← Big Technology Podcast
Big Technology Podcast
OpenAI and AMD's Megadeal, Sam Altman and Jony Ive’s Bumpy Start. Meta vs. Apple
00:57:35
Loading summary
Transcript61 lines
- [00:00]
A
OpenAI and AMD sign a megadeal where OpenAI could end up owning 10% of AMD. Sam Altman and Johnny I've run into some turbulence in their quest to build an AI God device and Meta and Apple and maybe the rest of the tech world are on a collision course. That's coming up with Spyglass's MG Siegler right after this. Welcome to Big Technology podcast where MG Siegler is here to make his first Monday of the month appearance and we have so much to talk about. We're going to cover the tie up between OpenAI and AMD and the fact that Nvidia and OpenAI's deal is really just a letter of intent at this point. We also have Sam Altman and Johnny. I've running into some bumps as they begin to build their AI device and Meta and Apple and maybe Amazon and Google all are basically on a collision course to build the same thing. And we will take a look at how the progress is going on all fronts and who's winning. So great to see you again, mg. Welcome back to the show.
- [01:03]
B
Great to be back, Alex. And I know we say it every single time I'm on, but there's so much happening and it's always, there's always more happening. It's just, it's wild how this is sort of just playing out.
- [01:13]
A
I love how we catch up every month and I was looking at the headlines from just this morning and I said we could do a full show on that. Last month we were talking about the rollout of GPT5 and whether OpenAI has broken ChatGPT with it. That is a.
- [01:29]
B
It feels like a decade ago.
- [01:31]
A
Yeah. Right. So let's begin of course, with, with Sam Altman. And a few weeks ago you called out an interesting line in a post of his about abundant intelligence and it was seemingly a wink towards intelligence. Interesting things to come. First of all, Sam says in this post, a very short, short post, he says, our vision is simple. We want to create a factory that can produce a gigaw gigawatt of new AI infrastructure every week, which is crazy. He says the execution of this will be extremely difficult. It will take us years to get to this milestone and it will require innovation at every level of the stack, from chips to power to building to robotics. But we have been hard at work at this and believe it's possible. Over the next couple of months, we'll be talking about some of our plans and the partners we are working with to make this a reality. We have some interesting new ideas and he sort of left it at that. And it seems like now with this deal with amd, which we're going to get into the details of, which by the way might have OpenAI ending up as a 10% owner of AMD, this is part of the interesting new ideas that Sam teased. So let's just start with the high, the high level here. Are we starting to see Sam's plan come to fruition?
- [02:54]
B
It seems like this is part of it. Right? It's, it's, this is, it's so hard to analyze this in part because of what we kicked off with. Right. Things are moving so fast and it's almost hard. It's obviously impossible for us on the outside to know how much of these deals are coming together sort of on the fly. Like there was the reporting about Obviously the Nvidia 100 billion dollar deal which led presumably to Sam's, you know, post there that you're talking about that that sort of came together while they were both on the overseas excursion with, with Trump and, and sort of going around and, and they were able to hash out that deal sort of one on one. And so with a deal like this now with, with amd, you almost wonder like did a similar thing play out? Was, was Sam with Lisa sue traveling around somewhere? Did, did she pick up the phone when she saw that Nvidia deal and decide like what happened? We need to, you know, figure out a deal basically over the next few days to sort of be able to, to respond to that in some ways. And so it's when, when talking about Sam's post, it's sort of like I'm, I'm unclear if he even knows all of the, the ways that this is going to go from his sort of nebulous, yeah, talk about these interesting new ideas for financing. And also that post, I felt the need to write about that both because I like the, the Sam Altman post, Kremlin ology of it all and doing breaking those down, especially when they're as short as this one, is that, you know, it's sort of, I kicked off talking about it felt like this was basically written because Sam felt like, well, we just did $100 billion deal with Nvidia, someone should say something about it. And then, and then, you know, it goes on for just a little bit with these, these very again, nebulous and grandiose terms as he often does. But, but this was also tied to, you know, some other quotes that, that were coming out of this from other companies. I think one was Was it the. Oh, it's the new co. One of the new co. CEOs of Oracle also had, you know, alluded to these, these strange new financing mechanisms. And of course Oracle's a big partner with OpenAI too. Is everyone's big partner with everyone these days. But, but. So it's like what, what's going on behind the scenes that all these folks are getting together and coming up with, with new accounting methods out of the blue that they, that they're hinting at but not fully coming into. And so yes, it's conceivable that this AMD deal is one of them, or it's conceivable that this AMD deal came together at the last minute because of the other deals going on.
- [05:27]
A
I have to give Sam credit. He's someone who can write like we're going to be telling you about some interesting new ideas we have about financing and building and most CEOs, I would say you re. You would read that and be like, there's nothing there. But with Sam, we've seen that there is something there and that's either. And maybe it's both of these options. I was gonna say it's either a testament to his ability to turn momentum and ideas into business reality, or a testament into the entire business world blindly following this one man and putting so many billions of dollars behind what he's promising in the future that it's. It's somewhat concerning. But I think we should just talk a little bit about the nature of this deal because it's such an interesting deal between OpenAI and AMD. So this is from the Wall Street Journal. OpenAI and chipmaker advanced Micro Devices announced a multi billion dollar partnership to collaborate on AI data centers that will run on AMD processors. Under the terms of the deal, OpenAI committed to purchasing 6 gigawatts worth of AMD chips starting next year. The chat GPT maker will buy the chips either directly through it or through its cloud partners. AMD chief Lisa sue said in an interview Sunday that the deal will result in tens of billions of dollars in new revenue for the chip company over the next half decade. Part of this, OpenAI will receive warrants for up to 160 million AMD shares worth roughly 10% of the chip company, at $0.01 per share, awarded in phases if OpenAI hits certain milestones for deployment. I have to admit, MG this latest one, it's got my head spinning. So let's just think I can. Can we recap. It is. It is not an investment in OpenAI it is OpenAI's commitment to buy AMD chips, but as it buys these chips it gets ownership in amd.
