Infrastructure deals are the new talent wars
While everyone obsesses over model benchmarks, the real AI competition is happening in billion-dollar infrastructure deals.
We’re watching the wrong leaderboard. Everyone’s tracking MMLU scores and benchmark wars while the actual competition moved to infrastructure partnerships months ago. The companies writing the biggest checks for compute and data centers are going to win, not the ones with the flashiest demos.
Follow the money, not the models
Meta, Oracle, Microsoft, Google, and OpenAI are each dropping billions on infrastructure deals. This isn’t about having the smartest algorithms anymore. It’s about securing enough compute to train at scale, enough power to run inference, and enough data centre capacity to serve millions of users without falling over. The model quality gap is narrowing fast, but infrastructure advantages compound over years.
The new platform lock-in
These infrastructure partnerships aren’t just about buying servers. They’re creating deep technical dependencies that will shape the industry for the next decade. When you’re locked into specific hardware architectures, cloud providers, or power grids, switching costs become astronomical. The AI companies making these deals now are building moats that pure software plays can’t cross.
The talent wars taught us that the best engineers follow the best resources. Now the best resources follow the best infrastructure deals.