← The Rundown
The Rundown
Is OpenAI's Sora Launch a Sign the AI Economy is Cracking? (ft. Kyla Scanlon & Alex Heath)
00:27:44
Loading summary
Transcript68 lines
- [00:00]
A
Welcome back to the Rundown, one of the best business podcasts in the world. My name is Zaydad Mani and in today's show we are talking all things AI with returning guest Kyla Scanlon and first time guest Alex Heath. Many of you guys are probably familiar with Kyla's work. She does an amazing job breaking down things impacting the economy on her substack and Instagram account. And Alex is one of the best tech reporters in the world. I've been reading his stuff for years now. He has a reputation of getting the major scoops and he recently launched his own newsletter called Sources, which we'll link in the description. In today's conversation, we touched on a ton of stuff ranging from OpenAI's big week, what executives inside the company feel about the rise of AI slop if all of this is a bubble, and why the US government is unlikely to stop the AI train anytime soon. This was an awesome and wide ranging conversation. Both Alex and Kyla are fantastic. They're really smart people. They really know their stuff. So I hope you guys enjoyed this conversation. All right, guys, we have Kyla Scanlon back on the show today. You guys know Kyla, incredible economic writer, observer of macroeconomics. Kyla, thank you for coming back on the show today. And we have Alex Heath, firsttime guest. Alex is a tech reporter, the founder of Sources. My man has all the tech scoops. Thank you for coming on the show today, Alex.
- [01:23]
B
Yeah, thanks for having me.
- [01:24]
A
Of course. I think we have to start with the big news of the week, which was OpenAI. I mean, OpenAI has all the juice these days. They are dominating headlines regarding the markets even though they're a private company. They launched a bunch of features this week with Chat GPT trying to turn Chat GPT into an operating system. They got the Sora app that's going viral. We're going to talk about all that stuff. Alex, I want to turn it to you first. You were actually at their Dev Day on Monday in San Francisco. Can you just walk us through like what the vibe was like at the Dev Day and what was your takeaway from. From that big event?
- [02:01]
B
Yeah, I mean, open eyes, crushing it on vibes. I would say there was packed as a lot of developers, everyone sees them as maybe like the next big platform. You know, you had like Facebook doing its platform stuff 10 years ago where it was going to become this distribution service for developers, the App Store, obviously, even before that. And I think ChatGPT hitting 800 million weekly users and them now letting developers build their app experiences inside ChatGPT where you don't have to leave to use Zillow, Coursera services like that is really exciting for developers. So, yeah, I would say the vibes were high. A lot of AI bubble talk as well. That was, there was that kind of underlying current of fear about that. But the product side, a lot of excitement.
- [02:48]
A
You know what I. My biggest takeaway as an investor was that they were announcing all these partnerships, right? They were announcing like features with Figma and Shopify and Etsy. And anytime they would announce these partnerships with a logo of the company would pop up on screen, the stock price of those companies would just shoot up. And it just kind of tells you, like, the appetite that investors have to get exposure to OpenAI because you can't directly invest in them and they just want to get exposure to it any way they can. But that kind of, that kind of leads me to like the bubble question. Kyla, I want to talk to you about something that you wrote about this week, which was a fantastic piece about AI. You talked about how now that OpenAI is launching Sora, which is like a social media/AI slop app, that this could be a sign that maybe the use cases for like the revenue opportunities for ChatGPT and the traditional AI applications might not be what was expected. So now they're turning to like brain rot to get to make money. Can you, can you talk more about that?
- [03:47]
C
Yeah, I mean, I felt very conspiracy theory writing that piece because the whole underlying thesis was that AI is obviously propping up the stock market. Like Nvidia, Microsoft, Meta are a huge component of the stock market and the returns that we've seen. AI is also propping up the economy. It's been a huge component of economic growth. And I think for the AI companies, it's kind of like, okay, so we're building out all of these things, but how do we, you know, pay back all the money that we've been spending? And the Wall Street Journal had a great piece talking about how the money required is like 5 times any amount of any subscription revenue that they could possibly generate. Like, it's five times larger than the current level of subscription revenue in the world. And so I think for them, they're like, well, how do we monetize? One really good way to monetize is by doing social media because you can sell ads against it. And so for them, like when I saw that happen and I've got checked this against a few people, it's like, oh, this might just be a Last resort sort of monetization model, turning to social media, trying to get advertisers and hopefully, you know, creating some element of profitability for all of this.