- [07:29]
B
Yes.
- [07:29]
A
What makes sense of this is this AMD giving away the company for OpenAI to buy its chips and is that really a purchase or is it just like. So it's not even. Yeah. And you go ahead and this is struggling here.
- [07:43]
B
And again this, this is sort of late breaking news, so I haven't had too much time to sort of go over the details of it. But at a high level, from what, you know, the report is here, it does seem like there's a, there's a number of things going on as always. But so the 10% thing is super interesting, right, because that' that's also oddly the number that, you know, intel, you know, was, was supposedly selling to the government, right, like for, for their deal. And there's, you know, there's other ownerships trading hands obviously between Nvidia and Intel and then a bunch of other stuff going on. So it's like it all feels like the same sort of general playbook just being executed in different ways where it's like now all of a sudden ownership is sort of table stakes to these deals. Right. Even when it's a company that's a, that's technically still a startup in OpenAI own would be owning up to 10 of a very public, long standing public chip company in amd. And, and what on earth does that look like from a finance perspective? Like their books, like how it's from again from what the reporting is like they have the right to buy those shares I think in warrants. And so like, I'm not sure that they necessarily will or, but they could. And, and obviously as you noted, it's sort of dependent on, you know, the business agreements between these two companies. And then there's like the other layer where as you just noted, it doesn't necessarily have to be OpenAI that's buying the chips, right. It could be one of their partners. And so I mean it could be Nvidia technically, I guess, you know, that would be buying AMD chips to help put together one of the Stargate operation things like. Because that's a layer of this too, right? That again, stepping back at the highest level, it seems like OpenAI realizes that they need to control their own destiny in many ways, but certainly on the cost front and to the ultimate extreme, they probably can't be paying one of the massive cloud providers, be it, you know, Microsoft as it's been to date. And then sort of Oracle 2. Now, all of them, they need to be able to control their costs in a better way. And doing that means eventually at some point sort of controlling your own data centers to some extent. And so it feels like a lot of these deals are coming together to allow OpenAI to put in place the building blocks to build their own data centers, quite literally. And this AMD deal presumably is also a part of that. But again, the wrinkle of it being one of the partners doing it, again, I think that that's related, just thinking on the fly here, related to Stargate partners being able to use it in their own data center. So like Oracle famously controls the one right now in, in Abilene, Texas. That's, that's the only one that's actually being built right now in the process of being built while the others are still just sort of being signed and, and squared away. And so again, like who gets what rights to purchase, you know, the, the processors? Who gets what's rights to purchase the, the shares in amd? And is AMD simply doing this again? Because it's sort of become like the table stakes thing. And by the way, it worked in terms of the narrative of it, right? Like is up a huge amount right now, is flying 27%. Yeah, right. So like obviously that comes into play too. They all see that these deals get done and then the stock goes through the roof because investors are excited all of a sudden again about, about AMD because they're doing a similar deal to, you know, what, what Nvidia and Intel and all the other players have done.
- [11:18]
A
Just want to pause on the 27% we've seen. This is the second time we've seen something like this happen, right? AMD is up 27% today. Oracle of course blasted into orbit when it made its deal with OpenAI. We're gonna, we're really not gonna focus on the stock market in this show. We're gonna talk mostly about product and, and partnerships. But it, it is worth noting, right, like this exuberance in the stock stock market is based off of infrastructure spend and not revenue. It's all based off of revenue projections. And those re revenue projections are extremely aggressive. Especially when you consider the fact that enterprises are moving exceptionally slow here. This is a red flag to me.
- [12:01]
B
And so that goes back to my post. The reason I wrote, you know, again off of riffing off of Sam's comments, because the only thing that I could think about, you know, in my head of like what they were, you know, presumably talking about with this New method of financing or whatever wasn't actually new method of financing. It. It's, it's basically using debt in a way to sort of get a flywheel going so that, you know, obviously leveraging debt, especially for, for data center build outs and things of that nature is not, is nothing new. But what, what you're sort of alluding to here is exactly what might be the new element of it in that they, these players are all trading revenue. And, and you know, again, it starts with OpenAI agreeing to do these massive deals with another partner. But as everyone knows, this has been widely reported. OpenAI doesn't have the money to do these, you know, the deals to the extent they don't have the money in the bank literally to do the deals, you know, to the extent that these, these are talking about. And so now we get these other deals like the new Nvidia one where it's like they are promising to, you know, commit up to $100 billion. It doesn't have to be that much. It starts with 10. Right? 10 it sounds like is, is maybe committed depending on how, how committed these two are to actually know finalizing this deal. But it's like 10 comes in at first and then another 10 comes in over, you know, and basically over nine tranches, they, they would get this 100 billion. But the timing of that all, like, is there actual money going into bank accounts to pay for something, you know, physically being built at some point or is it all just being based off of like lines of credit?
- [13:39]
A
You know, could just be chips. Maybe open AI is going to pay for these AMD chips with Nvidia chips. Maybe they're just going to take one of Jensen's trucks of chips, drive it over to Lisa sue and be like, we'll take 10 of your company right now.
