SpaceX 10GW in 2027 – Why It's Real, Will Drive $300B ARR for SpaceX
newsletter.semianalysis.com> both OpenAI and Anthropic can generate over $100B/GW/year of revenue when selling API inference on a GB300 cluster. This is significantly more than the costs of renting a GB300 cluster for a year at current neocloud prices.
This is one view and then there are Ed Zitron's articles on impending collapse of AI. I'm not sure whom to believe.
The specific predictions here are (1) SpaceXAI will borrow ~$250B to build at least 5 GW of GPU capacity in 2027 and then (2) OpenAI/Anthropic/others will pay SpaceXAI ~$300B/year to rent those GPUs. This money would have to come either from customer revenue (assuming something like $1T of token revenue in 2027) or investors. These are very large amounts of money. If you extrapolate current exponential growth trends, $1T is possible. But any number of events could pop the bubble in the next 18 months. The biggest risk AFAIK is simply investors running out of capital to invest. Another risk is that token prices collapse due to competition.
Mineral exploration / production VCs are salivating (well, experimenting with memes):
https://xcancel.com/robert_ivanhoe/status/208546990089877138...