the engine to build this. And I feel very fortunate waking up in the morning and knowing that there are people like Sandy in San Diego making a living building mods. And we're a big part of that and enabling that. I feel like that's a life worth living for. So what's next is just more of this. Our target is we want to get to a billion dollar for creators a year. But when we get to that, it's not like the KPI is going to change. It's probably going to be, all right, let's do 1.5 or maybe 2 or like, how else can we contribute to creators? This is sort of what juices us up in the morning. This is what we intend to continue doing.
- [164:15]
B
It's awesome.
- [164:16]
A
Last question. Because we didn't follow it super closely, what's going on in Counter Strike World and what's kind of the takeaways there? I know they made some. Some changes to their in game economy and a lot of people were pissed off, but I don't know a lot more than that.
- [164:35]
H
So you're right. There's a huge skin community around cs. Go Counter Strike. I guess the challenge is, and we're sometimes seeing this with studios, they make their own decisions based on what they think is right and based on what they think is right for the game. And once they make a change to the economy, obviously a lot of people lose a ton of value from their skins and As a result, they lose a lot of money. I'd say that in advance. Valve did not provide any guarantees that this is a thing that should happen. I don't know if they've intended for skin trading to kind of be as big as it was and they can decision to change. And sometimes as people creating in the ecosystem or working around the ecosystem, we're at the mercy of the IP owners and the IP owners have the right to do that. I don't know exactly, exactly what were the reasons behind the change, but those are some pretty smart people. I'm sure they have their good reasons.
- [165:31]
A
That makes sense.
- [165:32]
B
I'm checking the price of my M4A4 skin on csgoskins.com I think I'm down bad. I don't know. But I did buy it down 70% over a decade ago and it brought me a lot of joy when I played CS Go back in the day.
- [165:47]
G
Priceless.
- [165:47]
B
And I haven't logged in in five years or 10 years. So I think that I'm happy. I see it as a tip. When I bought it, I saw it as a tip. I was like, this game's free. I play it a ton. I'm happy to throw some money in. I understand that people do take it way more seriously and they even build whole businesses around it. And so there's a lot that goes on there. But for me, I feel like even if it's worth zero, the memories will stay with me forever of holding down Dust two on the Bee Bomb site. It's good times. Anyway, thank you so much for coming on the channel.
- [166:20]
A
Congratulations on all my milestones. Come back on when you hit a billion.
- [166:23]
B
We'll talk to you soon. Have a good one.
- [166:25]
H
Sounds good.
- [166:26]
A
All right, cheers.
- [166:27]
B
In other news, Numeral has a dot com now numeral.com sales tax on autopilot. Spend less than five minutes per month on sales tax compliance. We got to hit the gong for.
- [166:40]
A
Numeroal.Com that.com is too good. Also too good.
- [166:47]
B
ESPN. ESPN. BET is shutting down. ESPN and Pen Entertainment's 10 year $2 billion deal will end eight years early after the sportsbook saw lackluster demand. ESPN is now partnering with DraftKings. The interesting thing here is that everyone draws that Disney, that Disney. Oh, Walt Disney drew on the napkin. All the different interplay between everything. They didn't include gambling on there because Disney was gonna get into, you know, espn, bet. That's what happened. They did update the Disney chart, update the Disney market, the ecosystem chart and also go to Fin AI the number one AI agent for customer service. Number one in performance benchmark, number one in competitive bake offs. Number one ranking on G2. Our next guest is already in the Restream waiting room. We have Alex Israel from Metropolis with some very good news. Good to see you, Alex, how are you?
- [167:39]
A
Good to see you again.
- [167:40]
B
Welcome to the show.
- [167:40]
I
Good Jordy. John, good to see you.
- [167:42]
B
Good guys, thank you so much for hopping on the stream. Give us the news. What's going on?
- [167:49]
I
Busy day here in New York. We just raised $1.6 billion.
- [167:54]
E
Whoa.
- [167:59]
D
That is.