- [04:58]
A
Yeah, to me I think they're 100 leading into the ad model. Didn't they hire someone that used to be, that used to run ads at Meta or another tech company? Alex, correct me if I'm wrong, they did do that recently, right?
- [05:12]
B
Yeah. They have Fiji Simo, the running the consumer business now. She helps scale Facebook's early ads business. A scoop I had in Sources News actually a couple weeks ago is that she's looking for a head of monetization to do ads in chat, GPT and Sora right now.
- [05:29]
A
So how do you so talking to these executives? You know, when this AI stuff was really taken off a year or two ago, the promise was like, oh, we're gonna solve cancer, we're gonna, we're gonna do all these cool things. And now, now we're just getting slop, right? And they're just pushing that out. How do the, how do the people inside these companies think about that? Do they hear the criticism, do they care about the criticism? Or is this just like an ends to like meet the revenue goals so they can do the cool stuff?
- [05:59]
B
It's both. I mean, I think you can do multiple things at once. You can have your slop and you can still try to cure cancer. This is a big company. I mean, OpenAI is like approaching 4,000 people. So it's not this tiny research lab that it used to be. If you paid attention to all the press Sam Altman did this past week, he talked a lot about how excited he still is to solve science breakthroughs and how they see that as a key benchmark to reaching AGI. I think they also see that as a huge revenue opportunity. Obviously they're still focused on that. I think you can do multiple things. But you did see a lot of OpenAI employees on X actually when the Soar app launched, talking about the concerns they had about it. There were ones even criticizing the Meta Vibes launch, which was like Meta's ill fated Sora competitor they tried to launch the week before. Like they were criticizing that and the approach there. And then they did their own version of AI slot the next week. So yeah, there's definitely like a cultural divide here of like the people in the AI labs who want to just build the foundational research, create the science breakthroughs, get to AGI automate all human labor. And then the people they brought in from big tech and Startups on the commercial side under Fiji Simo, who have to make money, who have to pay for all of this, have to pay for the training runs and the research, which is only going to continue scaling. So you have to have one to feed the other. And yeah, I think Kyle's analysis is, is totally right. You have to do ads. There's no other, there's no other way to monetize a user base at this scale than advertising. And the vast majority of ChatGPT users are free. And so like the vast, vast majority. So they, they have to do something there. They can't just let that continue to be free and melt their GPUs. So I would be shocked if there's not ads in ChatGPT within 12 months.
- [07:48]
A
I think it might be even sooner the way that they're going.
- [07:51]
B
It takes time. You'd be. I mean, yeah, they move really fast, but you would be surprised also just given how fast ChatGPT has grown and that they didn't foresee becoming this major consumer platform, that the backend technical lift of chaining all the data together and getting it in a place where you can do ads at the scale that they want to operate in and all the markets they operate in, that usually takes years. So, yeah, I think, I think a year is actually pretty ambitious. We'll see. They still have to get the leader in to run this team that they're setting up for this. So they haven't yet, they haven't yet done that. And you've heard Sam saying like, oh, I like Instagram ads. And like he's starting to like be open to advertising in his comments, which he used to not be. So I think they're still warming the culture up actually to this idea because it is going to be a shock to the culture that thought they were all building something that was like the anti Metta. Yeah.
- [08:41]
A
Kyle, what do you think about that?
- [08:44]
C
Yeah, I mean, Alex is definitely the expert here, but I was at a conference that was around the future of work and every conversation just kept coming back to AI and what the AI companies are doing. And they're both like practitioners of AI, people who work at the labs, people who study and teach this stuff, like journalists who cover it. And I feel like everybody was like, why are they doing social media? And the whole takeaway was, was the money, right? The amount of money, as you know, that's been spent on, on building out these data centers is just astronomical. And they're turning more and more to credit financing for it. Like, I think they're like 15% of the investment grade market. And so, you know, that's when it gets really spooky. Like it's okay to cover things with cash, but when you start taking out debt to do it, like you, you better have some sort of plan. And so I think for them it's, they're probably like watching all of this, seeing how the market has ramped up and realizing that they, they have to have some sort of solution. But what Alex said, it takes a lot of time and I, I don't know how much time they have. I think that's the question that everybody's trying to figure out. Like, how fast could this bubble pop? Is it a bubble at all? And yeah, I guess we'll find out.
