Meta spends $100M+ on Azure while building its own rival
meta spends hundreds of millions on microsoft azure to access openai models for benchmarking, signaling that building proprietary inference infrastructure is costly and complex.
meta has become one of microsoft azure's biggest ai clients, spending hundreds of millions of dollars a year to access openai models via the foundry marketplace and utilizing trillions of tokens weekly. the spending primarily serves to benchmark meta's proprietary models against openai's offerings, even as meta develops its own api service that could directly compete with foundry. while microsoft derives roughly 70% of its ai revenue from openai's compute needs, the remaining customer base includes tech giants like bytedance and adobe. this dynamic underscores a period of interdependence in the ai sector where leading labs trade revenue for the compute access required to advance model development.