I just noticed something very interesting in this monumental funding round of OpenAI that closed last week. Everyone is focused on the number: 110 billion dollars, a valuation of 730 billion. But there's a technical detail that goes unnoticed and, honestly, changes everything about how we should think about the future of AI.

Amazon and Microsoft took very different things from this deal. Microsoft secured exclusivity in hosting OpenAI's stateless APIs — basically, every API call goes through Azure. It's a predictable cash flow, but with a problem: as models become more similar and price competition increases, this margin will only shrink. That's the present, you know? Quick integration, low friction, but commoditized.

And Amazon? It bet on the future. It secured rights to the Stateful Runtime Environment — that's the fancy name for an AI agent that actually works continuously, with memory, capable of performing real tasks, not just answering specific questions. Amazon expanded the deal from 38 billion to 100 billion over 8 years. This is infrastructure for when stateful agents become the core engine of business productivity.

The cool part is that this is exactly the bet Aakash Gupta (, the guy who has been following AI for years ), highlighted: in 2026 and 2027, companies will be buying systems that run sustainably, collaborate across tools, and maintain context. It’s no longer about isolated API calls. It’s about a digital workforce that functions like a real employee.

And look at the political game behind this. OpenAI was heavily tied to Microsoft — 27% of shares, infrastructure controlled, negotiation power completely unequal. Now, with Amazon stepping in strongly, a direct competition is created between the two cloud giants. For OpenAI, it’s the classic move: not to be tied to a single provider, using the future as a bargaining chip to negotiate better.

In the end, neither Microsoft nor Amazon can give up on OpenAI at this moment. When both are at the table, power shifts back into the hands of whoever controls the model. It’s a quiet but well-calculated repositioning. Stateless vs. stateful is more than a technical distinction — it’s the map of power in AI for the coming years.