Big Technology Podcast - Did Apple Get AI Spending Right?, Microsoft & OpenAI’s New Reality, Where’s Stargate?

Episode Date: May 5, 2026

M.G. Siegler of Spyglass is back for our monthly tech news discussion. Siegler joins us to discuss whether Apple made the right decision to hold off on spending a mid-sized country GDP worth of money on AI infrastructure, and whether the company's restraint will continue under incoming CEO John Ternus. We also cover OpenAI and Microsoft scrapping their "AGI" claude and whether Microsoft should've agreed to let OpenAI work with Amazon. We round off with a state of Stargate and where the risk is in the AI buildout. Tune in for a high-energy conversation about AI and tech's most pressing issues. ——— Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. Want a discount for Big Technology on Substack + Discord? Here’s 25% off for the first year: https://www.bigtechnology.com/subscribe?coupon=0843016b Learn more about your ad choices. Visit megaphone.fm/adchoices

Was Apple right all along to stay away from the AI spending race? And is that about to change? Microsoft and OpenAI are, where exactly after their latest new agreement? And where is OpenAI's Stargate program? That's coming up with M.G Siegler right after this. This week, I'm live at Knowledge 2026, ServiceNow's annual conference in Las Vegas, where Enterprise AI moves from Promise to Production. I'm sitting down with Service Now's president and CPO Amit Zaveri on the platform strategy powering it all.

their people and technology leaders on what AI means for the workforce, the engineering team behind ServiceNow's Nvidia partnership on what it really takes to ship AI at scale and Ultra Beauty on deploying AI across 1,300 stores. These are the conversations you won't hear anywhere else, and new episodes are dropping this week on my YouTube page. Insurance isn't one-size-fits-all, and shopping for it shouldn't feel like squeezing into something that just doesn't fit.

That's why drivers have enjoyed progressives name-your-price tool for years. With the Name Your Price Tool, you tell them what you want to pay, and they show you options that fit your budget. Enough hunting for discounts, trying to calculate rates, and tinkering with coverages. Maybe you're picking out your very first policy. Or maybe you're just looking for something that works better for you and your family. Either way, they make it simple to see your options. No guesswork, no surprises.

Ready to see how easy and fun shopping for car insurance can be? Visit Progressive.com and give the Name Your Price Tool a try. Take the stress out of shopping and find coverage that fits your life on your terms. Progressive Casualty Insurance Company and Affiliates. Price and Coverage Match Limited by State Law. Welcome to Big Technology Podcast. It is the first Monday of the month, which means M.G Mondays are back. And we're here with M.G. C. Lurter to talk about everything going on in the world of tech.

We have a lot to recap in the bigger picture talking about Apple's AI spending, whether it's about to ramp up KAPX expenses. are some signals that it might be. Microsoft and Open AI have a new, new agreement. It seems like there's another one every week, but this one might actually be a meaningful step towards letting open AI work with everybody else. And then of course, we have an update on Stargate. What is it? Did it actually ever exist? And is Open AI meeting expectations there or never actually going to build those data centers that it promised? So a lot to talk about today. We might even cover the Musk, Altman case, because we have some breaking news on that front. And it's great to welcome you back to the show, MG. Good to see you. Great to be back, Alex. And we joke about this every time

that there's like an unending stream of news. But it feels like the fact that the cook and Ternus transition happened, what, like a week and a half ago or something like that, maybe not even. And it feels like it was months ago, which is wild. Like, that's how wild the news is moving at this pace right now. Yeah, we're like fully into the Ternus era. It's like when there's a new president-elect and then all of a sudden you go from like calling this guy like by his name to like president-elect or president, whatever it is. And it's like that feels weird for like a couple weeks. And then all of a sudden it just settles in.

And we're definitely there because we're getting some indications of where Ternus is going. And to set it up, I think that we should talk about this great piece that you have about Apple's capital expenditures. And of course, to our listeners and viewers, it's no secret that big tech has been spending a lot of money on CAPX. We might see a trillion dollars in AI CAPEX this year. But you look at Apple and their capital expenditures are actually going down. And I don't think I've seen anyone illustrate just how stark the divide is,

as well as you've done in this recent piece called Apple's binary bet. Apple, and you say, is on track to spend 9 to 10 billion in CAPEX this year. Compare that with meta, which just raised its expectation to $145 billion. And it's clear that Apple just is not in the foundational model game. I always thought that this was a liability. But now maybe it's a strength I'm open to it. What do you think about this divergence? So this quarter and this time of year was interesting.

specifically because four of the technology, of the big tech companies, all reported on the exact same day, right? I think it was Amazon, Microsoft, Meta, and Google Alphabet all reported at the same time. And so it was, you know, some people called it earnings apocalypse. And Apple was the very next day. And so it was a unique sort of a lining of the stars moment where you could more easily sort of do the comparison of the Kappex spend. And it's still a little bit tricky because they're on different physical calendars. Microsoft famously is different from, I think, the rest of them. But still, the fact that Apple, who again reported the day after those other four, they came in with their Q2 numbers.

And as a part of that, they don't break out CAPEX in the way that the others do because they're not spending, as you noted, like to the extent that the others are. So I don't think that Wall Street sort of demands it. But they did give the CFO did give an update on the half year since it's again the end of Q2. And yeah, the number was insanely low compared to obviously what everyone else is spending. And it's, as you noted, it's basically it was three to four, maybe four and a half trillion, sorry, billion. And what it ends up being is that it's probably going to come in below what it's even was last year.

And then I went in, I went ahead and just did the obvious thing, which is porting those numbers and asking, you know, in this case, Claude to go back and look at the past five years of, of cap-expend and charted out and sort of normalize again for those different fiscal time periods. And it looks, the chart is fun because it's basically everyone obviously starting around 2022. So obviously that was, you know, a little bit after the chat GPT moment, starting in 2022, all of those companies basically started to really ramp up. Amazon, as you know better than most, was already ahead because they have huge cap-ex spends related to their, you know, core business of obviously delivering goods and they need warehouses and whatnot. And then they had

AWS, of course, which they also needed, you know, capex spend for. But the others, including Amazon, too, all really start to ramp up. And Apple, it just stays the same. It's always like in this line that's around 10 billion. One year they jumped up a little bit from like seven. But this was like five years ago from like seven till nine. And everyone's like, whoa, what's going on there? And it's like, I mean, that's hilarious to look back on now compared to where it. things are. But again, Apple's in the same place. They're down slightly this year. And it's, it's just a wild discrepancy when you compare it to, you noted what Microsoft is spending, but Amazon and Google, I think, are both guiding towards 190 billion now. And that's,

this is good updated every single quarter. So they very well could spend well over 200 billion for what ends up being the year. And Microsoft will be right there. Meta is a little bit lower, but meta, as you know, is getting hit left and right as from a stock perspective because they don't have the obvious returns because they don't have a cloud business, right? They're saying it's doing well for their ads business and certainly their numbers look good, but it's not as straightforward as it is for the other players, it seems. They don't have an AI product either, meta. No AI product. They sort of just rolled out their new models. And yeah.