- [13:52]
B
I mean if these chips are really at the shortage limits that's everyone's projecting like they're worth more than their weight in gold. And so you know, like they are somewhat of, of an actual value store, I guess. And that, in that way. But, but you know, the thing you're talking about how this makes you uncomfortable, the obvious elephant in the room that again everyone will know is like if even the smallest slowdown happens, not, not even like a massive correction. If a slowdown happens in several of, on several of these fronts, on any number of, sorry I should say on any one of these like several fronts that are going right now, it can start like, it seems like it would start a cascading effect that would potentially bring down the entire thing. And then we got debt being called and we got people unable to pay. And, and we've got new, you know, contracts being renegotiated and people taking out, you know, does the bank own, you know, do banks start owning some of this stuff? And, and, and does the government have to step in as, you know, they basically are clearly open to doing in deals like with intel, as we talked about. And so there's all sorts of weird things that are happening right now that I don't think that there's a ton of precedent for in history.
- [15:07]
A
Right. And as we've sort of danced around or even mentioned so far, there was a very interesting line in the Wall Street Journal story where the reporter writes, the Nvidia deal isn't completed yet. The two companies have signed a letter of intent and have yet to disclose specific terms in the regulatory filing. So is that, is that actually happening? I mean, yeah, I mean certain so.
- [15:29]
B
And so there's a few things here. One, that, what is up with OpenAI's letter of intense like love now? You know, they did the Microsoft one. They announced with, you know, a couple dozen words that they, they had a agreement to agree with Microsoft on new terms, but they have no actual terms to agree upon yet. And so it's not a done deal. Now this Nvidia one is also basically an agreement to agree to, you know, a deal. And you know, you have to sort of wonder if AMD didn't see that and, and think like, maybe we should swoop in here before this is a done deal and sort of get something else done. And, and you know, in the report there, they do say like, and, and I think Altman, I saw it, maybe it was even a tweet or he said it somewhere else, but he sort of makes it clear that Nvidia is still like a main partner. Right. This isn't like, this isn't them switching horses, you know, in, in the middle of a deal being done. But there is, you know, to continue the analogy, there is some level of jockeying that it feels like is happening here, right? Where it's like again, maybe Lisa sue saw that this deal was not finalized and so she could sort of come in and say what can we do? You know, alongside this, if nothing else, if we can't take some of this business, then there's a whole like this is apparently for inference. And you know, obviously Nvidia is in that game too, but wants to be own that game like they do with, with actual model training right now. You know, and so there's that whole element, there's the element of the Broadcom deal that, that OpenAI has, which kickstarted some of these Nvidia deal talks between OpenAI and Nvidia so that they didn't go down the path fully of making their own chips. It sounds like they're still doing it, but maybe not as fast as they would have previously because again, they want to hone the whole stack as we talked about. So, yeah, lot going on.
- [17:19]
A
I'm not saying it's going to happen, but I do think that you could in your mind imagine a scenario where you connect all these dots looking backwards from, you know, in the future, where there is a collapse here and you could say, how did we not like think this was a blatant collapse waiting to happen? Letters of intents, the losses, the size of the investment, the fact that enterprise is struggling to implement this stuff. Yes, chat GPT or Open AI is going to bring in 13 billion this year. At least it's on, on track to do that, but it's going to lose, according to reports, 120 billion by 2029. Again, like the caveat we always say, and I think this is, it's a very freaking important caveat, is that the technology is real chat GPT and, and, and all of these GPTs are, or all these LLMs are getting, getting better as we go. But the financing definitely is, is wonky, say the least.
- [18:15]
B
Yeah. And it feels a bit like, you know, when you do slow down for a second and try to think about this, it feels like a shock and awe campaign meant to leverage speed. And I do think that this is basically what OpenAI is running a playbook of, of like leveraging the fact that as they keep talking about, like this is a unique moment in time. We need to move as fast as possible. We have. There's so much. This is, this is the race to AGI or super intelligence. And, and we need to do this right now. We can't wait. If you're waiting, someone else is going to step in and do these deals like you're either in or you're out. And just using that time pressure to basically force the hands of people into pretty uncomfortable, what would normally be pretty, certainly for a cfo, a very uncomfortable calculation to, to make in these situations.
- [19:06]
A
Right. And I guess there is some logic to them doing that. Right. And yeah, for sure it's the competition. Right. And there's this, there's this account that I follow on, on X or seems to pop up in my timeline all the time you can't tell who I follow anymore because of our algorithmic overlords. But it's a person's name is Burkov and Burkov says Google, Anthropic and OpenAI are currently fighting a war of attrition. The problem for the latter two is their cash comes not from clients but from investors, while Google is profitable and no longer losing users. Add this to the fact that among the three, only Google has its own infinite supply of GPUs for the price of peanuts and draw your conclusion on how this war of attrition will end. Now, obviously this person is very bullish on Google winning this moment. It's quite possible. But I think maybe that's the sort of OpenAI fully understands the fact that it's matched up against them. By the way, also OpenAI also sorry Meta and Elon and then you sort of see why we have this rush the way we.
- [20:05]
B
It's a, it's yeah, completely unprecedented to have this many massive players with this amount of capital being able to be put to work. But I totally agree with that stance and I've written a little bit about this. Like the, even before, yeah, the riffing off of Sam's newest post, it was the, the main thing I've been sort of dancing around and finally was able to put this down. But it's like it really does feel to me at least at this moment in time given that Meta obviously is struggling a bit but is trying to reboot the system. Elon, like they're losing a lot of people. Like there's a lot of weird stuff going on there. There's a lot of spend going on there. He can fundraise like no other of course. So you know, he's, he's good for it now. But like there's weirdness there. Anthropic is a different beast. It feels like, you know, with their Martin, their new marketing campaign sort of plays to that. Like are they going to be now set up as the anti open AI type play but at a lower level? And then so to me anyway, long winded way of saying it feels like this is basically right now as it stands a two horse race between OpenAI and Google. And I do think that OpenAI looks at Google and just says like oh my God, they have everything that they need to do this at scale. While as you noted not having to burn investor capital they had, they, they're insanely profitable. They can do this on their own. And by the way, unlike Meta, which is also profitable. And Zuck talks up like, yeah, we can, we can fund this via profits, which is true, but they don't have the cloud infrastructure that Google does and they don't have the, the TPUs that Google does and they don't have all of the sort of other various expertise to sort of scale this type of unprecedented massive build out scale that Google is able to do. And so I think Sam and everyone else at OpenAI just says like there's a lot going on. All these things are varying degrees of noise and some is real competition but Google is the one that really has the opportunity. And so my headline idea was basically OpenAI needs to build Google Cloud before Google can build ChatGPT.
