Amazon just finished writing OpenAI a $50B check. The conditions weren't met. They paid anyway.

Amazon completed its full $50 billion investment in OpenAI on Friday.

The final $35 billion was supposed to be conditional. IPO. AGI breakthrough. Neither happened.

Amazon wired the money anyway.

Here's what $50 billion actually buys: AWS becomes the exclusive third-party cloud for OpenAI Frontier, the enterprise platform for AI agent teams. OpenAI commits to 2 gigawatts of Amazon's Trainium chips. And the infrastructure deal expands to potentially $100 billion over eight years.

This isn't a bet on OpenAI. It's a bet against the idea that one AI model provider wins.

Amazon already has Nova. It already has Anthropic. Now it has OpenAI. Three competing model families, all running on AWS infrastructure. The enterprise customer gets choice. Amazon gets the compute bill for all of them.

The real signal: conditions weren't met and Amazon invested anyway. That tells you Amazon sees the enterprise AI market as large enough to justify $50 billion in sunk cost before knowing if the bet pays off. This isn't venture logic. This is infrastructure monopoly logic.

If your AI strategy assumes model costs will drop because competition is increasing, recalculate. The companies building the models are now the companies owning the chips. The cost floor just moved.