And they're what are they done? Nothing. They've done nothing. So they're baking it into everything. And again, they were bringing it into their ads product. But again, back to Apple, even compared to meta, like Apple, if they do 10 billion this year and meta does 140 or 130, 140 billion, I mean, that's just an incredible, incredible discrepancy. Regardless of what you think about, you know, the future and AI, it's just like, it's wild how much lower it actually is right now. Yeah.

And you have this paragraph that I highlighted from your piece that sort of says, that puts this. and relief. Everyone thinks AI is going to be a core bit of technology that they each need to control, lest they be beholden to someone else. This is exactly why Microsoft shoved open AI aside, despite their early, extremely prescient bent on the startup that remains the leader in the field. Even though Microsoft owns 25% of the company with IP rights and access, they're not spending billions to build out AI on their own. So let's tackle this question. Because I also was on this path of like, well, if this is such a transformative technology, you would be best to have your own models

and not rely on somebody else. And even if you, you know, made a great deal with Google, let's say you got Gemini for free, which you won't, you're going to pay. You have less control over it because you're a step removed from the technology and it's going to be much more difficult to productize it. But the other side of the, the other side of this argument is,

it seems sound now, which is basically that like we are seeing this, you know, if Apple is right, you would say that they've seen that all these models are going to commoditize. The compute will commoditize. We talked about this a little bit on the Friday show. There's going to be a price war. Apple can just wait until the technology matures. The price war happens and then implement it. Maybe they can even implement it on device.

How much credence do you give? If that is the bet, and it sounds like that. That might be the Tim Cook Applebutt. How much credence do you give to this idea that they might be right? Yeah, I mean, hence the title, the binary bet. And that's, I think that there is a chance, right? I mean, obviously, I think in the short term, it's a lot more straightforward for that to, you know, if that plays out that way, because it's so relative, everyone believes that it's relatively early in this, in the AI game. And so there's a world, there's a world in which there's a few avenues.

where it works out for Apple. And it's not just, you know, the lack of spend on training these models. I mean, you know, we've talked about it before. What if LLMs don't end up being the end-all of AI that gets us to AGII, right? What if you need to sort of even not fully pivot, but you need to basically then ramp up sort of these other massive infrastructure arrangements in order to build up, you know, world models and whatever else comes next down the pike and robotics and and everything else in order to get us to what many would consider to be the sort of holy grail of AI. And so if even just in not choosing to participate in this LLM buildout, maybe Apple's wise in that they're realizing or lucking into, even if you don't necessarily believe in their vision of this

to date, that it's not just going to be LLMs, that the LLMs are step one of many steps to get us towards the future of AI. And so maybe Apple's bet is like they're okay sitting out the LLM part of it, but then they're going to come in hard on the robotics part of it, right? Like the actual physical robots part. But then you would say like, okay, but they still need models or something to run on those. And so even if they're getting unique data because they have hardware out there in the world, they still need, you know, to be in charge of their models.

But you could also say like, yeah, they're making a bet that they can partner on it, much like they've done, you know, with things like Google Maps, which ends up being an interesting analogy to that, right? Because famously, of course, they had to go against Apple at one point and launch their own product because they wouldn't agree to the terms that Google put out there in order to keep that deal going. And that's the downside, right, of being beholden to someone else at the end of the day, and this is Tim Cook Doctrine, right? Apple wants to build their own technologies. They don't want to be beholden to anyone else. This dates to the jobs days and even, you know, very early iterations of Apple when various software players, Adobe and everyone else didn't want to make

software for the Mac and for other earlier versions of Apple products. And so Apple decided that they needed to ultimately be in charge of that. That's not to say that they always do that. I mean, the most famous counter example is Google search, right? Apple decided they didn't need to build a search engine. And that's worked out great for them, right? Microsoft famously has spent billions and billions building up Bing. And what have they gotten out of it? I mean, it makes money now, but it's not, not Google and it's not, you know, something that I think is absolutely vital to Microsoft. And so what did they ultimately get out of those billions spent and all that time and all those resources and engineers and whatnot being devoted to that? And so I think you would say that Apple

made the right bet in that regard by partnering and getting billions of dollars sent to them, right? Famously, as we've seen now play out over the antitrust lawsuits, getting paid by Google to use Google search. And so maybe Apple in their head is thinking this, will play out similarly to the way that Google search has where we can basically just ride off of this and we don't need to own this and it won't end up coming to hurt us. I think most people, at least right now, would think that that's a little bit too risky of a bet to make. Again, speaking of what we already talked about with what they're spending in the Kappex number, but you can see why they're at least trying, you get back into why they're trying to do this

or why they're okay doing this. Yeah, let me throw out like the where this goes really well and where this goes really poorly for Apple. The way I could see this going really poorly is that AI becomes so powerful that we just sort of converge on this one sort of, you know, Star Trek computer-like AI interface where, you know, the open AI device really works. The assistant is amazing.

We access all information like we would through, you know, a screen that is effectively powered by AI and probably some audio-only or perceptory devices like headphones and a pin, right? So basically in that world, the only thing that matters is how good your AI model is, how it can operate all the programs and be your agent for you. And then Apple doesn't really have this opportunity to lease that technology because let's say it's open AI that's built the best version of this,

or Google that's built the best version of this. You're like sitting there as Apple, you're this like, you'll become a legacy device maker without access to the technology. That is the most important thing of any device. And I think that's a real risk for Apple. And that's sort of like it's far further away than I think many people anticipated. I think a lot of people expected AI devices to be more farther along here. And we're going to definitely talk about that on the show in the next couple of weeks.

