I believe that OpenAI (I'll get to SpaceX in a second) has a huge valuation risk because:

1. It's a bet that OpenAI will "win" AI and have a significant moat; and

2. Future hardware improvements won't massively devalue OpenAI.

I believe open source models will win here, mainly because China won't allow otherwise. I also think that nobody is really talking about the hardware decpreciations coming in the next few years, which is going to be really important from a performance-per-Watt perspective. B100s aren't going to suck. But a theoretical T100 will get 30-80% more performance for the same energy input.

So, SpaceX. I've previously said that SpaceX would've been a significantly better company without xAI. SpaceX was used to rescue Elon and other "investros" from the financially disastrous Twitter purchase. Starlink, Starship (which is a risky program) and the Falcon 9 are a solid business. They're just not a $2 trillion business.

So I believe that the AI bubble contributes at least half of SpaceX's valuation and when and if that bubble bursts, at least half of SpaceX's value is at risk.

Google announced they're throwing billions to rent GPUs from SpaceX. That might sound good. It solves a short-term cash issue. But as another commenter put it, it makes SpaceX seem more like a Commercial REIT. After all, renting out your GPUs is literally the lowest-value thing you can do with them. You're not building a business. You're taking rent so someone else can build a business.

So buying Cursor and I'm sure any number of other AI startups in the coming year or two, seems aimed at kicking that AI can down the street.

So I view the Google-SpaceX as a red flag in the short-to-medium term. SpaceX simply can't seem to do anything valuable with all the compute they have. And I also have way more confidence in Anthropic (in particular), OpenAI and Gemini than I do in Grok.