We’re feeling cynical about xAI’s big deal with Anthropic

Anthropic and xAI announced a major partnership this week, Anthropic purchased all the computing capacity of xAI’s Colossus 1 data center in Tennessee.

In the latest episode of TechCrunch’s Equity podcastKirsten Korosec, Sean O’Kane and I discussed what the deal could mean for xAI’s parent company SpaceX, as SpaceX prepares to go public and apparently planning to disband xAI as a separate organization.

Kirsten did her best to give “a positive view” of the collaboration; after all, it’s a new way for xAI to make money. But she also noted that this also suggests that xAI isn’t doing much when it comes to training its own AI models, and that it’s harder for the company to position itself as a “forward-looking, innovative” company if that’s the case.

Then Sean asked, “Why be positive when you can be cynical?” According to him, this seems like “a big heat check before the IPO.” Yes, becoming a neocloud may be “a more credible business” in the short term, but it is less likely to get outside investors excited in the long term. (And then there’s the environmental lawsuit xAI is facing over Colossus 1.)

Keep reading for a preview of our conversation, edited for length and clarity.

Sean O’Kane: I always like a surprise, especially when everyone is watching [are] on another ball, a great trial taking place. Seemingly out of nowhere, SpaceX and thus its AI subsidiary xAI – which apparently no longer exists, or soon won’t, as we can gather – struck a deal with Anthropic.

In short, the real version of the deal is for Anthropic to essentially take over all the computing power at the data center known as Colossus 1 in Memphis, Tennessee, to focus on Anthropic’s more enterprise-oriented AI products. Much has been reported on how [Anthropic’s] looking for more computing power […] and it seems like an escape valve for them to be able to make this deal and get access to all these computers.

In the short term, for xAI and for SpaceX, they are indeed a neocloud, in the sense that they had to do something with all this computing power they were building, because it certainly doesn’t seem like they’re going to need it for Grok – which, apart from X, isn’t setting the world on fire and becoming the next popular consumer chatbot.

Kirsten Korosec: And we should say that in terms of what a neocloud is, for those who don’t know, it’s the idea of ​​buying GPUs from Nvidia and the like and renting them out instead of using them for their own AI, and training their own AI models.

So this is a different kind of company, and the point that our AI editor, Russell Brandom, is making is that a lot of companies are building out data centers, but if they’re given the choice between, do they rent them out? [or using them to train their own models]they still prioritize using this computing power for their own internal AI model training. I think this is an important point and one that suggests xAI may not be doing much in the way of AI model training [side]

Anthony Ha: Right, and as Sean alluded to, most people wouldn’t necessarily consider Grok to be – not only because it’s known for its rather unpleasant, if not downright illegal, content, but it’s not necessarily super advanced either. Especially when we start talking about enterprise AI, which we’ll get into later in this episode, you don’t hear much about people using Grok for work-critical tasks.

And so the question becomes: how can xAI actually make money? And apparently just selling the infrastructure could be one of the key ways to do this.

Kirsten: And you could think positively about that, right? They figured out a way to make money. But I think when you position your company – in this case SpaceX-slash-xAI – as a forward-looking, innovative company, that’s a harder sell if you simply rent out your GPUs and don’t use them for that innovation.

Sean: But why be positive when you can be cynical? Which is to say, this seems like a major heat check before the IPO that we’ll soon see rammed into the markets with SpaceX.

Anthony, not only did you say that Grok is not used for large enterprise tasks, there are also reports that xAI employees were use of other modelsthey didn’t even use it [Grok] internally, and that caused this big commotion within xAI, post acquisition of SpaceXinvolving essentially all the co-founders except Elon Musk, [and] he’s basically saying he’s starting from scratch with xAI, despite the fact that SpaceX paid $250 billion for it in the lead-up to this mega-IPO.

And now he says that they are going to disband xAI as a separate entity within SpaceX. He starts calling the whole thing SpaceXAI, because this guy loves nothing more than ruining a brand that has any value – see Twitter.

This may be a more credible company in the short term, and so at some level I could see this perhaps being more attractive to investors when it comes time for the IPO because it’s a little bit more reliable and certainly more real than if they’re a frontier lab developer. But it’s also not the kind of company that – at least in a normal environment – will attract the same external investment that we see going into all the border labs.

That may be one of the biggest points of tension we’ve seen arise during this IPO process.