- [168:00]
A
Congratulations on saying the biggest number.
- [168:02]
B
Said the biggest number. Congratulations.
- [168:06]
A
Not easy, not easy. Easy to do, but today you got it done.
- [168:11]
B
Okay, but. But it's 500 million Series D equity, 1.1 billion term loan. A lot of people are getting jitters. They're asking for backstops. How do you manage a billion dollars in debt? Debt sounds scary. A lot of startups don't use debt. How. Why are you confident that debt makes sense in this, in this world?
- [168:31]
I
You know, I think for us it's all about the success of the platform. We've just scaled to such a significant rate. We're doing now over $5 billion of payments. We have over 20 million members on platform. It's given our scale, given our free cash flow profile, we're confident we can service the debt and we're confident that we continue to scale into new verticals.
- [168:49]
B
Do you think about matching the capital structure to the shape of the business? Is it important to have debt backed by physical assets or something? How do you think about the financial engineering that you're doing?
- [169:01]
I
Yeah, it's a good question. For us, it's all about the lowest cost of capital, the lowest dilution with the right partners that can facilitate the strongest level of long term sustained growth. For us, it's all about growth.
- [169:12]
B
Yeah. And what are the growth levers right now? Because some people, when they want to grow, they go film a Vibe reel. I imagine you're not filming $1.6 billion worth of vibe reels and trying to go viral every day. This isn't the clue that point by Vibe coding. All we're doing is Vibe coding Vibe reels. But I honestly would imagine that $1.1 billion, that's not even going towards R and D because that's something that if there's risky R and D, you'd want to be funding that with equity, walk through some of the growth levers that you're pulling over the next couple of years.
- [169:43]
I
So I think Jordy and John One of the last times we connected, we were pioneering the growth buyout. This idea that we could take a legacy old world business and we could acquire it directly by a traditional technology company. Right now we're pioneering something brand new which we're qualifying as the recognition economy. This idea that we can leverage AI and computer vision in the real world to drive seamless payment and commerce experiences everywhere. And we've proven that with parking. Now we're scaling into gas stations, car wash, quick serve retail. Soon you'll see our technology moving with you from a biometrics perspective into hotels and into other environments where you can facilitate seamless access and payments.
- [170:21]
B
So when you go and buy a company, obviously there's an established business, so that's accelerated. Then, then you're trying to use AI to accelerate growth and cut costs. How do you think about what you actually do as you build a portfolio?
- [170:36]
I
So we don't really look at it from a cost cutting perspective or a cost synergy perspective. It's all about revenue synergy. It's all about how do we drive gross profit and it's all about how we work with our partners to drive revenue to their businesses. So it's all about revenue synergy. Everything for us is about sustained revenue growth. I think too many companies think about cost synergies, too many private equity firms think about cost synergies and short term gains. For us, it's all about long term gains driven through revenue.
- [171:03]
B
Do you think too many private equity firms think like that? I feel like there's world for both. Like I've studied, like we've talked about private equity like titans. People have been in the business for decades where their whole model is, you know, they take a big company private, change the business model, grow it a touch, IPO it again, then there's sort of the bottom feeders that are picking up scraps and cutting costs even more. And I don't necessarily know that every company is built to grow forever. I actually am somewhat receptive to the idea that some companies like they've just done their job and they just kind of need to go out to pasture and retire and the company's retiring. And so I think that that type of private equity guy gets a bad rap. But again, I'll defend every, every private equity guy because I have, yeah, he came on the road.
- [171:50]
I
I love private equity as well, but I think the capital markets has a natural inclination to operate within boxes.
- [171:55]
B
Sure.
- [171:55]
I
And once you go outside that box, once you create a model where you're actually taking a Technology business and acquiring old world businesses that doesn't fit in a traditional box. And you have to take creative investors. You have to take best in class investors like Todd Boley, Tony Manila. You have to work with players like Aryeh Burkhoff that can think outside of the box and think how they can mitigate or create what I'd say is an asymmetric return profile. So all the downside risk mitigation of a traditional private equity play with all the upside potential of a traditional venture.