Like, where are the AI devices? That is that and Apple could be in some serious trouble because it will effectively have to beg open AI or Google for the technology that will enable that set of devices experiences. That's where it could go poorly. where it goes well for Apple is basically, you know, unlike, let's say, the software battles that it's had in the past or the times that's been forced to build its own technology, there is a powerful and almost equal open source AI development movement going on right now. They can definitely take advantage of that. And then if you look at those who have tried to develop proprietary models outside of that open source world, actually isn't as simple as we thought it was. Like we used to think it was spend,

uh, spend a hundred billion dollars and you have a world class model. Well, if you look at, if you think about it, uh, Amazon, obviously they have the infrastructure,

but Amazon Nova has gone nowhere. Most people don't even realize that Amazon has a foundational model. Uh, Google's done a good job of it. Microsoft after separating from open AI has not been able to do it. Meta has not been able to do it. XAI has not been able to do it. So it's not just throw money at the problem and build, build your own AI.

And then it would be almost the ultimate wise move from Apple to be like, we are not winning that race. We're just going to rely on licensing or open source. And we don't think that this universal AI device moment is going to happen. What do you think is more likely of those two scenarios or will we see a combination of both? So first, first of all, just to hit on the point you just made, because I think it's a good. one and an important one, that Apple might just be, it might be as simple as Apple recognizing, like, look, guys, we're behind an AI. The reality is, like, in order to try to catch up, we would have to spend like Elon has been doing, right, and building these massive data centers

in a hurry, and as Zuck has been doing, but it's tens of billions and it's ultimately hundreds of billions of dollars. And as you just noted, I've written about this, too, like the throwing the money at it, it seemed like when Elon was ramping up the, quote, colossus data center that maybe he could just all of a sudden jump back into the race because, you know, he was able to build a data center that could train an XAI model in record time and record speed. Ultimately, though, it didn't work. Like, it just didn't matter, right? And so Apple might be looking at those, that situation, the meta situation, obviously

famously, you know, acquiring scale AI and bringing in this new team and trying to get up to speed as fast as possible. And that's sort of not really working yet. And they just may be realizing, you know, in a, in a, in a, self-recognition way that, look, we're not going to be able to do this. No, we might have all the money, all the resources in the world, but that's not all that matters. We don't have the right talent, right? And what we would need to do in order to get that would be like, as we've previously talked about, acquiring Anthropic or something wild that would take hundreds of billions,

if not trillions of dollars to potentially do. And so, again, they could just recognize, like, we're just not going to get there. And so what's the point of burning these billions and billions of dollars, whereas, again, we can go down the path of making sort of these later bets, that it's not just LLMs and that we can sort of ride out the coattails of the others who are doing that well for the time being while we work behind the scenes, of course, and build our own stuff internally, but behind the scenes. But to go back to your question, which one is more likely? I do still think that there is a very good chance that Apple ends up a quote unquote winner of the AI movement simply because of the iPhone, because it's the best device out there to run models potentially and to run these applications. And I do think that that's the major fear, obviously, of Meta. Zuckerberg is famously sort of complained, you know, about the fact that they're beholden

to Apple with any devices, obviously the smart glasses, but anything else that they want to do. And then I think it's a problem, obviously, for Open AI, if and when they come out with their device, with famously Johnny I've designed device. I do think that there's a real scenario in which the iPhone ends up being the key device for AI and at the very least, Macs fallback to, you know, Mac computers, which, Obviously, the Mac Mini and the Mac Studio, it seems like they can't keep them in stock because of all the fun things that the fun work that people are doing with the agenda capabilities now. But anyway, Apple makes the best hardware. And so the fact that they can't keep doing that, they would still benefit from it.

And I think it's a matter of timetables, right? It's a matter of, is this a decade-long situation where the iPhone remains, you know, and I framed it in writing about Ternus? It's very possible that Ternus's entire tenure is still dominated by the iPhone. Like Apple is still the iPhone company by the time his 15 years is ending, right? And again, the world that you're laying out, that first scenario is that sort of the opposite of that where it's like, yeah, this open AI device, these meta devices, and even potentially some new devices from Apple come in and take over that mantle from the iPhone. I still just think the iPhone is probably going to be the central computing hub and maybe even more so.

And again, that's a big part of that bet. and why Apple maybe can afford to make that type of bet. Quick diversion on this. I mean, if that is the case, right, that the iPhone becomes the device of record for AI, that seems kind of bearish for Open AI. Now, this is like a caveman brain take, but like if you can serve AI models that are just as good or almost as good on device,

then maybe spending $1.3 trillion on a hardware or an infrastructure buildout is not the best bet. And I mean, there's, you know, yeah, that's, that's a, that may be the caveman argument of your year. That's not the, it's not only cavemen making that argument right now, right? Certainly. And, um, and there's, there's bets, tangential bets to open eyes, since you can't bet directly on it since it's not a public company, but there's tangential bets being made left and right now, you know, in order to, to believe that that's sort of the way that it plays out. A couple other points to, to your, your question.

One, I do think that there is also the world, you talked about how, yeah, the, the, other players are having a hard time at the frontier. The open source question is, is interesting, right? Because we just saw the new deep seek get launched out there, right, V4, after you and I had a good podcast a year plus ago when the deep seek moment, the quote unquote deep seat moment happened and it seemed like everything was changed and really what has changed in the intervening time. I mean, certainly there's been a lot of talk about all of the open source movement and how you build and how you train these models and can you do it for much cheaper and all the talking points over the past year. But the fact that Deepseek now has the

new version out, it's gotten very little buzz, certainly compared to what, you know, the last version was. And now the sort of the papers are starting to come out sort of comparing it. And it's definitely everyone acknowledges that it's behind still the frontier. But the real question is, is it actually diverging where it feels like the frontier is accelerating faster than where the open source sort of community is at right now? And I think that that's like sort of an interesting way to look at this as well. Like, what if it's just not playing out that way that open source is doing it? Now, the counter to that would be, we'll get to a point where, yes, like, the gains just keep getting more incremental on the frontier models, right? And open sources is quote unquote good

enough. But who knows? I mean, this could this in some ways, in an odd way, this might go back to the old iPhone versus Android and Clayton Christensen sort of debates of like, is, you know, good enough, good enough like, you know, the commoditized hardware and commoditized sort of features, are they, do they end up being good enough? And with the iPhone versus Android, it's the case where the iPhone is making just as much of the profit share and is more popular, you know, in many metrics than it's ever been before, even though we have the proliferation of all these other devices, thanks to the Android ecosystem out there. And so there are worlds in which it doesn't play out that way. The last point I would make is just the notion of Apple's other big bet. And I think a lot of

people have been dancing around this and we maybe have danced around this a little bit and you hit on it just now is that on device models could be good enough for almost everything that we're going to need and with what point does that do we hit that and how many years out do we get from there because if that ends up being the case and again apple has the best hardware that makes it a really really different sort of conversation i feel like yeah oh my goodness the apple chip versus the invidia chip will be a very interesting the ultimate battle. Could you imagine if it goes back to that?