- [09:55]
B
It's definitely a bubble. Yeah.
- [09:57]
A
I mean, Sam has said it himself. Like Sam Altman said that? Did he say that to you?
- [10:00]
C
He's just marketing. He's just marketing.
- [10:03]
B
No, I mean, I, I mean I was. They all, their execs did a Q A that I was at on Monday at Dev Day and they were all like, we have no idea. I mean, I think they're being facetious, obviously, but I do think there is a genuine level of they don't know how to deal with the fact that anytime someone is like mentioned in the same breath as them, their stock pops 20%. That's like a thing that they are trying to navigate that they did not foresee being that reactionary. I think they knew there was going to be this halo effect, but I think they are even surprised by the market fervor around this and I think it scares them. I got a lot of like, I just got a vibe of like the nervousness that they have about this being really real because they are potentially the bubble. And I think they're starting to realize that.
- [10:47]
A
I mean, Kyle, you make a good point though. Like how much time does OpenAI themselves have? Because the latest figures I believe is like they have what, $12 billion in revenue a year approximately. Google makes that every two to three weeks. Right. So like they're having to take on debt, they're having to raise more and more money to pay for these multi billion trillion dollar deals with Oracle and amd. How much longer can they keep doing that? And if the markets turn, whether it's through new tariff announcements from the Trump administration or whatever, is that just going to shorten that Runway?
- [11:17]
C
I mean they signed, I think the headline was they signed a trillion dollars in deals this year. So that's a huge, gigantic number relative to their revenue, which, yeah, I think is around 12 billion, but to the point of tariffs. Jolie. Joey Politano, who writes. I actually don't know how to pronounce his newsletter. It's like a prison toss. It's a really great newsletter. But he talks about how the tariff policy around computer products has actually been quite generous. So for the AI companies, they actually have a lot of support from the government to kind of go out, build out these data centers. They've been relatively protected from tariffs thus far. And, and so I think they're. They're insulated from some of the swings in fiscal policy that we've been seeing.
- [11:58]
A
And they're incentivized, like you said, just to. Just to not get in the way. I mean, there's some people that have called for regulations around, like, these AI terror, AI video apps like Sora. It's like kind of, you know, freaky stuff. Like, it can, like, really mess with boomers especially. There's probably not going to be any regulations around that because, like you said, no, the administration doesn't want to get in the way of anything that could slow down, slow down the AI story because that, that could solve. That could cause the stock market to kind of lose its steam.
- [12:26]
C
Yeah. And I think, you know, the, the thesis I put forward in my recent piece, and Alex, I'm curious what you think about this, but is that the stock market is essentially providing a floor to the administration to make the decisions that they've been making, like, whether or not you agree with the decisions. And so, like, it's giving them a lot of latitude. Like, they wouldn't have the freedom, I think, to, like, shut down the government for as long as they've been shutting it down. If the stock market was crashing, like, market has been ignoring this element of reality and has only really been focused on the AI companies. And I think the administration has an incentive to allow and protect the AI investments to kind of keep that loop going.
- [13:07]
B
I'm curious why you think that's happening. Do you think it's because the AI companies are putting forward such insane growth projections and constantly increasing their growth projections? Like Oracle, you know, was a great example the other week. Like, the fact that there are companies for the first time that are worth hundreds of billions of dollars are going like, oh, we're actually going to have an additional tens of billions in revenue. Like, that has never happened in, like, the history of, like, the world. So is that why the market is ignoring all this? Is because, like, the companies are putting out these Insane projections.