- [172:25]
A
Capital play, I have to double click on. You talked about biometric payments in retail, quick service hotels. I think you said fueling. Fueling. What does that actually look like in practice today? What are some of these?
- [172:44]
B
Biometric payment is when you're checking out and they take a pint of your blood and they run it through a centrifuge. A little more than a bank, a.
- [172:51]
A
Little more of a pint.
- [172:54]
I
Look, think about how much of our lives we spend wasting time checking out at a grocery store. Think about every time you go to a meeting in New York who are waiting to get into a class A office building, showing your ID to a stranger, giving all your privacy information over that actually ID has your address on. Think about a future where you just belong, where you can seamlessly walk into any class A office building. Think about a future where you go into your coffee shop and they just know what you want because you always order the same cup of coffee. So how can you leverage computer vision to create a new paradigm in privacy and reduction of fraud through leveraging biometrics? And that's exactly what we're doing.
- [173:30]
B
I feel like this kind of. Already you see glimpses of this in, in China and Japan and different Amazon's experimented with it.
- [173:38]
A
What do you think the key to success will be in terms of we have the technology now.
- [173:45]
B
Yeah. My question here is do we need the technology, the models, to get better? Is it an implementation problem or is it you need to buy a series of services and just implement them effectively?
- [173:59]
I
I think more than anything, you need to build trust and a dynamic with a consumer base. Right now we're crossing 20 million members on our platform that are leveraging our technology every single day to facilitate seamless access and payment at over 4,200 locations across the United States. It's that trust, it's that empowerment, it's that seamless access that people want to come back to time and time again so they're no longer wasting time. And if you have that network and if you have that flywheel I think that's the core ingredient in order to scale into new verticals.
- [174:31]
B
Are you on a collision course with Clear at all the biometrics? I don't know. There's something there. The chat was asking.
- [174:40]
I
No, it's a good question. I'll tell you. I think Clear has done a remarkable job at a very specific use case. And I think Karen and their team have built a remarkable product. And I think that we're very focused on mobility right now as it pertains to gas stations, car wash, quick serve, retail parking. Anytime a transaction is facilitated in a vehicle and then we're moving past the vehicle. But I think what that team has done is really unique as tied to identity. And we're scaling into additional and differentiated verticals.
- [175:10]
B
Yeah, differentiated verticals.
- [175:11]
G
Different problem.
- [175:12]
A
Yeah. Very cool to think about a world where you can drive through. Drive through a restaurant. Not or just.
- [175:22]
B
I like Ryan here in the chat who says, I will make it impossible for AI to ever guess my coffee order. He's gonna keep the clankers guessing.
- [175:32]
A
But a world where you can drive, drive through a drive through and just be like, this is what I want. And then you just drive forward and they hand it to you and you never have to do the whole like, you know, mobile payment, you're reaching out of your car or whatever. So imagine a world if it just.
- [175:45]
I
You know, could be precognitive, knew what you wanted before you ordered it.
- [175:48]
B
It pretty fast.
- [175:50]
A
Yeah, we'll work for everyone except for Ryan in the chat. But thank you. Thank you so much for joining. Massive milestone for you and the team.
- [175:59]
B
Congratulations.
- [175:59]
I
See you guys in la. Talk soon.
- [176:00]
B
See you.
- [176:01]
A
Talk soon. Cheers.
- [176:02]
B
Let me tell you about Adio.com customer relationship magic. Adio is the AI native CRM that builds scales and grows your company to the next level. And get started for free. Also send me an update. Analysis has launched ClusterMax 2.0, the industry standard GPU cloud rating system. And the number of AI clouds has exploded. We are seeing a fast takeoff in data center construction and Neo Cloud development. Very clearly fascinating newsletter. 200 page document breaking down how all the different clouds fit together. They tested 209 providers. They rated 84 of those providers. They have 140 customers that they surveyed. And the article is 40,000 words. I would encourage you to go read it. Sign up for semianalysis or tune in to TVPN on Monday because we will be interviewing the semianalysis analyst who wrote the piece. Until then, we'll bring in our next guest, Paul from FOMO Benchmark's latest investor.