Those two companies famously have a sort of a history where they don't like each other. They wouldn't let Nvidia GPUs for gaming run in Apple systems and Apple famously is the one company not buying up as many Nvidia GPUs for these training runs left and right as much as possible.

And even when they're doing them, they were using potentially TPUs and whatnot. But anyway, yeah, if that becomes that ultimate thing. And there's lots of talk, right? That NVIDIA wants to potentially, obviously build CPUs, which they've been doing for the flip side for inference and everything else that they're going to be needed for in the future of AI. But there's even talk that they would build their own, you know, potential actual PC systems, right, to go after sort of more of those markets and presumably go after more of the hardware front end because right now, obviously, they're the

biggest company in the world, but a massive player on the backside of all this AI work that's being done. But if they really want to be a front facing company and brand again, like, do they go for it and start to build their own machines, you know, with their capabilities? Yeah, that'll be interesting. All right. So let me just deliver the punchline to the beginning of our conversation here, which is, you know, is Apple making the right bet on, you know, with spending much less on CAPEX. obviously new CEO coming in and we have signals that he believes in some part that they've been going too light on the AI spending. First, when he came in, when John Ternus came in, first thing he said was this is the most exciting time to develop products, highlighted artificial intelligence right at the top.

Then, last week, he was on Apple's earnings call and signaled that Apple would take a somewhat different approach financially than it did under Cook. I'm just going to read a little bit from Bloomberg, an article by Mark German about this, and it's a little bit financial. I'll try to go through it at a high level, and then we can talk about what it means. German writes, under turnus, buybacks, or dividend hikes could potentially slow in size or frequency based off of the conversation that he had on the earnings call, and that may alter Apple's longstanding pitch to investors. We'll generate enormous cash flow and return it to you.

Now they have the optionality to do it less. Evercore analyst Ahmed Dharani said. In other words, Ternus could spring for a blockbuster acquisition or dial up spending on R&D. He could expand the company's AI infrastructure, something in Silicon Valley Peers are doing at a furious pace. I think that these things don't happen by accident. I don't think it's like, let me, you know, create the option for myself. Obviously, we know. Ternis has been preparing to take over the CEO position for probably a year or longer.

you had to have been thinking about this. This is one of his first major financial signals. Maybe the AI spending is coming. At the very least, I feel like this is optionality, right? Like this gives him the ability to basically go out and spend far more to a far greater extent than Apple has to date. Because to that point, like this is a big moment. Apple has a famous policy in place to try to be, you know, cash neutral, as you're noting.

They don't do, they don't actually do that because they generate so much. so much profit that they still hold a bunch of money on their balance sheet. But remember, there was a time, I guess, in the Steve Jobs days, where their cash balance was ballooning to the point where it was just way out of line, out of ban, versus what everyone else had in the industry, because they were generating so much profit and holding it all and not doing major M&A and not doing stock buybacks. Tim Cook changed that. And I do think that that's a big part of why the stock has done so well, right, under him. And it's become a $4 trillion company because he's, He does dividends and he does stock buybacks.

And basically they're signaling now, not necessarily that they're going to stop it entirely. And I think that there's still some mechanisms in place to do some of the buybacks that have been promised already. And so they're going to continue that, at least for the somewhat foreseeable future. But they are sending a signal. And it's an explicit signal. They wouldn't say it if it wasn't something there that they might not, they don't have that policy anymore, that they want to remain sort of that cash neutral position. and instead they could start hoarding cash again, and that could be used for a few things, as you noted. Either they do M&A, they start to do some M&A, they start to do more R&D.

And by the way, in the binary bet piece, I note, like, they did hit a record in R&D spend last quarter. And that signals something interesting that they're spending and, you know, looking into more and more future technologies. At the same time, all of their peer group is also spending more. You know, everyone's, of course, focused on the Kappex because the number is a, is huge. much, much greater. But still, they're all spending record amounts on R&D. And Apple is in that group, too. They're spending records amount. And this might help them to further accelerate that. They could also sort of retire some debt, I guess. You know, there's other more wonky sort of financial reasons why they might want to do this. But to me, again, I think it's sending a signal.

I think they're doing it now for a very good reason, which is that even though Ternis has now been, you know, announced in place, it's still Tim Cook running the company. And so this sort of gives them a little bit of cover. So it's like when, when, uh, turnus actually takes over in, in the fall, he's not, his first order of business isn't to, it's a blow up Apple's capital, Apple's capital sheet and, and freak out Wall Street. It's like, look, guys, we're guiding towards this. Like, there might be a time in the future where we're going to need to spend more. And again, to your exact question, I think that it's a reasonable assumption to make that they're going to start spending more on CAPEX2, even if they're not going to train models, they're not going to train an LLM.

inference and doing other things that you need to do. We already know that they're doing work behind the scenes. And another key part of this, remember, of this entire announcement, was Johnny Sirucci, who's Apple's chips chief and obviously a very vital part of the company, given that they announced on the same day that they're announcing a new CEO, they announced that he is a new title, a new C-suite level title. And I don't think that those two things are unrelated. And I do think that the fact that he's working on, you know, arguably the most important thing

that Apple has done over the past decade, which is their silicon. And if they're also, as has been sort of whispered about and rumored, if they're working on their own chips and able to do inference and be able to do more AI workloads, they need KAPEX in order to do that. And so again, all these things potentially are related. And I think this is signaling that more spend is coming for sure. Yeah, I agree. I think Apple's going to make some big acquisitions in turn as his first year for him to set the tone that it's going to be different. And I think that's good. I'm in favor of that. Two more Apple things before we go. First of all, you know, looking at the calendar, we are going to, you know, right now, we're saying maybe Apple is making the right move.