- [22:10]
A
Right.
- [22:11]
B
And that's how it goes.
- [22:12]
A
Yeah, yeah, couple more things on this. It's just so many interesting aspects to this and that. And speaking of Sam Altman criminology, let's take a look at the tweet that you referenced. Excited to partner with AMD to use their chips to serve our users. This is from Sam. He says this is all incremental to our work with Nvidia and we plan to increase our Nvidia purchasing over time. An astute X user said, everyone is scared of Jensen. This post mentions Nvidia more times than amd. Despite the deal being about amd. Jensen might bump them down in the preference list due to this.
- [22:48]
B
What do you think? Yeah, and I'm glad you found that one because yes that I think I saw that quickly and that is a great point about it. And that's another thing that's at play here. Like and also why it felt like at least a part of why Open AI and Nvidia struck that deal. It's like right now if you want to be a main player in this game, you've obviously got to work with Nvidia. And as everyone knows, like while they'll, they say they don't play favorites, like if they're going to invest whatever $100 billion into you, like you're going to get access to their chips, like they're not going to screw you over. But if you did a deal with amd, you know, behind their back, like maybe you're not first in line for those chips anymore. And so I think you have to sort of like you have to do with President Trump. You gotta sort of do the public fealty and show and show who's really in charge.
- [23:37]
A
Okay, here's my last question about this. Why would OpenAI want to own 10% of AMD? I cannot figure out why it is advantageous for them other than, I don't know, you tell me, what's the deal?
- [23:52]
B
I'm sort of at a loss too for that. I, I think like again the only thing I would come back to is what I said earlier. It's like this has now become like a weird table stakes thing of like these companies trading ownership, you know, and quite high percentages of ownership within, you know, each other. There's, there's going to be subsequent reporting about this obviously as to why that part became a big part of it. But it does, it's like, I think Lisa Su has a quote in there of like saying that it aligns their interests because OpenAI, you know, will now obviously have a potentially have an ownership stake in AMD and so they'll be aligned about like, you know, making sure that they're along for the ride, you know, wherever that ride takes them, I guess. But like, yeah, that's, it's a public company selling up to 10% of, of their shares. It's like. And again it's a private company buying. It's like very, very strange. I guess OpenAI might like it because it's like just something else to have on the balance sheet when, if and when they go public and they can say like look, we have these shares worth 10% of AMD and it's worth X, you know, hundreds of whatever it is, billions of dollars. And I don't know, I'm sort of drawing at straws here.
- [25:05]
A
Right. And now Microsoft and Nvidia are going to own AMD as well for the money that they put in. Yeah, helping build AGI. Well, if you don't get AGI, you can get some AMD shares. It's a, not the best constellation price, but worth, it's worth more than nothing.
- [25:23]
B
So weird.
- [25:24]
A
Okay, what is this Compute going to be used for? On the other side of this break, MG and I are going to talk about Sam Altman and Johnny. I've attempt to build an AI device and. But why that is running into trouble because of a lack of, among other things, Compute. We'll be back right after this. Did you know your credit card points and miles can lose value to inflation? Credit card companies often reduce the redemption value of your points and miles. Now imagine a credit card with rewards that can grow in value. With the Gemini credit card, you can earn Bitcoin or one of over 50 other cryptos instantly with no annual fee. Every swipe at the store or gas pump earns you instant rewards deposited straight to your account. Plus sign up now for a $200 Bitcoin bonus to kickstart your rewards. Visit gemini.com card today. Check out the link in the description for more information on rates. Again, if you're looking to invest in Bitcoin but don't know where to start, the Gemini Credit Card makes it easy. The Gemini Credit Card is issued by Web Bank. In order to Qualify for the $200 crypto intro bonus, you must spend $3,000 in your first 90 days. Some exclusions apply to instant rewards in which rewards are deposited when the transaction posts this content is not investment advice and trading. Crypto involves risk. The Gemini Credit card cannot be used to make gambling related purchases. Capital One's tech team isn't just talking about multi agentic AI. They already deployed one. It's called Chat Concierge and it's simplifying car shopping using self reflection and layered reasoning with live API check. It doesn't just help buyers find a car they love, it helps schedule a test drive, get pre approved for financing and estimate trade and value. Advanced, intuitive and deployed. That's how they stack. That's technology at Capital One. And we're back here on Big Technology Podcast with MG Siegler. He's the author of Spyglass, which you can find@spyglass.org, lots of great stories in there. Really an impressive September from you mg. I just like going through the stories before we talked and you are my go to source for basically everything that's happening in this crazy world, including this very interesting report that we have coming out really over the weekend today about how Sam Altman and Jony I've are hitting some speed bumps shall we say with their secretive AI device. I'll read from the FT briefly. OpenAI and star designer Jony I've are grappling with a series of technical issues with their secretive new AI device as they push to launch a blockbuster tech product next year. Their aim is to create a palm sized device with a screen that can take audio and visual and visual cues from the physical environment and respond to users requests. OpenAI and I've have yet to solve critical problems that could delay the device's release though. These include deciding on the assistant's personality, privacy issues and budgeting for the computing power needed to run OpenAI models on a mass computer device. And there's a person that tells the FT somebody close to I've that compute is a huge factor in the delay. Amazon has the compute for Alexa, so does Google, but OpenAI is struggling to get enough compute for ChatGPT, let alone an AI device. They need to fix that first. All right, we're going to talk about the nature of this device because we're getting like a picture of what it's going to look like. But first on the significance of somebody close to. I've complaining to the FT that OpenAI doesn't have enough compute for this device to run. That to me is pretty interesting and I think not a good sign and this is a multi billion dollar tie up. It's not a good sign for the fact that somebody close to I've is saying this, you know, for a story about the device hitting speed bumps. What do you think?