- [177:07]
A
What's happening?
- [177:07]
B
How you doing Paul? Good to meet you.
- [177:09]
A
Welcome to the show.
- [177:10]
G
Great to meet you guys.
- [177:11]
B
Welcome to the show.
- [177:12]
A
Big day.
- [177:12]
B
Introduce the company, give us the elevator pitch that you gave to Benchmark in their elevator.
- [177:20]
G
Yeah, it's a great question. I think there's a misconception about FOMO that it's a crypto trading app. And right now all of the assets currently available are digital assets and a lot of attention based assets. But the goal in the long run is to be the first largest financial platform ever built natively on crypto Rails. And I think that there's a rare opportunity here. The only other time I could think is the early Internet where 100 years of infrastructure was created. Amazing infrastructure. Everyone's trying to build on new infrastructure, but it's time for real users and I think FOMO's building the bridge to allow for accessibility to these new markets and so on FOMO you could download the app, you can onboard with Apple ID and second you could buy digital assets with Apple Pay and Core to the feature of FOMO is the social features and these are uniquely enabled by onchain. So on FOMO you can basically have full visibility into people's balances and discover new assets in real time.
- [178:22]
B
It's very interesting that the last deal that Benchmark did was chainalysis because both of these like you don't have your own token yet, is that correct?
- [178:31]
G
Yeah, that's correct. And there's no plans for a token.
- [178:34]
B
There's no plans for a token. So this is something where it feels like if I'm reading between the tea leaves, like Benchmark doesn't want necessarily the underlying exposure to a fluctuating asset. They want a business that's. But they're, they're interested in businesses that are built on top of crypto Rails. And that's where you come in.
- [178:49]
E
Correct?
- [178:50]
A
Yeah.
- [178:50]
G
I think Benchmark is just incredible builders of consumer products and I don't think, I think we're really thinking about the exit right now. We're trying to build generational company to compete with the behemoths of the industry. I think that 10 years ago it made sense to have a centralized exchange where you have to kyc to onboard where you have to do all these checks where it takes three days to deposit from a bank account. Right. But you also only have access to about 50 to 70 to 90 digital assets on, on most of these exchanges. On FOMO when you download the app, you have access to millions of assets instantly and I think this is where the future is headed. And what's beautiful is if you guys went on the app, I can go to Jordy John's profile and I could see every trade you ever made on any asset. So right now it's mostly digital assets. But eventually you'll have equities, you'll have real world assets, you'll have prediction markets bonds, you'll have FOMO savings accounts where users can earn on chain yield, which is super hard to access and everything else.
- [179:50]
A
What's, what's the pathway to equities as an example? Like, you know, this is something a lot of people are excited about, but I'm curious, like what, what the practical path is to on chain equities. Some, some of they sort of exist in various ways already. Is that, is that correct? But what. I guess, I guess doing it in a way that is compliant.
- [180:19]
E
In a.
- [180:19]
A
Regulatory environment that's kind of constantly evolving.
- [180:23]
G
Yeah, I think it's inevitable that everything gets commoditized into digital assets. Right. It's just the most efficient architecture for value transfer that's ever been created. So given that given the ability to permissionly access, the ability to hold stuff in self custody, a lot of institutional players are starting to create equities on chain. But I think with the market structure bill and other things in Congress and over time you'll see more clarity on native digital asset equity issuances. Right. But in the interim it is wrapping a lot of these traditional equities on chain movers like Robinhood, Kraken and a lot of these big players are putting equities on chain and the goal is to access all of those permissionlessly in decentralized non custodial wallets. The problem is historically that's impossible. Right. It's really hard to do these things. If you want to onboard to any of these AMMs or these on chain protocols, you need to onboard to a centralized exchange. It takes three days to get funds there. Then you need to download a decentralized wallet, then you need to create and save your seed phrase. Then you, then you connect to the exchange and there's tons of fraud. There's not great fraud warning and users are just getting rugged. And then you go to a block explorer and the user experience is terrible. So we're trying to take that whole experience and make it super simple and then have this social layer where people can discover new assets really easily.
