Comment by paxys
2 days ago
TL;DR - financial analysts look at current charts and project them foward by 5 years and go "wow the numbers look bad".
Sure OpenAI may well be bleeding money into the 2030s, or may even go bust completely depending on how pessimistic you are, but the analysis completely skips:
- They are building their own data centers, and will be less reliant on renting compute from Microsoft and Amazon over time.
- Once the AI bubble has subsided costs for GPU purchases and rentals will decrease significantly. Plus there will be more advancements and competition in the space (e.g. Google TPUs) and Nvidia will no longer be able to name their own price.
- We will write more efficient software for training and inference.
- Once user growth is tapped out OpenAI will no longer need to have the overly generous free tier that they do today. And if they decide to turn up the advertising faucet these users could bring in a ton more revenue than in the projection. Thinking that every AI company combined will capture only 2% of the total digital advertising market is ridiculous. AI apps are already challenging social media for scrolling time.
Basically, the entire space is evolving so rapidly that it's pointless to make a projection with the assumption that the landscape isn't going to change from here on.
I think you are right that the entire analysis is flawed. The Amazon and Microsoft "rental" deals have inflated price tags because of the circular financial arrangements between them and OpenAI, and because those future revenue streams can be used notionally to finance CapEx. All of the Stargate DC build is being done through for-profit SPVs, so the financials are murky, but building the infra gives them collateral for debt, and they are going to lease the compute to the highest bidder, so there is a whole scheme for getting out of the non-profit box, creating a self-perpetuating loop of borrowing to build, using what they build as collateral for more borrowing, raising additional revenue and hedging by leasing compute to 3rd parties, and then using the for-profit SPVs to cross-subsidize OpenAI proper. That plan has enormous risks of its own (can the leadership team of OpenAI effectively build a competitor in the hyperscale compute space?) but whatever happens, it won't just be straight line scaling their current deals with existing hyperscalers.
[dead]