Demis Steps Down, Apple’s Memory Problem, Microsoft’s Clever Trick

22:29But also, something that wasn't really talked about a lot is they said the services growth is going to be a little softer than usual. So what do you think about that? Like, it's a balance, right? You want to make your money from the iPhone. It's clearly more than 50 % of your revenue, most important product. But services is a big portion of your valuation. It's a big part of your growth. And you kind of can't mess with that.

22:49M.G. Siegler:Yeah, that was surprising. And that obviously surprised Wall Street a bit because that's been the growth story in the face of sort of slowing growth overall with the iPhone just by nature of law of large numbers elements. It's at such a state where it can't grow so massively every quarter anymore. They've been relying on services, you know, because they have this this massive install base and upselling them on different things. And so it was a bit surprising to see that it didn't grow as fast as they were expecting it to this past quarter. And they're blaming some interesting things, one of which is sort of the regular the regulations around the world that are coming into place to have to change the app store, which is fascinating, right?

23:29M.G. Siegler:Like because that obviously has not happened so far in the United States. But there's obviously a lot of pending lawsuits and pending antitrust just, you know, cases going into place that could upend that and all these various things that are on hold right now with what happened with the Epic, Apple versus Epic trial. And so Apple seems like they're hinting that there could actually be a real change to the App Store model. And you and I've talked about this a little bit, and I've been writing about this. Like, I wouldn't be shocked if at some point one of John Ternus' first moves is to change the App Store model a bit.

24:03M.G. Siegler:It didn't happen at WWDC, obviously. Would it happen at an iPhone event? I don't know. That seems sort of like maybe a distraction from the actual products themselves. But at some point, again, just given the regulatory pressure and given now that they're acknowledging sort of how it's impacting the actual business, I wouldn't be shocked if they do start to move on some of that, changing perhaps the overall split, tweaking it a little bit of what the cut is. But they also blamed other things like just to slow down, it sounds like maybe more secular in the gaming element, because games are such a huge part of the App Store.

24:36M.G. Siegler:That's like the main thing that sort of drives a lot of that App Store revenue is the gaming purchases and in-app stuff. And so, you know, they had a few other things to blame and, you know, China and everything else. But still, it was surprising that it's not growing as fast, especially because all you hear about and all you see them touting now is Apple TV and all the big Hollywood stuff and F1 partnership and MLS partnership and how well all these things are doing. But obviously, those things aren't doing well enough to sort of buoy what the real part of services is, which is the App Store.

25:11Yeah. For John Ternus, do you think it's kind of an interesting sort of strategy here for him? Like if he was to come in and cut some of these fees or give in to some of these EU pressures, he'd either be seen as somebody like a conciliatory CEO or somebody that gets pushed around. And I think that Apple has really worked hard to make sure that no one pushes them around. So I'm kind of curious how you think Ternus would play that.

25:38M.G. Siegler:I doubt that Ternus sort of gives in to some of the EU stuff if for no other reason than Donald Trump is still president and that would look bad upon him, right? And that would make for awkward conversations potentially with President Trump's buddy, Tim Cook, who's still going to be around, potentially managing that relationship. So I wouldn't expect them to do anything in the short term with the EU other than continuing to fight and sort of putting out the back and forth rhetoric around what side is right. And I also just believe, I do think that they think that the EU is being unreasonable with a lot of their demands to open up the platform in ways that would potentially make it less secure.

26:14M.G. Siegler:I think that there's some points that the EU has that are fine. But I overall, I mostly side with Apple on that stuff. Maybe that's my American bias kicking in. But I live in Europe, not continental Europe, but the UK. And so I see some stuff of, I just came back from actual continental Europe, and I see some sides of it on arguments on both sides. But I do tend to side with Apple on that stuff. So I don't see them moving on that. But I do think, yeah, the App Store, I think you bring up an interesting notion that yet can turn us be made to look weak, like he's going to backtrack on all of the things that Cook and everyone else has put in place for so long, dating back to the Steve Jobs days.

26:55M.G. Siegler:I think the better framing of it would be, you know, I'm entering in with a blank slate, and I'm going to revisit everything. And there's no sacred cows, and we will look at everything from first principles and whatnot. And again, we've likened it in the past to Satya Nadella coming in in those early days and Microsoft and releasing the iPad version of Office and that sending a signal more than anything else, right? So it does turn us to do something to send a signal. I wouldn't expect it really again at that iPhone event, but maybe sometime next year. We'll see. Okay. You know, conspiracy, whether my conspiracy plays out or not, we're going to probably see this come to a head pretty soon because in the earnings report last week, Cook did say that Apple did have a stockpile of memory and they've been drawing it down and the buffer seems to be dwindling.

