1. Pick the game 2. Pick the players 3. Run the scenario AI is an ongoing negotiation over two things: today's profits and tomorrow's moats. Explore now: ai-gametheory.com/.

Select any two players, and view the relationships between them. What each side is optimizing, what each "game" means for the pair, what are the built-in tension and dynamics that persist through games.

We map eight players in the stack: Chips. Clouds. Frontier labs. Open-weight labs. Inference. Harnesses. Applications. Enterprises. We place every pair inside six games: Intelligence.
Distribution.
Execution.
Inference.
Data.
Users. Because the model is non-commutative, there are 9 players, 8 counterparts, and 6 games, which leads to 432 possible scenarios!!

Hope this project surfaces key questions: - Why do clouds back multiple model labs? - Why build their own chips? - Why do chip firms support open models? - When does cheap capital become a distribution edge? - When is “irrational” spend rational given real incentives? And more..