- [29:21]
B
MG yeah, yeah. The behind the scenes element of that is super interesting. You have to wonder. To me it's sort of felt almost as if it's, you know, probably, I don't want to say it's an intentional leak, but I do feel like it's directionally interesting that they're basically telling the market that this device is not going to be on time like it because their original reports, right, were that it was going to be probably sometime at the end of next year, the end of 26. And so, you know, this to me signals that this is more like a 2027 type thing now. And I think that that SOR quote maybe helps that you know, get out there so people aren't expecting, you know, have a countdown all of a sudden a for a year from now when, when this device doesn't hit the market. But I also think like there's a lot in that, in that report, including yeah, like what this device, you know, potentially is. They don't come out and say it directly but you know, a lot, I think a lot of us who read those original like reports when the, when the two sides both started working together but then obviously when the acquisition happened, which also feels like 20 years ago, but was only a few months ago, which is again wild. But I think, you know, we all sort of zeroed in on it being not necessarily a wearable, even though that was, you know, what meta and, and you know, others were working on, but that it could be some sort of thing that's, you know, a voice first device that you carry around alongside at least to start, certainly a smartphone. So it's a, a complementary device to a smartphone. And so there's, there's that element of it and then yeah, the data element of like what would be required to make this work. It's sort of, I'm curious why that compute complaint is in place other than the general compute complaint that we're seeing with Sora and everything else right now. Right. Like they're always compute constrained to some degree and you know, right now they're rate limiting the creation of Sora videos. They, they just jumped it down. I saw recently from 100 in 24 hours because I keep hitting it to now it's 30. And so they're really feeling constraints. It feels like feeling the compute pressure on that. And so with this presumed voice first device, the report is that there will be cameras involved with it too. Like is that really going to be that big of a data hog? And maybe the part of it that is, is the fact that it's apparently always listening, it's always on, always maybe recording. So that's maybe part of it. But it's, it's a curious, it's a curious bit of reporting that, that, that's like the issue, one of the main issues.
- [32:09]
A
Right. And I think what you're getting at is the compute shouldn't be the main. I mean, obviously it's an issue. It shouldn't be the main.
- [32:17]
B
The main thing that would stop a rollout. Yeah. And they could just limit.
- [32:21]
A
It was.
- [32:21]
B
Yeah.
- [32:22]
A
Right. And even if it was and you and someone from Johnny I've team wanted to leak out that were having some issues, you know, the fact that they. The leak was not just we're delayed but finger pointing to me points to the idea that these teams aren't working well. I mean like if you have a sports team and they collapse, right? A sports team that loses a couple games, you hear the athletes and they're always like, we need to do better as a team, we need to get better. And then let's say they lose six games in a row. And I'm familiar with this because all of my teams lose six games in a row, that's when they start finger pointing and they say, well, the defense wasn't holding up, it's part of the bargain. Or you know, we, we really, you know, are hitters at the bottom of the order. Need to, need to come through. And for them to say OpenAI is struggling to get enough compute for ChatGPT, let alone an AI device, they need to fix that. Ooh, that is that. The fact that someone would go to the FT with that is, is I think fairly concerning hearing you say that now.
- [33:22]
B
It's sort of brings into my mind like the idea of sort of what we were talking about with Sora because that's like the, the prime example of, of their compute constraint at the moment, at least forward facing wise. But it sort of feels like, okay, there's all these different things going on at OpenAI in the moment and now famously they brought in another CEO to help, to help deal with all that in Fiji. Simu. Simu. And so it feels like there's maybe some jockeying happening for making sure that each team gets their level of compute that they want to get or feel like to get in order to, to fully be able to, to bring their product to market. And again, like the happens to be the time when SORA is really sort of just blowing up their servers at the moment. But obviously they're working on other models and there's other needs for compute that are going on behind the scenes, there's other products, there's all the agentic work that they're doing. And you know, it feels like maybe this is just a way to ensure that like hey, you guys spent a lot of money to bring on the IO team. We need our, to make sure that we have our, our resources too.
- [34:29]
A
Yeah, I think that's.
- [34:30]
B
And they feel like they're not getting them for some reason.
- [34:32]
A
Now that you contextualize it. I think that's, that's the exact right read. And it sort of sparks in my mind the fact that this has always been an issue at OpenAI teams fighting for compute. And I think the safety teams, you know, of yesteryear that got upset and left got upset mainly over the fact that they didn't have enough compute to either steer the product roadmap or do the checks that they wanted to or the experiments that they wanted to. So let's.
- [34:57]
B
And it's also like what, what Microsoft, you know, remember, like this is what OpenAI's main complaint about was, was with Microsoft, like that they're not getting the resources that they need and they feel like they're, you know, Microsoft's working on their own projects and why aren't you being fully committed to helping us? And, and you know, so now it's moved internal.
