Oracle's $300 Billion Revenue OpenAI Deal. Will It Actually Happen?
What is the probability that Oracle’s $300 billion deal with OpenAI actually happens? Low in my view.
The deal would see OpenAI spend approximately $60 billion per year in Cloud Infrastructure (e.g., data centers) with Oracle (ORCL) over 5 years beginning in 2027.
OpenAI is tracking toward a $13 billion revenue year for 2025 - a far cry from $60 billion of Revenue - never mind cash!
GPT 5 was an incremental improvement over GPT4.5. Anthropic’s Claude is better at coding. Gemini 2.5 Pro is better at coding in my view, although admittedly I prefer to interact with GPT 5 over Gemini 2.5 for non-coding interactions given that Gemini’s personality is quirky. All this to say I don’t believe that GPT 5 is the best LLM.
Where will OpenAI find the cash to meet its $60 billion annual financial obligation with Oracle?
Will OpenAI continue to raise capital at nosebleed valuations in large amounts?
The $100 billion build LLM is not happening in 2026 or 2027 (or 2028 or 2029). Perhaps we will get an $8-10 billion LLM build in 2026, which will probably burn 2x that amount in cash.
The $300 billion Oracle deal is due to begin in 2027. Where will OpenAI find $60 billion in cash per year over 5 years with which to pay Oracle for the datacenters that Oracle is to construct? Does OpenAI expect to IPO at an outrageous valuation that will enable it to raise $500-600 billion? The numbers don’t work.
My guess is that this deal will never be fulfilled and that the contract will be adjusted significantly downward or cancelled outright over the next couple of years.