But even still, they have a, they have a terrible AI product, like nothing. We're going to find out in a month about whether all this like, you know, sort of, what would you call it? Wise restraint is actually wise or whether they're just being stupid and they have no answer. to build these products because new series supposed to come next month. Yeah, and I mean, I think it's wild that it's that quick already that it's happening. At the same time, like,

if you believe the reporting by, you know, Mark German, it's basically like, they've still delayed it a couple of times, right, from when it was going to roll out in potentially beta form to get it out there early. And, you know, maybe it's just wanting to keep it back in order to make a big, wow moment at WBC

when they do actually launch these things and unveil them to the world like, look, we finally did it. Syria is finally fixed and we can make a big to-do of it. You hope that's the case. It's also possible that it's the opposite of that, that they're having still some troubles integrating it and the edge cases, right? Because much like Amazon, as you brought up with sort of their own work and Alexa and whatnot on AI, dating back to the Siri days, earlier Siri days too, they've had a hell of a time.

And Google has as well sort of transitioning from their original models, Google Assistant to Gemini, to Bard and then to Gemini, and then Alexa to Alexa Plus, and Apple, Siri to, I assume it's still going to be called Siri. Maybe it shouldn't still be called Siri, but it's still like, you know, there's a world in which Apple is having a hard time sort of making that transition, just as the others have before them. I think Alexa Plus is the best of the bunch, I have to say. I really, I really like using it. It is good. You know, relative, I think, to where it was. add it's a good sort of conversational AI. Do you use it, though, for any of the things that Amazon, you know, hopes to use it for? Do you shop with it, for example? Sometimes. But I was shopping with old Alexa, but what I really use it for is to settle arguments in the house. When we're debating things like my wife and I, I think I've said this before, but I'm American, she's European. We have

strong opinions on which system is better. So we'll just sort of get into a debate and then all of a sudden and we'll just summon, we changed this name to Ziggy like it has some other name, so we'll summon Ziggy. And that tends to settle the argument for us. And I mean, that sort of speaks directly to what, you know, the iPhone point earlier. It's like it sounds like you use it basically because you have devices around your house, right? And it's a good conversational AI. And so you can use it instead of taking out your phone and bringing up chat GPT or Claude or whatnot. And I think that that's, again, a case in Apple's court.

Like, what's the best AI device? It's the one you have on you. Exactly. The one in your hands of your pocket. It's not the one that's not the theoretical one out there. It's the one that billions of people already have. That's right. Okay.

Last thing. And then we're going to go to break and then talk about Microsoft and opening eye. I think we can spike the football for a moment and you in particular that when this turnus talk was starting to surface, I think we both felt strongly that there was something there. And there were voices out there that said, nope, Tim Cook is going to stay for a while. And here it happened. the transition is underway.

Yeah. I mean, again, I, we had talked about this and I feel like beyond sort of the reporting, I think financial times, right, ends up looking really good in the way that they framed it. And I think that, you know, that was accurate. And I think it was accurate for a reason. I think that there was, you know, that someone wanted it to be known that Cook was very seriously weighing this decision.

And maybe it wasn't a slam dunk, you know, that it was going to happen for sure. maybe he wanted to take the holiday time to think it through and and make sure that uh you know like unlike bob iger in the past that uh when he handed off that first go around to bob jpec and then covid hit and and the world blew up and it ended up being a really bad situation for disney you know maybe cook also wanted some optionality and just to see you know what what it was but we had talked about it seemed like it was there was some smoke to you know some fire behind that smoke because it just was a perfect time for for him to do it if he chose to do it because of the way that the earnings were going to play out, which they did in Q1, record all-time record earnings

for Apple. And because of the way that the stock was likely to react and it's been $4 trillion now, and because of Apple's 50th anniversary in April first. And, you know, Cook obviously, that was important to him. He made it very clear. And so all of these things came into place. The one thing that I think was downplayed for the most part that I didn't read about too much, but I think was in my mind, at least, a big indicator of this was the sort of board machinations of, and maybe this is just because this is sort of, you know, my world of investing in VC and whatnot on board dynamics where when Apple announced that Arthur Levinson and Ronald Sugar would be able to stay on board after their 75th birthdays, which is the historic step-down time for Apple board

members, and the fact that Levinson was the board chair, that to me indicated, that it was very likely that they were waiting on something, right? They didn't want to appoint a new board chair if Tim Cook was going to come in within a very short amount of time and become that board chair. And I think that that's, you know, in some ways it might be as simple as that, that they basically extended, you know, the contract of those folks and allowed them to stay on, but because they needed a good transition, you know, for the overall company. There's been a lot of turnover, as has been clear over the past many months, beyond just the cook stuff. There's a lot of, you know, of executive turnover and even some on the board. And so they wanted a steady, steady hands.

But you also want steady hands when you're doing a CEO transition. And I think, again, you could read a lot into that tea leaves wise. And I think that that ended up being the case. All right. Let's go to break. We're going to talk about Microsoft and OpenAI doing away with the AGI clause and Microsoft allowing OpenAI to work with any cloud provider, which is fairly big news that I don't think has been played up enough as far as the Open AI story goes. Of course, there's been a lot of stuff in the news. So hard to follow everything, but we'll try to cover this story and also give you the latest on the Stargate infrastructure build out when we come back right after this. Look, if you have a kid in school right now, you know the

drill. What should take 20 minutes of homework ends up taking two hours and usually ends in tears. And every good tutor, well, they're fully booked for months. This episode is brought to you by Brainley. Brainley is an AI-powered personal tutor built by educators, not a general purpose chatbot. It doesn't just give your kid the answer. It walks them through step-by-step explanations so they actually understand the material. It learns how your child learns, diagnoses when they're struggling, and builds a personalized learning path in under three minutes. Available 24-7, there's no scheduling headaches, and it's just a fraction of the cost of a private tutor. Finals are coming. Build your teen's study plan now. It only takes minutes. Go to brainy.com

slash big tech to get 50% off your first brainly subscription with my code big tech. That's B-R-A-I-N-L-Y dot com slash big tech. Most leaders know how work is supposed to happen, but when it comes to how it actually gets done day-to-day across tools, teams, and handoffs, they're mostly guessing. That's exactly the problem Scribe Optimize was built to solve. Trusted by over 80,000 enterprises, including nearly half of the Fortune 500, it gives leaders a live into how work is really happening across approved business apps without interviews, manual process mapping, or extra effort from the team. And because it's continuously analyzing real workflow