- [13:40]
C
Yeah. And like no other parts of the economy are moving. Right. So like nearly 40% of US GDP this year I think has come from AI is like 75% of S&P 500 returns, has been the companies, they hold a significant portion of the debt, like I said. And the only jobs that the US has been adding is in like health care. And so all of these other components of the economy are frozen in place, understandably because of the situation with fiscal policy. Like tariffs really frees up a lot of industries. And so I think, yeah, like investors are like, this is the only. There's a good piece in the FT called United States or the United States is AI now. Like it is just AI. And I think that's true. It's, it's just AI based on what.
- [14:24]
B
You'Re saying and also just kind of want to understand of how open AI thinks. If I was an investor and I'm not, the, the data point that I would be closely tracking is open AI's revenue run rate and ChatGPT's growth rate. Because if that decelerates or even reverses at all, I think it's a massive correction because Allman has even said, I mean, I think it was with Ben Thompson at Strathecary, he was like, how are you going to pay for all this? And in a more direct way than I've heard him say it before, he was like, well, check our revenue. Which is like duh. But like it was good to just hear him say it because it's like, where's the money going to come from? And a thing that I've noticed is that ChatGPT is getting fairly saturated in the markets where it can monetize the best. And like in India, which is I think rapidly approaching its largest market, they're actually cutting the subscription cost aggressively because you can't scale a 20amonth subscription in India to hundreds of millions of people. So the margin profile on the business is looking really worse actually the bigger the platform gets until they turn on ads. So they're in a race to turn on ads in that sense because the bigger they get, the less their growth. Their revenue growth rate continues. And there's other, there's other inputs there. Like they have the developer platform, the API business, which I think they're going to push very hard on in the coming months to make up for ChatGPT revenue decelerating. But that's a big thing. I do think there's other just wildcards like up the sleeve that we may not see coming to like, I think they could theoretically start charging companies thousands of dollars for a single prompt. Potentially millions of dollars for a single prompt is actually what Altman was talking with me about a few days ago. Because if the reasoning models get as advanced, continue getting as advanced as they are, and they can really get inside a business, I think they're excited about putting like an entire data center to use on one problem.
- [16:22]
A
And that's probably pricing power to do that though. I mean, they have anthropic.
- [16:25]
B
If the outcome is good enough, right? If the outcome is like solving a fundamental, applied like material science problem, yes. But again, this is also theoretical. It sounds like sci fi, right? So the wild cards that are like, oh, they pay for this is like something out of a Netflix show. And then what was practically happening is that ChatGPT is getting saturated. It's getting a lot of competition from Gemini and from all the other chatbots. And its margin profile in the subscription is getting worse as it scales. And so they're in a really tough spot. I mean, they have to execute. I interviewed Greg Brockman, the co founder who runs Stargate, a few days ago and he was like, yeah, we know, we know that it's on us to execute it. We have, we feel the, we feel the pressure there because now you have to go and actually like let put the shovels into the ground and connect the racks and set these data centers up and do it at record speed. And they've never done that before. You know, they're starting this all from scratch as a company to be their own hyperscaler. So I would be chat. I would be chat. I would be. Yeah, I mean I would be tracking ChatGPT and I would be tracking the build out of these data centers pretty closely.
- [17:33]
A
What I worry about though is like OpenAI is private, right? Like we're not going to get quarterly reports. So now it's going to be up to reporters like you, Alex and other, other people that can get the scoop to figure out what the actual story is inside these companies. And if it's legit, you know that that's, it's all going to depend on that. Otherwise we're not going to get the real information.
- [17:54]
B
Have faith. Have faith. I think they have a lot of investors and they're a pretty leaky ship. So I think, I think their revenue numbers will continue getting leaked.
- [18:02]
A
That's a good point. Kyla. You know, something that I liked about in your piece was that you talked about how gold is also going up at the same time where like you know, OpenAI is worth 500 billion and you know, Nvidia is at 4.5 trillion. Those two assets don't typically go together at the same time. Usually you have risk on assets like tech companies and you have gold go the opposite direction, but now they're going up at the same time. Is gold just a hedge against AI slop? Is that, Is that what that is?