27:44At least that's what he says. So get ready to pay more for Apple devices. Bottom line.

27:49M.G. Siegler:Yeah. Yeah. And, you know, I think one interesting thing that has been talked about a bit, but obviously Cook has been lobbying hard to try to get Apple to be able to use memory chip suppliers in China. Right. And it's super contentious because obviously everything going on with with AI and China and all the back and forth and trade negotiations and whatnot. But if If Apple can actually get an unlock there from – it would have to be from Trump, presumably, to be able to at least use memory chips for iPhones built in China to use those Chinese memory chips, that would be a massive unlock because all of a sudden they could just push all the rest of them to the rest of the world, most notably, of course, the US.

28:30M.G. Siegler:And they don't have to worry about that. But can really either side be viewed to be okay with that at this point given how contentious the US-China relations are at this moment? I don't know. But obviously, Cook is pushing for that. Yeah, I wouldn't be stunned to see that go through. But that being said, I don't know, I kind of, I mean, it's just memory. But do you really want that type of hardware in your in your iPhone? I don't know. But before we move off of Apple, let's talk about Apple as a service, which you wrote about maybe instead of, you know, you buying the iPhone at a higher price tag, you just subscribe to it monthly.

29:06M.G. Siegler:Yeah, so and this is what I was alluding to, obviously, in what we were just talking about. This, I think, is strategically very smart of Apple. And I feel like this is sort of the perfect strategy to launch with these iPhone price increases potentially coming down the pike. Because again, I laid out, I thought, like, Cook gave the quote to Wall Street Journal, raised the prices of other Apple devices, and then the third leg was going to be eventually raising the price of the iPhone itself, obviously the most important thing. But there was a one more thing happened before the iPhone price, which is that, as you're noting, basically, they have this new service called Apple Upgrade, which is effectively a way for you to pay a monthly fee for an iPhone.

29:49M.G. Siegler:Now, they've had this somewhat before with the iPhone Upgrade program, which I've used in the past and is okay, but wasn't great because it uses a third party banking partner, rather than Apple itself sort of to be able to pay for these things. But this thing is a little bit different. It's definitely bigger in scope. It's not just the iPhone. It's also for Mac. It's for iPad. It's for Apple Watch and maybe other things down the road. It's US only right now. But given that their partner here is Klarna, which is a European company, you would expect that probably this goes at least to Europe, if not worldwide, you know, eventually sooner rather than later.

30:22M.G. Siegler:And I think it's going to be vital again for all the stuff we were talking about with the price increases, because this is the sort of the most optic easy way to negate those. And you can basically see Ternus getting up on stage at the iPhone event and saying, flashing up the screen that they always do that shows the prices of the iPhone. And when people see them all of a sudden$200 higher than they were previously, you know, you can imagine maybe some oohs and ahs in the crowd of, you know, seeing the little sticker shock, even though they all knew it was coming. Still, it'd be a little bit of a sticker shock.

30:52M.G. Siegler:But what if instead now they have, or you can pay, you know,$19.99 a month or$30 a month for an iPhone pro model. Um, you know, if you're interested in using our Apple upgrade program. So again, I think that this is sort of a little bit of, um, of a smart, very savvy, uh, certainly marketing move, but also, yeah, just a way to alleviate the potential pain of, of having to raise these prices. Um, and I do think a lot of it is optics, but I also think it will help, right? Like people have proven that they're okay, sort of, uh, paying a monthly fee if there's nothing, you know, there's no upfront cost.

31:27M.G. Siegler:And, and over time, they will potentially pay less, but they won't own the device, but then they can buy it, you know, straight up for the remaining price. And so I think it gives people optionality to be able to opt into sort of getting that expensive device in the iPhone and the Mac and iPad, iPad and whatnot. And so I think that'll be especially key potentially for this$2 ,000 plus iPhone ultra slash fold. Yeah, I love that. I mean, would you pay$30 a month? Like, does that sound better? Or$2 ,000 right out of your pocket right away for your iPhone to fold? I don't know. I think people will stomach that$30 price much more than the$2 ,000.