activity, the insights stay current instead of going stale the moment a process changes. You can see which workflows are happening, where time is going, and which tools are involved. It automatically surfaces top issues, explains why they're happening, and even recommends ways to fix them with estimated time savings. And importantly, it's built with privacy in mind. So activities only captured in admin-approved business apps and user-level data is anonymized by default. The kind of visibility that used to take months, now it's just always on. If you're ready to stop guessing and start seeing, visit scribe. How slash big tech. That's s-c-r-I-B-E.com. And we're back here on big technology podcast with M-Gis-Siegler of Spyglass. You could find it at Spyglass.org. Highly recommend you go check it out and

sign up for the newsletter, one of my favorite reads. MG, let's talk a little bit about the clause, right? So Microsoft and OpenAI famously have had very interesting and weird iterations of their tie-up, including this one clause in the contract that

if OpenAI sort of reached AGI, then Microsoft lost all rights to its technology after that. And last week, basically what happened was Microsoft and Open AI came to this agreement, where, first of all, Open AI could work with any cloud provider,

and then it immediately ran and announced this deal with AWS. So it's now available for Amazon cloud customers. And I anticipate, and given my conversations that I've had with Thomas Currie and the head of Google Cloud platform, that it will come to Google fairly soon as well. He wants it there. So that's one. But then the second and maybe the weirder thing was this AGI clause is now gone. And basically Microsoft has access to Open AI.

IP till 2032. And that is the state of the partnership today. Is it too simple to read this as Microsoft has been kind of annoyed working with OpenAI and has this huge stake in the company realized that two companies would basically be better as far apart as they could be. And meanwhile, wants Open AI to grow because it has such a large stake in the company that it says go ahead and work with anybody. You're not helping our Azure business as much as we thought you would.

And so we'd rather just kind of get out of here, be as big as big as you can, and we'll just take the money. Nope, not too simple to read it that way. That's the way I read it. I mean, this is fun. It's just following this whole drama and this saga has been fun, especially because they keep insisting that there's, you know, nothing to this.

Like, everything is great. And, you know, we'll get the Sam and Satcha picture tweeted out. every once in a while. But obviously at this point, like there's been enough reports, there's been enough just of these weird deals. The fact that this deal, this new, this new update to their agreement is happening six months after the last one, which was this massive undertaking, right in, and with a lot of changes made so that open AI could clear the path to eventually become, you know, the public benefit corporation and get actual equity holders in the company, and that, of course, would pave the way for them to eventually go public, which they're

obviously still working towards. But the fact that the AGI clause didn't go away in those negotiations was interesting, right? Like, they tweaked it, but they didn't kill it outright. And presumably, that still annoyed Microsoft because at the end of the day, they still had this weird thing hanging over their heads where even though they tweaked it to the point where previously, if I have it right, the Open AI board could have announced AGI and just torpedoed the deal. Basically ended the relationship with Microsoft and they were able to do whatever they wanted. And so they had this unilateral power over the deal. So Microsoft got away, did away with that in the last go-around of those negotiations,

but it's still there. They just had to sort of co-agree. And so it seemed like a de facto going away of it, right? Like was Microsoft ever going to agree to let them go away from their. obligations from a business perspective? Probably not. But now it's much cleaner. Now there's no more AGI clause. I don't know when I wrote about this, I didn't know if to read into that, that if there was anything to read into it, this is the notion that maybe Open AI thought it was coming sooner or thought it was coming later, right? Like that they felt like that they could do this now

because they didn't care. But ultimately, I'm not sure that that matters at all. I think that all that matters is they had a deal that they wanted to do. with Amazon. They had famously taken a huge investment from Amazon, which we had talked about previously. Microsoft almost sued them for because they were going to sue them. And because it was a question of the definition of, you know, like state versus stateless models or, you know, getting into the weeds of what the, what they could legally do. And so that was obviously going to be another headache and battle going forward. And at the same time, Open AI, I think that there's a very real case to be made that part of the reason why Anthropic has seen the sort of incredible surge

by all accounts that they've been seeing is because they have this path where they can sell their models via Amazon and via Google, right, and now via Microsoft as well. And so they have a much, much more greenfield opportunity because they have all of the major cloud providers that they could sell on, whereas Open AI was restricted. and that they could only really do it through Microsoft. And so I think that the two sides, it sounds like Sautja and Sam got together and basically hashed out a simple, at least relatively speaking, more simple agreement, get rid of the AGI clause and allow us to run our models anywhere.

Oh, and by the way, the most interesting thing to me about this entire announcement was while Microsoft no longer has to pay revenue share back to Open AI, Open AI still has to pay it to Microsoft and they have to pay it even when they're using AWS or potentially Google in the future. And so beyond what's your point about Microsoft being incentivized with the, you know, with their equity holding, which obviously they care about and want that to be as big of a company as possible so that their percentage of that big company is bigger. The fact that they're going to be making more and more money off of their chief rivals in the cloud potentially, obviously help grease those wheels and make that happen, I feel like.

What's more valuable to Microsoft? I mean, it's hard to know exactly, but having Open AI grow, let's say, 50% bigger and having your 25% stake, you know, grow in that same nature. Or if you believe, because I think in your recent pieces, you've stated Open AI has the best models. It certainly seems like in terms of released models, GPT 5.5 is the best on the market right now. I think, you know, maybe being the only place, making Azure the only place you can get those models is just, even if it, I mean, I'm thinking from a Microsoft standpoint, even if it hampers Open AI a bit, it's almost like no matter how annoying Sam might be to you, like going and working with Amazon, that is a tough thing to give up. That exclusivity is a tough thing to give up. And while I can understand how like, you know, physiologically that Microsoft might have been like kind of like to have. hell with it, go do what you want. I still kind of think that it might be a mistake to let Open AI go everywhere. So, and this speaks to, you know, the point of my piece writing about the end of the clause is that there's another potential clause in here, which I think was downplayed

a little bit in that, and I can just read this part of it. Microsoft remains OpenAI's primary cloud partner and Open AI products will ship first on Azure unless Microsoft cannot and chooses not to support the necessary capabilities. So to me, that and is very important, right? Because it's basically saying, assuming that they worded this, like, in very very deliberately in this way, which you have to believe that many lawyers poured over such a statement. This was a statement in the blog post.