- [181:44]
A
How are you thinking about building the team? Feel like there's a lot of pressure in crypto, obviously. Jeff at Hyper Liquid has kind of set the bar on scale to team size. But how are you guys thinking about it?
- [181:58]
G
It? Yeah. So our team is about 10 right now and this is by far our competitive advantage. Our team is incredible. All ex builders of protocols like DYDX, Uniswap, OpenSea that paved the way for consumer crypto. And it's the missing piece of the puzzle here is most teams have been either really good at building trading infrastructure, which is a really hard problem. Right. You need to have a router aggregator, potentially build your own routing. There's tons of mev, which is people taking spread on user when they are trying to execute on chain. But then you also need people who can build an amazing user experience. Right. And those two pieces matched is what's been missing so far. So I really believe this is the first accessible platform for digital assets on chain that has ever existed that's really accessible to the mass market. We launched six months ago. We've already onboarded 120,000 users. We have added many of those users to our platform recently. We're doing 20 to 40 million dollars in volume A day. And currently Solana on chain volume. Thanks. And currently Solana on chain volume is a $320 billion market, meaning we're doing 0.1% of that quarterly volume. And that's just Solana. So our goal is to add all of the chains and dominate the whole on chain spot trading market in the next six months.
- [183:19]
B
That's fantastic.
- [183:20]
A
Ambitious. Love it. Let's give it up for this. Let's give it up for the sunset in the back too. That thing is just absolutely.
- [183:26]
B
How much did you raise?
- [183:28]
G
We raised $17 million.
- [183:29]
A
There we go.
- [183:31]
B
I was gonna say Vlad had a.
- [183:33]
G
Nice camera, but I got the nice sunset.
- [183:35]
A
I mean, that's. That's insane. That's insane.
- [183:38]
D
You're.
- [183:38]
A
You're in New York.
- [183:39]
G
Yeah, in Soho, New York.
- [183:41]
B
Is that. Is that a real sunset or like virtual background?
- [183:43]
G
It's a real sunset. It's actually really incredible right now.
- [183:45]
B
It's just.
- [183:45]
A
It's just real deal.
- [183:47]
B
Well, thank you so much for coming on the show.
- [183:49]
A
Yeah. Congratulations on the round and I'm sure we'll have you back on soon.
- [183:53]
B
Have a good one.
- [183:54]
G
Awesome.
- [183:54]
A
Thanks, guys.
- [183:55]
B
Bye. How'd you sleep last night, Jordy? Back in your 8 Sleep? You can go to 8sleep.com, get a pod 5 code.
- [184:04]
A
I've just settled into the 80, 80, low 80s.
- [184:07]
B
Low 80s because I was in the high 80s. I got an 87 and play how.
- [184:11]
A
Many hours for me 6 hours 44 minutes. I got 7 hours 11 minutes from.
- [184:18]
B
Beacon is in the Restream waiting room.
- [184:20]
A
I'm excited for this one.
- [184:20]
B
Neelam. He's the ex president of Instacart. Welcome to the stream. How are you doing?
- [184:25]
A
What's happening? Welcome to the show.
- [184:27]
B
Thanks so much for taking the time to talk to us.
- [184:29]
F
So excited to be here. Big fan, excited to chat with you guys.
- [184:32]
B
Thanks so much.
- [184:33]
A
Great to have you.
- [184:34]
B
Get us up to speed, Give us production on yourself and the company, please.