- [18:31]
C
I mean, I think people have different views on it. You know, there's, there's like a world outside of the United States. Right. And so I think some people are looking at, you know, the available options for investing and saying like, you know, it used to be that the US was kind of the safest place to park your money. And it doesn't seem like that anymore. Like China has really tried to establish themselves as the custodian of gold reserves and are, you know, pitching that to people as a way to, you know, implicitly de dollarize. And so the most interesting take I saw on it was somebody who went on the Odd Lots podcast with Bloomberg and he was essentially like, tech is ripping apart society, so it's a good investment. But then also like, people are buying gold as a hedge against that destruction. And so it creates a rather complete portfolio in that way. But yeah, I think people are seeing, you know, a lot of the concerns with how the Trump administration is running policy. Again, whether or not you disagree with it, it is quite volatile and sort of looking for an alternative. I think a lot of investors are pretty spooked about the run up that we've seen in the S and P and whether or not that's sustainable. Like, you know, the question that we asked a little bit earlier is all of this a bubble? I think literally everybody who invests is asking that at this point. And so I think people are saying like, gold is a good thing, it's the best performing asset this year, which is a very, very strange environment.
- [19:56]
A
Yeah, I think, you know, I was thinking about this as well, like the fact that they're both going up at the same time. I think there's a chance that we have a 20% drawdown in the S and P then before we have a 20% drawdown in gold. Just the way that things are going right now.
- [20:08]
B
Now, because didn't Jamie Dimon say like 30% drawdown risk or something that what.
- [20:12]
A
He said, but he said that, he says that literally that's just, I don't know. He's a, he's a banker. He has to be worried about the economy every year. I'm curious to get yalls take though on. We're in the middle of the froth right now. Everyone says that it's a bubble. How big does it get? So if I was to say, does OpenAI get to a trillion dollar valuation as a private company, yes or no? Just curious to get y' all stake.
- [20:38]
B
They're currently at 500 billion as a private company. I think that's pretty rough. I would put the odds at like 40% as a private company.
- [20:47]
A
Kyla, what do you think? Yeah, I know if I was to say over under OpenAI becomes a publicly traded company before 20 over. Under 2028. Do you guys take the over or under.
- [21:04]
B
Over? Just because if they don't, that's the bubble. They have to, they have to have the liquidity. They have to.
- [21:15]
A
I mean they're gonna be before 2028, right?
- [21:18]
B
Yes, yes, sorry. So yeah, like, like they can't, they can't continue raising at the scale they're at as a nonprofit. And the markets, it's just like it's too pent up. So I do think they, they have to for many reasons and I think they're worried about it for reasons beyond the non profit. They have to like clean up their act a little bit internally I think. But yeah, I think they need to, they need to, to fund this, to fund the Stargate stuff. And they know that.
- [21:53]
C
Yeah, $500 billion, right. Allocated to Stargate at least.
- [21:58]
B
Yeah.
- [21:59]
C
Do you think that's viable like that project?
- [22:03]
B
I mean in some form? Yeah. Like they are going to build data centers. They already are and albile and they'll build more. Is it going to be a thing that they actually spend 600 or 500 billion plus on and build? Almond has a goal of like 250 gigawatts and like by like 2030 or something or 2035. It's like that stuff. No, that's like Jeff Bezos said something recently about like, oh, we're just going to send data, data centers into space and that's how we're going to like keep building data centers. And it's like yeah, if you're going to actually do 250 gigawatts, like yeah, like I'll believe it when I see rockets like with data centers going into space. So maybe that happens.
- [22:42]
A
The power is going to be the biggest constraint moving forward. It's not even going to be chips.
- [22:46]
B
If we want to live in the matrix and like be plugged into like tubes underground, then sure. But I don't think anyone Wants that.
- [22:52]
A
What do you guys think Jensen was. How do you think Jensen reacted on Monday morning when he saw that Sam Altman did this deal with amd? You think he. I saw his reaction partially on cbc.
- [23:02]
C
So catty about it.
- [23:03]
A
Yeah, like, understandably so.
- [23:06]
B
He probably did.
- [23:07]
A
So he probably had no idea that.
- [23:08]
B
That he said he didn't know.
- [23:10]
A
Oh, he said that.