- [35:14]
A
Right. And let's talk a little bit about what this device is because we, we did get some interesting detail about what it's going to be. So this is from the FT report. It's roughly the size of a smartphone. Users communicate with it through a camera, microphone and speaker. It is designed to sit on a desk or a table, but it can also be carried around by the user. The device is always on rather than triggered by Word or Prompt. Its sensors gather data through the day that would help it build its virtual assistant's memory. One issue is ensuring the device only chimes in when useful, preventing it from talking too much or not knowing when to finish the conversation. An ongoing issue with ChatGPT. It's interesting. Sort of reminds me of the. The friend.compendant. yeah, for sure.
- [36:02]
B
Friend limitless. There's a bu. There's a few others. Right. And obviously humane Rip and. And rabbit, which also feels like decades ago that. But yeah, there it does again. This is. It feels a bit obvious that this is the path that they were gonna go down. I. I feel like this is. And, and I'm somewhat biased because I've like, long written about this, dating back a decade where I thought that there would be like, you know, these vocal computing devices, not just Alexa, but like, yeah, on the go and, and always on. You type things. And it hasn't played out exactly that way yet, but I do again, I feel like it's just a matter of time until someone nails these. And I do think that when chat, when OpenAI rolled out GPT4O and alongside that they did the voice element of it, which obviously they fame infamously got in trouble with for the Scarlett Johansson voice, which is what mainly what people remember now. But if you remember back then, you know, it was about a year, a little over a year ago, I think it was. It was just an insanely impressive, like, voice technology. Right. Compared to what the state of the art had been up until then. And it's been, you know, getting better. And now obviously Gemini and everyone else has it integrated within their systems as well. But it feels like Alexa and Siri and the first sort of versions of these, these types of vocal computing devices, they focus so much on like the individual devices and the sort of rudimentary utility stuff. I just triggered all of my devices in my office here. But. And so once, once they did that, it sort of held back from being able to do like the newfangled versions of these. Obviously they couldn't have known at the time where LLMs would go and how this would. How this would evolve. But it feels like with that GPT4O rollout all that time ago and now the continued evolution, the voice computing element is there where it needs to be for a device like this to work. But the hardware part is now the hard part, which is always the hard part. But, you know, companies like Amazon or Apple or Google can figure that out because they have their massive companies at scale. Question is, if OpenAI can do it with the team that they acquired with Jony I've's team and the team, most importantly that he brought over from Apple who have done this at scale for decades.
- [38:21]
A
And this brings me to this idea that they're all sort of, all these products are consolidating in some way. Like we just named a bunch of companies OpenAI, Amazon, Google, Facebook, Apple, they're all, it seems like they're all building the same thing. And, and so the hardware might be the tough part but then you sort of will, you'll eventually probably differentiate based off of the assistant inside. And this was your take about what's, what's going on with, with the OpenAI device compared to the Apple's Apple device that we're going to see. So, so you write, just imagine a world in which jony I've introduces OpenAI's first hardware, a small screenless digital companion for your life at the same time that Apple unveils their next hardware. A pair of glasses with cameras on your face and perhaps a screen in your eye. One runs on Chat GPT, the other on Siri. That would be quite the dichotomy and.
- [39:17]
B
It'S sort of angling that way right like these, the timelines are sort of starting to line up especially with what we're talking about with delays. You know, potentially It's a open AI's device and where we know that Apple from the, you know, most recent Mark Gurman report is, is dropping sort of Vision Pro work to, to go for full steam ahead on these, these SM S's to compete, you know, presumably with Meta and meet them in market and there's the whole subtext thing going on with Johnny. I've being, you know, perhaps upset about the world that he helped usher in right with screens and, and everything else with the iPhone. And the fact that not only is Apple now, you know, going full steam ahead on, on glasses with cameras, but if they're, you know, again the reports are that they're going to go down the path that Meta just did with, with the screen in the eye. Like now we have screens in our eyes and, and Jony I've is making, trying to make advice with zero screens and so that is a very interesting butting, butting of heads. I, I would, the only other thing I would add to that that I didn't write about in there but I would be, I'm slightly concerned that there's too much focus on it being a screenless device perhaps. Right. Like I'm not saying that it needs to be another smartphone certainly and I do think that there's probably a world and I think there's increasingly a world where people are maybe you know, happy to move at least a bit away from, from being reliant on a smartphone. But I'm also worried a little bit that it would be harder to use with zero screens. You know, like there's sometimes when you just need a screen for something to showcase versus doing it orally or, or via another mechanism. Now maybe they could say well they'll send a notification to your phone if they, if there's something they need to show you or something like that. But I'm a little worried that it's, it's, it's trying to do it just to make a point about anti screens at the same time. The other, the flip side of that would be a screen as everyone knows is the biggest battery draw of any of devices. And so like if they could do it without a screen, they can do a lot better in all day battery life and maybe weak battery life.
- [41:23]
A
I think they're all going to go screen. I remember when Amazon announced the Echo device and it was so cool because it glowed at the top and talk to you and we had Panos Panay from Amazon, the head of devices and services on the show in March and he beseeched me to go get the Echo show with the screen and said the experience wasn't going to be the same without. And I think all these devices will end up getting there and it's just kind of interesting how they're all converging in their own way. Let me see if I get this right. You have OpenAI pushing Apple on the Assistant device. You have Meta pushing Apple towards glasses. You have Amazon getting pushed by everybody to have a better Alexa. You have Google kind of standing out there to the side but it has Ambient Gemini and is also working on glasses and Amazon apparently is Amazon on glasses. Is the, is. Is the tech world just is. Are all tech products and services just converging to this, you know, eventual yet to be named singular interface with assistant inside.