- [184:38]
F
Yeah, we, we, my co founder Divya and I started the company a year and a half ago. What we are is a AI first holding company. We acquire, operate and then we grow niche vertical market software businesses that serve under under looked at markets. We call them Main street businesses. So think about campground operations, K to 12, private schools, construction companies, things that usually don't have large IT departments. They're not the first place big tech goes to bring technology and innovation to. Divya was leaving Sequoia, I was leaving D1. We both were thinking about ways that AI was going to transform Main street and that's what got us together. And we felt like access to the best technologies usually doesn't diffuse evenly. And we thought the easiest way to get the best technology in the hands of end customers on Main street is by partnering with entrepreneurs that already run these businesses, already have these relationships and then turbocharged them with our team tech process. And you know, having been there and done that and scaling businesses in the past. So that's a bit about what we're about.
- [185:53]
A
Yeah, very cool. Where, where can companies that you've acquired to date or plan to acquire get the most leverage out of AI so far? We had Vlad on from Robinhood earlier. He was talking about getting leverage at the CX in cx he was also.
- [186:11]
B
Talking about fly fishing.
- [186:12]
A
Well, yeah, but getting leverage in CX getting leverage in engineering. What are the other kind of buckets that you think AI efficiency is, is, is real today? Because we've been having this debate. You know ChatGPT is an amazing product but at least in our day today and I think many of our of our peers, it's not exactly like replacing a full time hire at this point.
- [186:36]
F
Yeah, yeah. I, I think Vlad definitely can use some of our products to find a spot to, to fly fish. We own three business and campgrounds. But, but other than that.
- [186:46]
B
Yeah, yeah, truly.
- [186:47]
A
Okay, so, so yeah, perfect example because I feel like some of these categories just looking at the business and saying how do we build software? Not Even AI to make them more efficient. There's already efficiency that you could create. But yeah, where are they really getting an edge?
- [187:09]
F
Let me give you that one. So let me give you the campground example since we're talking about it. We partnered with an entrepreneur who was building a business out in Saskatoon in Canada that served private campgrounds. There's 18,000 of them across North America. He built great software to allow campgrounds to take reservations and you know, manage bookings, manage space allocation, all of that stuff. We bought the business, we Mike runs it for us. We tend to work with our entrepreneurs and they continue to stay with the business after we acquire it IT and the first thing we worked on with him was how can we apply AI to help him find new campgrounds. And this sounds like a very basic problem, but most of the businesses we buy do no outbound at the point that we buy them. So they all of their sales happen through word of mouth and referrals. They tend not to have what we know in Silicon Valley, the formal go to market motions. So we built some in house tech to be able to first find Mike a list of 18,000 campgrounds, then score them, do some lead scoring, some multimodal communication like when do you text them, when do you call them, when do you email them, when do you LinkedIn them? Many of them are not on LinkedIn and within the first year of running that business, it's going to grow more than 80%. So that's an example of AI on the go to market side. The second thing we did with Mike is we were doing product roadmap review and one of the features he had built was, was a lot of his campgrounds actually had boat slips and marinas attached to them. So he built a feature so that you could actually kind of reserve a spot in a marina within a campground operation. And I knew nothing about the space, but we started helping Mike. We put someone on the project with him to scope out and do calls with marina operators. So we found him a bunch of marina operators, figured out their tech needs and then we gave him all of our companies get Cognition instances. So we gave him and his development team Team Cognition. Our forward deploy engineers worked with him and then we launched a brand new product called letsboat.com so like a brand new product, we built the UI Cognition and our developers built the product and now we're live with that, which is a, you know, kind of a net new company within, within our campground operations. And we, you know, when I was building Instacart, that whole Process probably would have taken us a year. We did it in like a month and a half. And I think, you know, AI for code development is just table stakes now. So I don't think it's a competitive advantage. But the matching like customer needs customers who are in Main street that no one's talking to with tech. That's where I think Beacons magic comes in.