32:05M.G. Siegler:Yeah. And when you compare it to what, you know, cell phone bills are these days, right? If people are already paying$50 to$100, you know, potentially for that service. Like, this is a whole beautiful device that you get to use those services. And arguably the most important device probably in most people's lives. And so, you know, in terms of management and day to day stuff. And so I think a lot of people will be into that. And I do think it like obviously it points to and this is sort of the point of the piece that I wrote. Like I've been long sort of looking for Apple's equivalent of Amazon Prime, you know, basically the all encompassing subscription service.

32:40M.G. Siegler:Obviously, it would look different. Apple doesn't do deliveries. But but still, like, you know, they have Apple one, which is which is sort of the software layer of this. But what if they do rope in the iPhone? And now it's here. It's not baked into Apple one right now, but you can see them eventually merging it. And you being like in this whole ecosystem of, of just paying Apple a monthly fee for all of your different services like cloud and, and the iPhone itself. And then, you know, from Apple's perspective, that's basically someone who's, who's even more locked into the ecosystem than they have been by blue bubbles or whatnot, right?

33:12M.G. Siegler:Like you have to continually pay Apple in order to keep your device and keep up to date and everything. And so I think that that's pretty compelling, both from a business side and from a, you know, potentially a consumer side. Okay. I want to take a break and then come back and talk about this clever accounting trick that Microsoft is using to make it look like it's spending less on the AI build out. And then we will end with Google and what's going on with its AI efforts. That's coming up right after this. Summer always changes how I get dressed. I want pieces that feel lighter and more comfortable, but still put together.

33:47That's where Kintz comes in. They focus on well-made essentials in breathable linen and soft organic cotton, the kind of basics you will keep coming back to. Everything is priced 50 to 80 % less than similar brands because they work directly with ethical factories, so you're paying for quality, not brand markup. Personally, moving between AC and summer heat, a lightweight, zip-down sweatshirt from Kintz has been a life changer. Upgrade your everyday. Download the Kintz app for app-exclusive offers or go to kintz.com slash bigtech. Get free shipping on your order and 365-day returns. Now available in Canada and the UK too.

34:34That's q-u-i-n-c-e dot com slash bigtech. There are a lot of things AI can replace. Teamwork isn't one of them. and covering technology every day I'm seeing firsthand just how quickly the relationship between people and AI is evolving. For me, there's constantly new information to research, interviews to prepare for, and ideas I'm working through with my team. What I like about Notion is that AI becomes part of that process, where we work from the same context instead of sitting off on our own. Notion is the AI workspace where your team's knowledge, projects, and agents all come together in one place.

35:07Teams using Notion move faster, cut friction and cost by consolidating tools and stay aligned. That shared context means AI can help answer questions, surface information, and handle some of the busy work while the people on the team can focus on the thinking and the collaboration that actually matters. That's the future of AI at work that makes sense to me. Technology strengthening teamwork rather than replacing it. Learn more about how Notion can support your business at notion.com slash big tech. That's all lowercase letters, notion.com slash big tech to try Notion today. And when you use our link, you're supporting our show.

35:40This episode is brought to you by AvePoint. Everyone's racing to roll out AI right now. Co-pilots, chatbots, agents doing real work. But here's the part nobody loves talking about. All that AI runs on your data, and most teams have no single way to see it, secure it, and prove it's under control. That's exactly what AvePoint does. For 25 years, they've been the trusted layer beneath the world's most demanding data, now extended across your entire AI estate, your data, your cloud, and the agents acting on your behalf. It's how more than 28 ,000 organizations deploy AI with confidence. So innovation scales without scaling risk.

36:20It's a single platform instead of a pile of tools, bringing security, governance, and resilience all together. AvePoint, the unifying trust layer for AI. Learn more at avpt.co slash big technology Podcast. That's avpt.co slash big technology podcast. And we're back here on Big Technology Podcast Friday edition with M.G. Siegler of Spyglass. You can sign up to M.G.'s great newsletter at spyglass.org. Highly recommend it. M.G., let's talk a little bit about Microsoft. So last year, last week, Wall Street really praised Microsoft for not raising its capex at a time when it seemed like everybody was raising the capex.

37:03And then I read your piece on it. And it seems like Microsoft's restraint is actually just a creative shuffling of the numbers. Tell us a little bit more.