- [42:37]
B
That actually reminds me of one, one sub point of, of my piece which was while I just said, you know, the, the negative side of like the no screen thing. I do think it's interesting that OpenAI is trying to go down at least a little bit different of a path like you won't see presumably unless there's success from it from someone be it OpenAI or someone else. I don't think you'll see Apple do like yeah like a screenless AI device I don't think you'll see except for the glasses.
- [43:08]
A
They could do the glasses with no screen.
- [43:10]
B
That's true. And it sounds like the first version will be, but I mean like a sort of a secondary piece of hardware. Yeah, like exactly what is rumored to be, you know, what OpenAI is working on. It feels like they're the only ones besides those startups as we mentioned, that are going down that path, at least right now. Now you could say Alexa got him to trigger it again is sort of like that in a way. Right. But it's obviously meant for to be plugged in and you don't take it with you. But I think you're right. Obviously everyone is, is sort of coalescing around these same ideas, including the startups all working on the things that we noted. And the answer to that is like, I just think everyone is still trying to figure out like what the next device is despite all of the AI hype and everything that we've been talking about. And obviously everyone, you know, we're real believers in it, but we're unsure about the financing models for it. But everyone believes in the technology angle of it, but it's still at the end of the day, people are looking for the next device because they think that this can be the next thing that becomes like the massive platform in the next iPhone. Right. And I'm still of the mindset that's the more boring mindset, that's like nothing's ever going to be the scale again of, of what the iPhone achieved and instead it's going to be all of these devices, a bunch of different devices, including glasses, including your smartphone, maybe you know, potentially including a, a chat pin, walk around thing, desktop device, whatever it is, and, and several other things, you know, along those lines. So yeah, I think that they all just feel like they need to have these initiatives. You mentioned Amazon because they, as you noted, like they sounds like per reporting, they do have a glasses project as well, but they hired the Jay Allard, which was a name I hadn't heard in years and years out of Microsoft to work on, you know, these new. Some sort of new ideas around devices and AI and whatnot.
- [45:06]
A
Yep. So they do have. So here's a plug for, for an upcoming episode. Panos is going to come on. I just recorded with him in New York following the Amazon Devices event and we do talk about that workshop that they have where they're trying to figure out the future of the new AI device. And we had a really interesting interaction. It's kind of Fun where I said, well, what do you think about pendants? And he goes, oh really? Trying to go all over the place here. And he goes, pendants. You know, so anyway, I encourage people to tune into that because it does seem like they're at least considering a pendant there. But you know what's going to be interesting? It's going to be the design philosophy of all these companies. Like you wrote in a post about how Amazon said let's just get it out, let's get Alexa out there after they saw that Apple, you know, didn't release or get as much of Apple intelligence out there and they got killed for it.
- [46:01]
B
Right.
- [46:02]
A
I actually, I want to end with a couple things. First of all, I want to get your take on well, second of all, I want to get your take on the Vision Pro before we leave because we've said on a couple times, a couple times on the show that Apple has deprioritized that so we should at least take a moment to talk about what that means. But before that, it's just very interesting to see Meta pushing Apple in these directions. And so I'd be curious to hear your perspective on where the. It's really a blood feud in my perspective, maybe that's too strong of a term, but where the Meta vs Apple product war stands and how you think each one of those companies product philosophy will be an advantage or a disadvantage as they move forward trying to ship these AI devices.
- [46:48]
B
Yes. So I've written about this a number of times dating, dating back a few years now. Basically, you know, once, once Apple tried to destroy Meta's advertising business, it felt like that was the kickoff of a real cold, you know, not even cold hot war. In many ways that's, that's been happening I think at the highest level for Meta. You know, Zuckerberg has said it multiple times from, on Joe Rogan's podcast no less, like talking about how they feel like they're in a just terrible position because they're beholden to Apple and to a lesser extent Google in terms of the smartphone, you know, needs for their products. And it's not just for distribution of their apps anymore. It's now for things like their glasses where they feel like they can't have the same level of connectivity that like an Apple watch does because they're not first party hardware, you know, and they, and they can't tap into the, the iPhone the same way that Apple itself can with, you know, low level Bluetooth and all these other things that they use for connections. And so I think that Meta has been trying basically anything to make, you know, something that could again replace that smartphone as the next computing device. And again, Zuckerberg has not been shy about talking about that. I'm super skeptical, as we just talked about, that anything will ever get to that scale. But I do think that Meta has done a good job and an admirable job by the way of sticking with it. Right. Like lots of companies, famously Google, you know, with Google Glass and, and their own VR efforts, like they just do something and then when it's not working, they back away. And, you know, there's something to be said for that too. But a lot of these newfangled devices, it just feels like they will never work unless someone just continually pumps in the billions required to sort of be in the right place, right time. Right. So much of this is based around timing as so many things in the world. And I just feel like a lot of these devices have never worked because they're not right place, right time, and people give up before they would ever be in the right place. Because you can't really time the markets right. You can't know, know that it's going to be the right time for it. You can, you know, do as much as you can to try to make it the right time for you. But, but again, Meta, to their credit, support almost maybe a hundred billion dollars now into their efforts, you know, in terms of the Quest headsets, but now also with the glasses and, and all the other projects and the AR glasses that they're working on. So credit to them for, for still trying to do it. I'm still a little bit, bit skeptical about what scale they can get to with that and if they can ever fully truly break. Yeah, the, the need to be paired up quite literally with Apple and, or Google and it feels like it's going to be a long time at the very least until they can get to that spot, that part. I haven't used the new glasses yet. Curious if you have. But like, you know, they're impressed. It sounds like they're impressive. It sounds like the technology is impressive, but it also doesn't sound like the type of thing that's, you know, know people are going to leave their smartphones behind anytime in the next five to ten years.
