I'm still waiting for the real news to drop- in the next 6 months we're going to start hearing some big moves from Space X AI. Early this year they lost pretty much all their leadership, it's very clear they failed to keep up with the frontier models and Musk has essentially given up for now - renting out their compute to Anthropic and Google. But that's not sustainable, everything they say about their IPO is that AI is the core value driver. So at some point Musk is going to have some decisions to make about who he brings in to drive that. I imagine once they get that person in and start building a team around them the deals with Anthropic and Google will be ended.
I guess the cursor guys will be happy because they got their pay day, but I'd be very aware if I were them that their future is at the whim of whoever Musk appoints and it's difficult to tell who that would be right now.
I guess now is the time to take bets, so I'm going to bet an early OpenAI employee like Sutskever gets the job and they acquihire him in. Here's a bit of a laugh - at this stock price Musk could probably tempt Demis to come over, that would be wild.
Why is it not sustainable? Right now it appears to be a better business than selling access to models.
It's a good business. Maybe even a great business. But it's not going to justify a valuation like Space X. In the same way that Tesla has slowly become less competitive in automotive, I don't think it's sensible to assume Space X will have some durable edge in building data centres, especially when basically everything going into those data centres are commodities. And if it is just a neo cloud, then you have to contend with the pro-cyclical nature of that. It's also just clearly not their plan, they're not promising to be a neo cloud. They're promising to own the full stack.