37:13M.G. Siegler:Yeah. And so I want to caveat this. I don't think that Microsoft is doing this to deceive the market. I do think, though, that they are happy to benefit from the optic advantage of this. And basically, what you're talking about is they were able to, in this quarter, they decided that they were going to basically change the way that they account for their data center leases. And so historically, they've actually considered them to be 15 years. And that basically meant that they were, you know, taking on the cost of a 15-year lease. and because of different accounting rules, which I don't fully, you know, I'm not an accountant.

38:00M.G. Siegler:So take this with a grain of salt. But from my understanding, and obviously from AI's understanding and some of the reporting understanding, it seems like there's accounting rules which allow them to, if they were to change that 15-year cycle, which is what they did for at least some of their data centers to change it to a 25-year cycle, that basically takes some of their spend off of what has historically been in the CapEx section of their spend and put it into OpEx instead. And so basically, this allows them to still say, and they were up front about this in the earnings, they basically said, we're not changing our outlook, we're not lowering, and we're not, as you noted, we're not raising, but we're keeping it the same.

38:44M.G. Siegler:And that was weird to people because at the same time they said that, it looked like they actually had spent about$15 billion less than what they thought they were going to, at least at the top of the range that they were guiding to before. And it's like, whoa, did Microsoft, are they pausing? In the past, it looked like they had been doing that before when they were sort of shifting and reallocating stuff as they were trying to unwind from their open AI partnership. And so was Microsoft taking a pause again? Was this a sign of the end of the bubble and whatnot? But no, instead, it looks like it was just this accounting change, which basically means, again, they're spending the same amount overall.

39:19M.G. Siegler:So it's not like they're spending less. It's just that it changes it from going under the CapEx line to moving it over to the OPEX line. And so again, optically, I think that they were happy to make it look like, look, we're being more prudent and we're not sort of upping it like Google is and potentially like Amazon is and like Meta is right now. And instead, we're going to be more mindful about where the spend is. But it still looks like they're going going to be spending on par with everyone else. It's just that change. And it was even more confusing because when you looked into it more, as it turns out, actually, to their credit, Microsoft was actually super conservative in this regard.

39:58M.G. Siegler:And the other rest of their peer group had long ago sort of changed from a shorter time horizon for this data center leases. And so they were already taking advantage of that. Now, I do think Microsoft and probably Amy Hood, their CFO, smartly had this timed well. And you always want to do this, it sounds like, at the end of a fiscal year, which is what they're at. So they timed that well. But they also knew that like, look, the single biggest point that Boyle Street seems to care about right now is capex spend. And we have a way to alleviate the pressure off of that. And again, they were in a position to be able to do this.

40:31M.G. Siegler:And in some ways, you could argue that they were actually probably, you know, reporting above what they were actually doing relative if you if you compared it to what the peer group was doing. So yes, it's it's sort of a little bit of a funny accounting thing. But I think it's a savvy move and I think it gives them a little bit more headway. And again, you saw the stock. I think they had the single best stock day gain of anyone in history. Yeah. They beat out NVIDIA a few years ago. So that's incredible. Okay, but MG, when I think about where the AI story can crumble, this is definitely one of the places.

41:04And I'll explain why. So as we talked about, they said that the useful life of a data center is going to go from 15 to 25 years. And it gets them this ability to reclassify that expense. Okay, it sounds good. But you and I both know that GPUs are not going to last 25 years. Now, on the low end, people say they'll last three years, right? So you're not going to have the H100 potentially go for a decade. Some people say it lasts three years. Maybe it does go for a decade. But this idea that you're not going to replace those chips for 25 years, it just doesn't hold up to any scrutiny at all. And of course, the chips are 50 % of the data center build out cost.

41:44So how do you put that on a 25-year scale? Yeah, 25-year scale just makes zero sense to me.

41:50M.G. Siegler:So there's levels to this. So 25 years, from my understanding, is the actual facility itself. Now, inside, servers are a whole different matter. And this has been a big point of contention across the board, both from big tech and certainly with the NeoClouds. Because, as you know, they're all spending tons and tons of capital and all this circular financing is going in. And a lot of it breaks down to, yeah, exactly what you're pointing to. Like, what is the actual lifespan of these servers? And they range. Yeah, like you noted, from three years, some are six years, some are seven, I think. And Michael Burry, famously from Big Short, has sort of talked about this as a reason why he's shorting a lot of these companies, because he believes that they're sort of doing funny math on a lot of these schedules.