Or at least co-pilot. A couple of co-pilots. Maybe co-pilot hallucinated the entire thing, the agreement. But if this is to be believed, it's basically that... Microsoft still has a right of first refusal on these new models and new capabilities that OpenAI puts out there. So say, you know, when they roll out GBT 5.5 for the first time, there might be a world in which they could say like, hey, we want this to be exclusive on Azure for a set period of time. I'd love to know what those terms look like, how long they could actually, you know, keep such exclusivity if they're allowed to. could be something, you know, like, and if they roll out a new product,

if Open A Eye rolls out a new product, they have to offer it, you know, first on Azure as well. And it's basically open eyes only out in that situation is, again, if Microsoft cannot support whatever it is, or they choose not to. So it's not even the fact that they would say like, yeah, but you guys don't have X, Y, and Z capabilities in Azure. And so therefore, we're going to roll this out on AWS. We're going to roll this out on Google Cloud. They have to, Microsoft has to explicitly say, we will not roll out whatever X, Y, and Z is. Now, who knows, you know, if they ever get to that point.

But you could see a world in which, you know, say Open AI and Amazon come up with some new amazing product offering that they're saying is only available, only able to be done on AWS. And Microsoft getting, again, mad because they would say, like, well, look, we can roll that out on Azure too. So I don't agree with your ability to roll this out exclusively on AWS. You know this is going to come up again. Like there's no way this is the end of this conversation and the end of like the animosity between all the parties here. Like because to your point, the fact that Microsoft is okaying

OpenAI, their 25, 27, whatever it is now, percent holder, you know, giant equity portion of the company working with their chief rival, the biggest cloud provider in the world is sort of incredible. And so obviously this is still going to come up time and time again, but there's new clauses to be had here. It seems like anybody who can say the sentence, Open AI would not exist without us tends to get very annoyed at Open AI, whether it's Elon or Microsoft or whatever it might be.

Yep. It's never going to end. And this is, yeah, this whole, when they do a deal, to your point, they're obviously going to do a deal with Google. They're going to be in GCP at some point. And then it sort of becomes, you bring up an interesting point around Azure and Microsoft's game plan. You know, historically, I think Amazon has more run the game plan, but now to the point

that all the clouds are doing it, they all want to offer all the models, right? They're saying, like, we want the customer to choose whatever model they want. We want them to be operating on our cloud, but we want them to choose. We don't care if it's our own model. We don't care if it's an open AI model, if it's an anthropic model. So they're reaching to the point now with Open AI's capable new unlock that they'll all be able to offer basically all of the models. And Microsoft, I guess with these new or smaller clauses, might be able to say, you should pick Azure because even if you can run all of the other players, you know, models on the other clouds, we're going to get the first look at the new, the latest open AI models. And maybe that's enough to sway some companies, you know, to go, you can still choose Claude if you want.

You can still choose maybe not Gemini, but you can still choose something else, some of these other models, Deepseek and whatnot if you want. But we're going to have the first look advantage still, even if we don't have the exclusivity full time. It's almost like movie windowing. It's a weird, it's a weird new world. Right. Now in theaters, GPT 5.5. Before it comes to DVD on Google Cloud. the U.S. in early April.

Oh, my goodness. Yeah. No, there's truth to that. All right. I definitely want to cover this Stargate story because it's one of those big stories where, if folks, if you recall, Open AI announced this $500 billion buildout of AI infrastructure. Elon Musk famously said they don't have the money. And then Stargate, there were reports that like, oh, the Stargate company never formed. and Stargate itself didn't, didn't, you know, get started.

And now Stargate is kind of this like catch-all word encapsulating all of OpenAI's infrastructure. And MGI, you know, let me know if you think this is, this makes sense. This is like a good window into Open AI's infrastructure as a whole. Because I think the conventional wisdom that Open AI was going to go out and build a lot of its own infrastructure. And actually, it's shifted to Open AI has. secured those data centers as opposed to building those data centers. And this has kind of happened in a really quiet way that the headlines in the beginning were very loud and, you know, obviously got a lot of attention.

But what's happened afterwards certainly has not. So take us into that story. Yeah. And it got loud to the point where remember, this was President Trump's first day back in office. He has Elon Musk, sorry, not Elon Musk, famously not Elon Musk there, even though he's a part of the new Doge part of the whole administration. He is not there, but there with him is Sam Altman, Elon's big rival, Masaan, and Larry Ellison.

So the three of them are there alongside President Trump in the White House announcing the Stargate, Project Stargate, which was going to be a $500 billion, which was a massive deal. it looks sort of quaint almost now when you talk about these numbers given where these companies are valued at. But it was a massive infrastructure deal. And obviously it was a big win for the Trump administration. We could argue that a lot of these deals were maybe already in place and maybe they were repurposing things and whatnot. But that's sort of par for the course. It seems like with some of these announcements. But still, it was regardless, it was a great PR moment and a great moment for all these companies. And it looked like a great moment, you know, potentially for America. we were going to have this huge, huge new AI initiative backed by these major companies and including,

obviously, the government. And so fast forward over the next year, it basically went into place that the first Stargate, which was going to be in a data center in Abilene, Texas, which, by the way, did not start out as a Stargate project. It started out, oddly as an Elon Musk project that Oracle was building. A whole range of people were building. Blue Al Capitol is involved with it, a lot of other players are involved with it. But basically, Open AI was able to bring this in with Oracle to make it now the first Stargates. And it was going to go from there. But fast forward again over the next year, there were just a lot of weird reports about deals getting backed out of and nothing sort of coming online in the way that they had hoped. They again, it talked about of the 500 billion,

100 billion was committed right out of the gates and they were going to come out swinging. And again, it was just really slow going. And the subsequent reports had it that like, yeah, well, maybe we're not going to use this facility. And maybe SoftBank was having a hard time raising the debt needed to get this facility off the ground. Maybe Oracle was having a hard time raising the debt needed to get out. Oh, by the way, Oracle and opening out are backing out of this facility, the extension of the Abilene site. And who's coming in? Microsoft, oddly.

