CapEx, Clauses, & Compute

This month's discussion between Alex Kantrowitz and I on his Big Technology Podcast was all about money. I mean, I think it always technically is on some level, but this one was all about CapEx, "The Clause", and Stargate.

Specifically, while overall Big Tech CapEx spend may be on the verge of hitting $1T combined soon, Apple... stays the same at their roughly $10B a quarter. And actually, they might be down a bit this year versus last. They are ramping R&D spend, but so are all their competitors. It's just such a binary bet that they don't need to fully control AI – at least right now.

And it could pay off in a few ways. If this plays out like Apple vs. Google Maps, where Apple had to rip off the band-aid and then try to catch up, as painful as it was for Maps, it seems like they might not be able to do that at all with AI. But if it's more like Google Search, obviously that has worked out extremely well for Apple – to the point where Google is paying them billions to use it. Meanwhile, Microsoft spent billions to try to compete and, well, Bing is not Google!

That's sort of another way of framing the notion that owning the iPhone (the distribution and eventually edge compute platform) may matter more than owning the models. The reality will undoubtedly be more nuanced, but it's an interesting, if insanely risky, bet by Apple when you just look at what their peers are doing.

Further, there is a way to look at this through an even simpler lens. Perhaps Apple realized they were behind but also realized that in order to try to catch up, they'd have to spend billions – likely hundreds of billions – and while xAI and Meta have done that, it hasn't really worked out for them. At least not yet. So maybe Apple, in a moment of corporate self-reflection, simply realized that wasn't a smart play. And perhaps they'll focus on the future of the industry with robots and what not. Again, we'll see!

They are clearly signaling they're about to start ramping spend on something. Might it be more CapEx spend, or might it be M&A? (Sadly, Anthropic may be a bridge too far now, but Disney?!) Could it be going deeper on chips with Johny Srouji, maybe even further down the AI path to eventually compete with NVIDIA in ways? Perhaps only John Ternus now knows...

WWDC in a month should be fun though!

Speaking of fun, "The Clause" is dead, long live some new clauses in an agreement between OpenAI and Microsoft. Is it as simple as the latter realizing that their massive ownership stake in the former will be far more valuable if they're actually able to get their models on all clouds? Perhaps. But that's obviously also somewhat risky for Microsoft given that those other clouds compete with them.

But perhaps those new clauses which give Microsoft first-look rights at OpenAI's newest tech is good enough. Are we entering a world with movie-like windowing for AI – "See it first in Azure, coming soon to AWS"?

At the very least, OpenAI and Microsoft get to avoid more lawsuits – with one another no less - just as the trial in which they're both defendants against Elon Musk kicks off...

Finally, the Stargate situation remains a sort of fascinating encapsulation of the AI data center strategy overall. Certainly for OpenAI, which keeps moving the goal posts ever since they announced it at the White House.

It was a great PR moment to announce something which... basically hasn't happened. At least not as it was laid out. And while OpenAI may be getting gaslight-y about this, clearly the strategy has shifted. Owning and controlling data centers may have sounded good – or even necessary, to combat Google – 18 months ago. But now it just sounds insanely expensive, slow, and risky. Better to offload all of that to partners. At least for now.

Perhaps if and when OpenAI is public, they can revisit. As presumably they'll have more avenues to raise the needed debt, versus relying on SoftBank, Oracle, and NVIDIA (before the latter just tried to quietly get out of doing that). Regardless, CFO Sarah Friar is probably happy not to have all of this on OpenAI's balance sheet right now...

At the same time, you could certainly argue that OpenAI spending like "drunken sailors" on data center deals over the past couple of years has given them a real advantage over Anthropic at the moment. And now we're seeing Anthropic have to take their shots and break out their pocketbook to spend under pressure. Money wins.