42:39M.G. Siegler:For Microsoft's thing, I do think that they can get away with, again, this is what the peer groups have long done. Amazon and Google, in some cases, are far longer. I think Google has some that are 40 years. Amazon might, too. But again, Amazon has different warehouses, right? They have all their logistics warehouses and whatnot. So it's a little bit hard to dive into and break down at a granular level what they're allocating towards each. But I do think that they can make a credible argument, certainly relative to the market, that they should be allowed to do it over this 25-year schedule. They'll just say, like, look, yeah, we'll replace the servers within them, but we'll have the actual facilities for that long.

43:17M.G. Siegler:Now, if something changes drastically with AI, maybe they're in a bad place where they just don't need them or something else happens and they need to be repurposed or something like that, or they have to sell them off. And who knows how that can play out. That's 25 years a long time, certainly in tech and even more so in AI. But to your other point and question of like, how much of it do they have to bake in because so much of it is NVIDIA cost, basically. and yeah I mean all of that is is a little bit of a gray area I feel like right now. Yeah now last week we saw some crazy numbers with the cloud growth 82 percent from Google 30 something percent from AWS 43 percent from Azure.

44:01There's a discussion that has been had that like yes these companies are seeing pretty significant growth from cloud but most of that is just OpenAI and Anthropic, which have raised hundreds of billions of dollars just investing it back into cloud. So if that VC money were to fall away, and some of it, by the way, is coming from these companies themselves, like you wouldn't see the growth. Like for instance, Amazon's going to invest$50 billion in OpenAI. And so like, are we shocked that they're having 36 % growth in cloud? How much of this is just kind of like growth that we're seeing on the back of AI investment?

44:40and how much of it do you think is real?

44:42M.G. Siegler:It's a good question. And it feels like it's one that Wall Street isn't really answering right now or asking even right now, because as you saw, like with these numbers, so Google dropped, Meta dropped, and a lot of that is spend related. And with Meta, it's more nuanced. It's related to, you know, that they don't have a cloud, you know, whereas the other players do. But Microsoft, as we just talked about, you know, had an all-time surge and Amazon had a massive surge too. And a lot of that seemed to be based off of the fact that it felt like Wall Street feels like that they're starting to get a return on some of this AI investment.

45:15M.G. Siegler:And to your exact point, how much of it is actually circular? How much of it is just one or two companies that they're all heavily invested in, the two largest AI companies versus random, long tail, both mom and pop businesses, but even smaller Fortune 500 companies actually using AI. And you don't hear a lot about that right now, but I think it's an important and key question because, as you note, if anything were to happen to slow down Anthropic or OpenAI, that has a cascading effect that sort of happens across the industry, and it will definitely impact. Certainly, it impacts Amazon. Certainly, it impacts Microsoft, Google as well, but they have some of their own, obviously, dependent on Gemini itself too and TPUs and whatnot, Meta would be in a good position then, I guess, given that they don't have the cloud right now.

46:08M.G. Siegler:And Apple, of course, would be sitting most pretty of all. But yeah, I think that that's a good key question that no one is focused in on right now, because growth has just been crazy. When Wall Street sees those numbers you're talking about in terms of year-on-year growth at the relative size that they're at, it's just incredible. And so why would you want to question that, Alex? Yeah, I mean, I am surprised that there hasn't been a deeper discussion about this because obviously, like, it's not all the spending. Like, you add up the growth in all these companies and it's not just entirely OpenAI and Anthropic, but it is a sizable percent of it.

46:46And the fact that we're not, like, talking about that in terms of, like, you know, oh, like, Amazon going from 17 or 18 percent quarterly growth to 36 percent. to me, it just kind of boggles the mind that it's not a bigger discussion topic.

47:02M.G. Siegler:But and, you know, you bring that up just jogs my memory. That might be exactly why Satya Nadella is taking the stance that he is right now with Microsoft, where it seems almost like antagonistic about his big partners. They still own 25 percent plus or whatever of OpenAI. And now they own, you know, a few percent of Anthropic, too. and yet Satya Nadella like in the press and in his blog posts and stuff just keeps talking about you know the need to be Switzerland and diversify away from just those two big behemoths again which he owns massive percentages of and why it could be exactly what you're talking about like he knows that if if they're too beholden to those two if those if something happens there and inevitably even if down the road like those two actually do make it out and you know break out on their own to become two of the new, you know, tech behemoths that are public companies, they will eventually obviously just have their own clouds.