And so all of these weird things going on. But you're exactly right. that to me, this basically encapsulates the change that Open AI has made over that time. And I think that there's other plays in here, which are sort of tangential but related, including the big deal with Nvidia, which is another interesting deal, right? We have talked about it previously where Nvidia announced with much, much pomp and circumstance with Jensen and Greg Brockman and Sam Altman that they were going to invest $100 billion into Open AI over an extended period of time.

and then all of a sudden that quietly went away. Well, what was a big part of that deal? A big part of that deal was clearly in order to secure the debt required for Open ad to continue these buildouts. Open ad has been talking about in various blog posts over the time spans since Stargate came out that they needed to come up with interesting and mysterious new ways of doing capital financing for these data center buildouts. And yes, to your exact point, the thought process seemed to be that they felt like they

needed to own these. They weren't going to own the Abilene site because that was already. sort of, again, backed into that was already in place. But going forward, they wanted to own a lot of these. And to me, that just read like they realized that their real threat was Google. Google, of course, famously owns their data centers and owns all of the infrastructure that they need. And so how was open ag going to be able to compete with that? Well, they couldn't as a massively money losing startup be able to raise the debt needed in order to build out these facilities.

And so they were partnering with Oracle, with SoftBank, eventually with Nvidia. But again, all of these sort of came back. and were pulled back to varying degrees. And now the narrative, it seems like, has shifted fully. And I've written about this because it felt like that was an easy thing to read in between the lines of, that they were changing that narrative, that it was no longer going to be Project Stargate building up, you know, these massive infrastructure for open AI, whereas it's going to be now, we're going to figure out how to build up this massive infrastructure individually, all these companies.

And maybe we'll rope them into Stargate. And Stargate will just be the catch-all term exactly as you laid out. now that seems to be the world. And the only thing I would say the other layer of this is the weirdness is related to sort of the evolving PR strategy of Open AI where they now write blog posts that are sort of gaslighty in the way that they're basically saying like, yeah, as you know, we announced this as a, it was always going to be the plan that this was this was sort of going to play out this way. And that's just not the case.

Like if you look at those old blog posts, they were not laid out in the way that they're currently laying out the current Stargate initiative. And you see it in the pullback again of Stargate, not just in the U.S. either, in the UK and all around the world. Like they have these initiatives. And it's all tied to, it seems like it's all tied to obviously the need for them to corral spending a little bit if and when they're going to really try to go public. Yeah. I mean, one thing I'll say is that this new strategy also shifts all the risk to like the partners. Yes.

Through the oracles and invidias. And a lot of the neocloughs that kind of exist like the nebius and the core weaves. I mean, we had core weave on, and the one thing I couldn't, you know, I thought they did a great job explaining their business when they were here on the show. And obviously, I think their stock popped like 12% after their appearance. So folks put your people on big technology podcasts. But, but I think the weakness in the whole story is your whole business is based off of Open AI keeping this promise of building out the infrastructure. And last week in particular, when there was that Wall Street Journal report, which Open AI pushed back on a very, strongly that like Sarah Fryer, the CFO is like, if we don't make more money, we're not going to build more data centers. That caused a real reaction in these partners in the neoclouds. The whole wave of them got hit pretty hard. And that's, that is like if you're looking for areas of risk or weakness in this whole AI story, that's the place I would look. Yeah. And going forward, the question is going back to their old strategy or what seemed to be their old strategy where they

felt like maybe they had to combat Google, right? And Google, again, owns the infrastructure. And what's the reason why you'd want it? Why wouldn't you just want to offload all this, right? Because it'd be great. Well, at the end of the day, like, you're still like, you're paying people in order to do this, this work, right? And if you own it and control it completely, just like we talked about with Apple, owning, owning, you know, the whole stack and whatnot, you're going to have better cost advantages over, you know, the people who don't. And so Google's in a much better position, it feels like, at the highest level of being able to do all of the AI work that they need to do because of their massive infrastructure capabilities,

whereas Open AI just doesn't have that. And look at now what's happening even with Anthropic, where they're scrambling to ramp up on the infrastructure side because I think Open AI is a good point that, you know, for all the talk of the past many months that maybe, you know, I don't remember how exactly they framed it, but that Open AI was spending like drunken sailors or whatnot, you know, like in terms of their infrastructure plans, there was an upside to that in that they now

have the capacity to meet the demands, whereas anthropic, it feels like does not. And so, again, does Google have those problems? No. Does Microsoft, I mean, all the cloud providers are saying that they're capacity constrained, and they are to certain degrees, right, because it comes down to ultimately chips and or power. But still, if we believe that we're going to get to a world where those issues are eventually resolved, Google's just overall owning of, again, that entire stack when it comes to the infrastructure is going to give them massive cost advantages over what Open AI is going to be able

to offer up at that steady state. We're not in that steady state right now. And so it's sort of masked a little bit. But there's a reason why they wanted to get there, right? They're not, are they going to be able to compete on a cost basis with the Googles of the world? And, you know, increasingly the Microsoft's, the Amazon. of the world as we go forward with these massive clouds.

And so that's what it comes down to. And that's going to be, but that's like, that's a few years down the road, it feels like, right? And maybe open eyes public already by then. Maybe they get in a situation where it ends up playing out okay from them because they're now a public company. They're able to sort of, they were able to corral spending a bit, focus. I'm laying out a, you know, a good case scenario for them. And they're able to sort of use different monetary instruments to raise.

what they need, to raise debt and whatnot to build out their own infrastructure. And because of the way that things are constantly evolving, maybe it doesn't matter so much that they don't control the infrastructure at Abilene. And it doesn't matter so much they control, you know, the first few Stargates. What matters is the Stargates going forward because those are the ones with the latest chips and the better inference and able to do the better inference and whatnot. And so there's a world, I guess, that they can do that. But still, it's going to be, there's going to be.

there's going to be so much money needed to get from A to B where that's going to be the story of the next couple years, a few years at least. Most definitely. This hour went by extremely fast. We were talking about potentially covering the Musk Altman trial, but I think when you're on in June, we'll actually maybe have a verdict to discuss. So we'll leave it until June. Talk about things that are moving fast. Like that's going to change 10 times in between now. Exactly.

So we'll keep an eye out on that. That'll be interesting. and again, folks definitely recommend you go check out Spyglass at Spyglass.org. MG, great to see you again. Thanks for coming on. Thank you, Alex. All right, everybody. Thank you so much for listening and watching, and we'll see you next time on Big Technology Podcast.