- [49:54]
A
Yeah, I, I haven't used them yet, but I, I actually want to pick up on a, on a item that you put in a recent post about it that I think Exor Exotica actually balked at the idea of naming the latest generation Ray Bans because they were bulky and then Meta sort of floated some money towards them and now all.
- [50:10]
B
Of a sudden, right again, another, another tech company that owns an ownership stake in a, in a public, public company.
- [50:16]
A
Exactly. And, and so I think what I draw from that is what might be playing in Meta's benefit here is they don't have a standard. You know, they don't have to worry about decades of a design legacy that they have to uphold to make something feel Meta ish. They're just like, all right, if these are ugly as hell Oakley glasses that you know, talk back to you and some people like them, let's just do it. And that sort of, I guess, spaghetti at the wall type of approach I think is a very interesting contrast to the, it has to be perfect way that Apple is developing things.
- [50:52]
B
Yeah. And one riff off of that idea is I do think the smartest thing that they did was partner with Exor Exotica. Because the, the biggest strength that Apple obviously has beyond all their in house expertise is also the retail channels. Right. Like they can move product because they have stores around the world and they can ship something and it be in, in every single country basically around the world aside from I guess Russia and maybe a few other places. But, but Meta has, has one store or a couple stores. Right. And even Google has just a few, a handful of stores. Like no one else has that capability. So Meta, smartly, it's not just the design of the glasses which obviously mattered too, like you know, and getting to use the Ray Ban brand, but it's also the, the foot, the retail footprint that they can get from that partnership is, was a hugely smart thing. And it's something that I've been wondering about when thinking through what Apple does with Smart Glass is like do they also do some sort of partnership there? Not because they need it from a distribution perspective, but more because like Meta has now set a standard in a way that look you're wearing the same brand that you wore for your regular sunglasses or your regular glasses, but now they have these, these tech enabled things and like Apple, as great as they are at design, like, like do people, you know, like there's the difference between wearing an Apple watch which is, you know, well designed and stuff, but it sort of resides under your shirt. And there's a very different thing between something that's right on your face all the time and obviously Google ran into that head first, pun intended, you know, with, with Google Glass. And I think again Meta was very smart in, in out of necessity. They had to Be smart about that partnership. And so what does Apple do there with. Do they have to cut a deal to do a partnership in a similar vein?
- [52:39]
A
Okay, a couple more minutes left. Let's just do the requiem for the Vision Pro. I remember back in the day hearing Apple make lofty promises about the future of spatial computing wasn't the Metaverse, it was spatial computing. That's right, it was big vision. And now the Vision Pro has been a disappointment. But not only that, it seems like it's going to be deprioritized within Apple's suite of products. And you even suggest in one of your recent posts that this might be it for.
- [53:11]
B
Yeah, and you know, German and others have said like, look, they're still committed. They're obviously about to release a new version with just an upgraded chip. It's apparently going to have a slightly better headband, which is nice because the current one is not any good. But, but beyond that it will just be a chip upgrade. The one that you're talking about of course, is that they're, they're gonna at least put on the back burner the idea of a Vision Air or just the Vision without the Pro part of it, the slider, the smaller, lighter and potentially less expensive, you know, version to actually scale this thing. And I think that they just probably looked at the market and said like, obviously where we're at right now with the Vision Pro is not where we want to be. But even if they did a quote unquote Vision or Vision Air and say it was like even $1,000, do you think it's going to be a massive scale, you know, hit device? I don't think that it would be. And so I think that they're being realistic about that. And you know, as you're talking about Meta sort of showed them what the market is like, at least for the foreseeable future in terms of a wearable on your face. And again, others had tried, we talked about Google Snap, of course was, was an early mover with Spectacles and still trying it. But Meta really sort of nailed I think both sort of a technological perspective, ease of use the part, partnership angle and timing again. And I think the, the for the Vision Pro, when we talk about does it ever, ever go anywhere? I don't think that it's going to go anywhere anytime soon. And so it's a matter of if Apple does hit with something like the smart pair of smart glasses and the next version of it. I, I wouldn't be shocked if they do just back burner the notion of the Vision Pro, it's still there as like a legacy product. It's all roped into this vision quote unquote vision umbrella of like, well, this is just the, the high end of what we always wanted to do and, and you know, we're working at it on multiple fronts. It's too bad because I do feel like they've started to come into their own. It's with some of the content and they're getting better about it, but the device itself is just way too bulky, way too heavy. Everything about it is just inconvenient to use and it's just the time the market completely wrong and obviously they couldn't have known AI was coming as in as hot as it did. But like, like, you know, they, they went all in on this product and they probably shouldn't have done that.
- [55:29]
A
I know we said to start this episode that the past month or 30 days or whatever it's been, has been crazy. And so it's hard to possibly capture all the action that's happened. And, and I, I think that we'll have the, the same thing repeatedly month, month on end because none of this seems like it's slowing. And mg, it's just great to be able to break it down with you every month. So, so thank you so much for coming on the show.
- [55:56]
B
Of course. Thanks as always, Alex. And I'll see you in a month, see what's.
- [55:59]
A
All right, see you in a month. And folks, remember to check out spyglass.org you can set up sign up for MG's newsletter there, become a pro subscriber like I am and get the added, added articles and added insights, which I highly recommend. All right, on. On Wednesday, uh, Anthropic Chief Product Officer Mike Krieger is going to be on if all goes according to schedule, talk about the company's latest model and, and discuss the fact that we've gone from Sonnet 4.0 to 4.5 in a very short amount of time and whether the AI product release cycle is speeding up. Thanks again to mg. Thank you all for listening and we'll see you next time on Big Technology Podcast.