47:54M.G. Siegler:They're not going to be relying on Microsoft and Google and Amazon indefinitely, certainly not to the extent that they currently are, you know, and they're obviously working to build out their own data centers right now that they could potentially have their own clouds on. And so maybe, you know, it's as simple as Satya Nadella sort of trying to get ahead of this and also, you know, reading the tea leaves with how upset everyone is about pricing and everything like that. And so, yeah, positioning Microsoft to potentially take advantage of the short term, but really be in a better position for the longer term.

48:26M.G. Siegler:And one more thing, to your exact point on the Amazon 50 billion thing, that's one of the craziest things about all of this, right? Like there was a report that everyone knew that Amazon had an option to put into up to 50 billion, but the reporting suggested that the option was basically tied to either open ad going public and or achieving AGI. Neither of those things has happened, as far as we know, at least with AGI right now. And yet, Amazon apparently closed out that$50 billion, that remaining$35 billion tranche, you know, not such a small tranche, to close out. Why did they do that? There's no reporting on it yet.

49:05M.G. Siegler:Either they know something that's going really great at open AI, and there's some inklings, right, that open AI is making a bit of a comeback with Codex against Anthropic and that some of the new models, 5.6 and what's maybe coming with Atlas, I think they're calling it as the next model, are actually going to have a real breakthrough and potentially could vault them back to the top. So maybe it's as simple as they saw that, but it is fascinating. I mean, they put in$50 billion in 2026 into OpenAI after they already have the massive investment in Anthropic. And so they are now arguably more levered on these two companies than anyone else, including Microsoft.

49:45M.G. Siegler:When you combine those two stakes together, they just bought, again, they bought, what, 5 % full on of OpenAI in 2026. Sort of wild. Yeah. I mean, either they saw something big coming and maybe they did, or they just made a handshake agreement that that money comes right back into AWS, which is obviously healthy. That's a fair point. Fair point. Yeah. No, maybe a little bit of both. Yeah. I would imagine. Okay. Before we go, let's talk about Google. You know, a couple of months ago, MG, I think we were kind of early on this talking about, you know, did Google just fall behind? Clearly they have.

50:18And I think you and I are both continually puzzled that they continue to fall further behind. They have their next big model kind of stuck in review phase. They're nowhere to be found on these coding agents. You think that you've written that this is potentially culture related. So let's just get your thought about what's going on within that company and how it could fix it.

50:45M.G. Siegler:Yeah. I mean, so with the disclaimer, obviously that I worked at Google for over a decade, but I've been gone for a couple of years now. So I don't have inside information, you know, current about this situation. But again, just reading about it from afar and sort of knowing what we've talked about and watching IO this year and seeing, you know, obviously they promised on stage, Sundar Pacheco promised that they would have the 3.5 Pro model, their flagship model out, you know, in June, you know, a month after that conference. And now here we are in August and it's still not here. And, you know, the reporting now suggests that it's months delayed and it sort of feels a little bit like the llama situation for Meta back in the day, right?

51:22M.G. Siegler:Where it's like they kept promising their, I think it was called behemoth model back in the day. And And it just never came, right? And that's what led to ultimately the reset. Now, I don't think that Google is going to full-on reset and buy like a scale-like company and, you know, boot out the old team and bring in this new team. But I do wonder what is going on there internally right now. And it does feel like, you know, as you noted, that there's a bit of a culture clash and there's been reporting on this, right? Like Google is such a big company and there's so many smart people. But the downside of that is that they have fiefdoms within the company and they have smart people who disagree with one another about how they want to do something like, say, agentic coding.

52:04M.G. Siegler:And so when there's one team over here that wants to do it this way and another team over here that wants to do it this way, probably a third team over here, they butt heads and they're all competing for resources because not only does Google have to have internal resources, as we noted, they have their cloud. And so they have to have the external resources for people to use, too. And so there's clearly sort of this internal arms race going on for competing factions. You would have thought that this was was reconciled, you know, when it felt like that DeepMind basically and Demis Hassabis took over sort of the group and they united DeepMind and Google Brain back in the day.

52:40M.G. Siegler:But clearly there's now splintering factions again, you know, per the reporting on it. And so so where does this leave them now? You know, they might be, you know, they keep trying to talk up the notion that the Gemini force, you know, is already in training. And so that's going to be great. But again, that just sort of suggests that they really messed up with 3.5 Pro. And are we ever actually even going to see it? Are they going to go right to four? It sort of feels like they're going to go try to jump right to four now. And if they do that, four had better be really good. Otherwise, they're going to have a real narrative problem.