- [189:57]
A
Yeah, it's interesting that at least the edge today seems to be. There's a bunch of businesses that you can now build software for that previously would have not been able to justify engineering hours that suddenly can. I mean, we experienced this. Tyler on our team builds a bunch of internal software that we use to run the show more efficiently. And you know, this was, I don't know, 10 years ago. We probably wouldn't have been able to justify it. But he can just be so productive because of that. What about some of the other categories I'm curious with? We've talked about Alpha School on the education side, using AI in the curriculum. What are the ways in which you can get efficiency in education at kind of like a business level or are you thinking more at the actual curriculum level?
- [190:49]
F
It's probably more today at the business level. So we partnered. We have a number of software companies in the education space. I'll bring up one because I was talking to their customers earlier today. We work with a company called College Kickstart. It's based in California and their customers are K12 private schools, specifically the college counselors in K12 private schools. And they vacuum up a bunch of data and signals publicly available to be able to help College counselors in private schools work with parents, high anxiety parents and students through the matching process of, of hey, where should my kid apply? Where should they apply? Early decision, how should they help present themselves to the kind of target schools, et cetera. And really at the heart of it, that's a data problem and it's a computation problem. So we partnered with College Kickstarter. It's one of our newest businesses. It's a situation where the general manager was retiring, so he's going to be transitioning out over the period of this year. We brought in a great general manager from Whatnot and Uber and his name is Tom. He was previously at Teach for America. Tom's going to work with the old founder and the CEO to be able to build net new algorithms to help college counselors make better decisions for their students. And that's like I have a 12 year old daughter. The anxiety that goes around where your kid's going to end up in, in college is like a really big deal even for a 12 year old.
- [192:23]
A
Right.
- [192:23]
F
So something like that. I think AI can add a ton of value.
- [192:26]
A
That makes sense. How much is traditional private equity pitching AI to their LPs to your knowledge, I imagine.
- [192:38]
F
You know, I think a ton. Yeah, honestly, I think a ton. And look, I think we, Divya and I very purposely designed Beacon as a holding company. So we buy our businesses, not to sell them. We want to hold them forever. And there's a couple of reasons for that. First, there's like a financial benefit to duration and reinvesting the cash flows that come from these businesses, either within those businesses or in other businesses that we own. But the real benefit is, as we thought about the AI transformation, I don't think anyone really knows how long all this stuff's going to take. People have great ideas. I was listening. You had Brett on earlier. He has great ideas. There's tons of of really smart people with great ideas. But my best guess is no one can handicap how long it's going to take for the like full transformation to happen. So we want to have patience and we want to have duration to be able to keep improving our products and keep building net new stuff for our customers. I don't want to be held to having to sell these companies in three years or five years or seven years if that's like the moment that the magic's really going to start to kick in. Private equity can't make that promise to founders. So what they end up doing is. And not that there's anything wrong with this, their model tends to be buy a business already with a plan of how to exit it with the underwrite that's based on a bunch of cost takeout, which they do in the first 18 months. Then they start dressing it up for sale in the second 18 months. And increasingly smart founders, the ones we work with, that's just not the model for them.
- [194:08]
A
Yeah, that's a huge advantage for you guys. Well, super exciting, massive milestone and you're officially our AI roll up correspondent now. So I love it. Expect a call as more of this stuff is in the news.
- [194:23]
B
Yeah, yeah. What an amazing story. Thanks so much for coming on and break it down.
- [194:26]
F
Thank you guys. Take care.
- [194:27]
B
Great to meet. We'll talk to you soon.
- [194:28]
A
Cheers.
- [194:30]
B
Let me tell you about adquick.com, out of home advertising made easy and measurable. Say goodbye to the headaches of out of home advertising. Only Adquick combines technology out of home expertise and data to enable efficient seamless advertising. Bang across the globe. Elon Musk's pay package has been approved. Tesla shareholders approve Elon Musk's $1 trillion pay package. For reference, it received 75% of the vote. The $1 trillion stock award is tied to business goals. It nets Musk 878 billion. I guess. Board chair warned Musk could leave if pay was rejected. Many unions and some investors called the payout excessive. I'm sure there will be pushback on this, but if he does, it hits the milestones. You kind of got to give him some credit. And what better way to give him credit than with $1 trillion. He's at half a trillion right now. He has 500 billion to his name. This would be, would more than double it. But he's within spitting distance even without this.
