Inside the Trillion-Dollar AI Buildout | Dylan Patel Interview
- Source
- youtube.com
- Author
- Invest Like the Best
- Date
Context
In an Invest Like the Best interview, Dylan Patel explains how to read the large compute deals around OpenAI. Patel's central point is that compute has to be bought before the revenue it supports exists: a cluster is needed to train a better model, and again to serve it. Rivals that are far larger companies can fund this from cash flow, which Patel says creates a risk of OpenAI being too small to matter without allies.
Patel puts rough numbers on it as an estimate: renting about one gigawatt of data center capacity costs 10 to 15 billion dollars a year, and OpenAI signs five-year commitments, so each gigawatt is a 50 to 75 billion dollar obligation. In Patel's telling, that is why OpenAI is striking deals with Oracle and Nvidia, and why Oracle is making a large bet that the commitment will be paid. It is one analyst's interpretation, not a disclosed contract figure.
Checking sign-in…
Loading comments…