53:13Yeah, there's a great piece in the algorithmic bridge, which is a sub stack that says that talks about the actual reason Google fell out, quote unquote, of the AI race. And basically, the piece makes the argument that Google has effectively given up, although I think that might be too strong. But it sets out like basically this dynamic where you have startups and then you have vertically integrated companies like Google. And then you have two different types of AI approaches like the LLM approach and the world model approach. And so basically what you have on one side is startups that are committed to LLMs because they have to make a profit.

53:44They think that this is the way to go. And then you have these vertically integrated companies that are less profit sensitive and they're able to explore other areas. Right. So basically the point is that Google is less profit sensitive. It doesn't have the urgency. It can basically dedicate itself to the field of research that it thinks is going to pay off more. And it seems like Google is just making this bet, like the world model type approach is going to be the way that's going to get it to what it wants to get to, like AGI level AI. Demis has obviously always talked about that LLMs may get them to AGI, but probably there's going to be another couple of breakthroughs needed.

54:25So I think the piece argues pretty much that they're working on those next few breakthroughs and have effectively ceded or put much less priority on LLMs. And that's why they are starting to struggle. What do you think about this?

54:39M.G. Siegler:Yeah, I agree with you, first and foremost. I think it's overstated, but I think it's overstated in a way that's interesting, right? I think like it's a good thought experiments to think through like what, what, why is Google in this place right now beyond sort of potential for internal conflict? Like they have to know, because if nothing else, they've already been through the point where, you know, they needed to reorient themselves and sort of come back in AI. So how did they let themselves fall behind again? you can certainly make a compelling case that it's less that um they don't see it and more just that exactly you know what you laid out there and what the piece lays out that from demis has to be on down he just basically views world models as the way forward as the most important thing and they're willing to sort of take a back seat on llm development um so in order to push forward with that and you sort of saw that with io you know again um where you know demis was out on stage basically saying like talking about the Omni model, which is their version of the world model, and basically talking that up while sort of downplaying every other thing that they were doing that.

55:41M.G. Siegler:Now, did he do that because again, 3.5 pro wasn't ready and he wanted to get people excited about what's next? Or did he do that because that's really what he's excited about in every interview that he's given, um, you know, and talk to you and talk to others about this, right? He's talked that he's excited about world models and, and, and obviously he's not alone there with Jan LeCun and a bunch of others sort of all now working in this space. Now, the one other element of this piece, which I thought was super interesting, is that it's not just that, you know, Google could be going down the world model path as they view that as the path potentially to get to AGI or the next phase of AI development.

56:16M.G. Siegler:But it's also that if Anthropic and OpenAI are battling, they could be correct too, in that they could find that where they are in LLMs right now is actually going to trip over into the recursive self-improvement stuff, the RSI. And basically that would unlock the ability for the LLMs themselves or whatever, you know, the next version of the models are as they're improving themselves to basically come up with the world models themselves. So people don't have to do it right. Like basically they could, they could trip into RSI and then all of a sudden that would just create this like, you know, uh, potentially exponential, um, growth of, of AI itself to the point where world models and God knows what else comes into play.

56:57M.G. Siegler:And so that is an interesting battle if it is framed that way. The website is spyglass.org. MG Siegler joins us the first Monday of the month, every month, although this week or this month, we're doing it on Friday. MG, it's always great to speak with you. Thanks so much for coming on the show. Likewise, Alex. Talk soon. All right. Thanks everybody for listening and watching. And we'll see you next time on Big Technology Podcast. Athletic Brewing Company crafts award-winning non-alcoholic beers for those who want to be part of every round. With over 185 flavor awards, they're exceptional NA beers that fit your lifestyle and any social occasion.

57:35M.G. Siegler:Summer's full of good times, and athletic fits right in. Go to athleticbrewing.com to have brews delivered to your door, or find them at a bar, restaurant, or store near you. Near Beer. Athletic Brewing Company. Fit for all times. Hello. Look what TJ Maxx dragged in. The Devil Wears Prada 2 is now streaming on Disney Plus and Hulu. We are digital. We are downloadable. We are streamable. The fashion event of the year is certified fresh. Pull yourself together. We have work to do. Critics say it's smart and witty and the perfect sequel. That's all. Get runway ready for The Devil Wears Prada 2 on Disney Plus and Hulu.

58:17M.G. Siegler:Rated PG-13.