- [195:35]
A
Yeah, this also allows him to have I think 20% of the voting power, 25% of the voting power at Tesla, which is something he's talked about publicly around building this robot army and wanting to make sure that he can't be terminated for political reasons or other reasons. If he's going to go and scale.
- [196:05]
B
Look at the way this package rolls out. So the operational milestones are more interesting than the market value milestones in my opinion. We all know the goal is to get to 10 trillion in market cap, which is unprecedented. But at the same time Nvidia is half the way there already. So if you can just build a company that's two Nvidias, you're good to go. But the operational milestones are interesting because they tell more of a story in my opinion. The first is 20 million vehicles delivered, then 10 million. The website just updated. Here we go. 10 million active full self driving subscriptions. Then 1 million robot deliveries, 1 million robo taxis in operation. And then you get to the EBITDA and you want at the end of this 400 billion in adjusted EBITDA to really get up there in those tranches. So you can't just pump the stock. Now the stock is pumping. It's at the highest level of price to equity since 2018. So Tesla shares are at $447 each, which is 230 times earnings per share. And analysts see the automaker reporting over the next 12 months. So it is at a high price to earnings ratio. The S&P 500 for comparison is trading at 23 times expected earnings. Nvidia, at the center of the AI revolution is valued at 32 times earnings, while Microsoft has a PE of 30. And Meta Platforms is lagging at just 21. Tesla, on the other hand, 10 times the price to earnings ratio.
- [197:51]
A
Absolutely wild.
- [197:52]
B
Well, one reason why it might be so highly valued is because Tesla shareholders are optimistic about Optomis. The humanoid robot. Elon Musk called Optimus an infinite money loop.
- [198:04]
A
You love that on the call just now.
- [198:06]
B
Oh, yeah, he said. He predicted other uses for the Optimus robot. But Musk declared Optimus is an infinite money loop. In the future, there may not be any money. He mused. Don't sell your shares. He added, I guess what I'm saying is hang on to your Tesla stock. I love it. Let's go.
- [198:25]
A
Not financial advice.
- [198:26]
B
We have to hit the gong for Elon Musk.
- [198:28]
A
Let's hit it. Sorry to Alex from Metropolitan. He thought he had the biggest number.
- [198:37]
B
Of today's show, but Elon came in from the top rope also. Elon, you got a trillion dollars. You got half a billion. Half a trillion already. Head over to getbezel.com buy every single watch they have on sale. They have 25,000 luxury watches fully authenticated in house by Bezel's team of experts. Elon, you could buy all 25,000 today.
- [198:57]
A
26,000.
- [198:58]
B
All 26,000 today. You could also book every single wander forever, Elon, so why don't you do that as well? Head over to wander.com. find your happy place. Find all of the places that will make you happy. Imagine just having a network of every wander permanently reserved for you wherever you go. Land, land.
- [199:19]
A
Go long. Land.
- [199:20]
B
We're going book a wander with inspiring reviews, hotel grade amenities, dreamy beds, top tier cleaning, 247 concierge and it's a vacation.
- [199:25]
E
No, but better.
- [199:25]
A
On that note.
- [199:26]
B
Yes.
- [199:26]
A
We got to get on with London.
- [199:28]
B
We do.
- [199:28]
A
So thank you for tuning in to the show today and we will be back tomorrow. We have a number of very cool guests. We have Mikey from Suno is going to be coming on. We've got Colin and Samir. We have Katharine Boyle from Andreessen joining as well. So I cannot wait and we hope you have a wonderful evening.
- [199:49]
B
Goodbye.
- [199:51]
A
Bye.