- [23:11]
B
He said he did. No. Yeah, yeah. I mean they used. I mean, yeah, I'm sure he doesn't love it, but AMD is giving away up to 10% of the company in exchange for selling GPUs. So he doesn't have to do that. So it's. He was kind of, you know, being catty about that, which makes sense. I think he's in a much better position. But, you know, it's. I think OpenAI is going to get the compute anywhere they can. And it's interesting also that like Microsoft is not mentioned at all. You know, Microsoft was not mentioned at all at their conference this week in any of these recent headlines, their initial hyperscaler. Right. That owns a huge chunk of the company. It's totally absent from these current data center discussions. I think that's pretty telling.
- [23:56]
A
I think there's, I think there's definitely beef there. I think we got time for one last question. I'm curious, have you guys tried some of, have you guys tried the Sora Apple and the Vibes app? Alex, I know you cover social media for a long time. What was your take on it? And then Kyla, I want to get your take as well.
- [24:12]
B
Yeah, yeah, I've been, I've been making Sora videos. I played with Vibes a little bit. Wasn't really for me. Sora feels like a Vine. That was like my initial reaction to it. It was like it really reminded me of vine in terms of how fun and lighthearted it was and goofy it was. So I like that approach. I like the mixing real people with artificial. It's better than just pure artificial for me. But it's a very buggy app, you know, and there's. It's still invite only. I think they have a lot of work to do on it. But I like it. Yeah, it's like, it's like a thumbs up for me.
- [24:46]
A
Kyle, did you try it?
- [24:48]
C
I'm a little grouchy about it. Yeah, I think I'm just grouchy about everything. I just like the IP stuff is crazy. They're like, you have to opt out. It's like that's not a good thing to do to people. And as somebody who, you know, my face is out there on the social media apps, I don't particularly enjoy, I think the direction that it's all headed. And I think Meta really, really just disrespects their user with Vibes and it's just, I mean, it's a very Zuckerberg type thing. He knows it's going to make money, but I think there's better things that we could do with all the resources that we're allocating to this stuff.
- [25:20]
A
Vibes is so bad, right? Like, I saw the Vibes thing and I'm like, what? This is the definition of AI slope. You can, you could. I could get on board with, with the Sora thing that it's fun. I shared a bunch of videos with my friends making goofy videos. Vibes is like I'm watching a screensaver. Like, what is this? All the billions of dollars in compute going towards making three, you know, 40 screensavers. Like, I just don't understand the direction that they're going here.
- [25:44]
C
It's a very Zuckerberg type direction.
- [25:47]
B
Yeah, I think everyone is kind of like moving in the dark here and like figuring out what actually is going to work. And I think OpenAI's first stab was clearly more culturally resonant. But I do expect this space of synthetic social media to evolve very quickly and to become much more prominent across all our feeds whether we want it to or not.
- [26:09]
A
Awesome. Well, I appreciate you guys hopping on today. Two legends in the space. Thank you so much for Yalls takes and hopefully we get to do this again soon. I had a good time.
- [26:18]
B
Yeah. Thanks for having us.
- [26:19]
C
Yeah. Subscribe to Sources News.
- [26:22]
A
Sources.
- [26:22]
B
Subscribe to Kyla Access podcast.
- [26:25]
A
Yeah, Access Podcast for Alex. Sources for Alex and then Kyla on Instagram substack. I mean, she's everywhere. So, you know, it's hard. It's hard not to see Kyla's stuff. Thank you, guys.
- [26:35]
C
Thanks, Sid.
- [26:36]
A
Well, all right, guys. Hope you enjoyed that conversation with Kyla and Alex. I was really happy that both of them came on at the same time. It was just great hearing each of their perspectives on where we are in the AI space. I want to give a shout out to Alex's new newsletter one more time. It's called Sources. You can find it at Sources News. And we'll also put a link in the description. The description as well. Alex also co hosts a podcast called Access where they recently interviewed Mark Zuckerberg. And for Kyla, you can follow her stuff on Instagram and Substack. Let me know in the comments on Spotify and YouTube on what you guys thought about this episode. What were some of the takes that you guys agreed with or really disagreed with? And if this is your first time listening to the rundown, just a reminder, we post a new 10 minute episode every day breaking down the markets. So make sure you guys are subscribed to the podcast to stay in the loop. Thank you guys again for listening, watching and commenting. Shout out to Mike and Connor for all the work behind the scenes and we'll see you guys back here on Monday.