I would like to hear your take on these three points:
1. AI is a low margin business like energy/manufacturing
The idea is that, unlike software, AI has high marginal costs per query. Serving 100M queries costs roughly 2x as much as 50M. So as models get more complex, computing costs (electricity and water)scale up linearly. Doesn't this trap AI companies in a CAPEX-heavy, OPEX-intensive, and low-margin game instead of the high-margin SaaS story everyone's betting on?
2. Anthropic has a better business model than OpenAI
OpenAI relies heavily on consumer subscriptions (ChatGPT Plus), which are volatile. Anthropic gets 80+% revenue from enterprise/API deals much stickier. So Anthropic's actually in a stronger position long term?
3. $1 trillion OpenAI IPO doesn't make sense
Above reasons plus they're burning $14B+ annually by 2027, mostly going to Microsoft for cloud credits. Plus, the circular logic of their investors funding startups that buy OpenAI credits. Sounds like WeWork all over again?
Best,
Matt
On AI being low margin, some businesses indeed will see lower margins while others will be higher margin. The compute level is indeed likely to be lower margin, but if there are just a few big winners here, getting a lot of absolute dollars at a low margin can still be a good business. Then on top of this, there is likely to be other companies that get higher margins that are more focused on the software side. We are not sure if this is the best analogy but we might compare it to a telco that is lower margin but providing the base 'internet' services, and then 'the Googles' that build the high margin businesses on top of it.
On #2, a lot remains to be seen here and a lot can change. There will be different models and a bit of diversity in strtegy and approaches is probably a good thing overall.
As for any OpenAI IPO, they probably have a bit more work to do before they go this route but we think it is not at the same 'level' as WeWork in terms of promotional and aspirational claims. OpenAI has certainly made some comunication missteps in the last month or two but the business is far more scalable and is actually a first mover, where WeWork was really just a different concept of an existing business model.