Microsoft openly competes with OpenAI and Anthropic: Nadella urges enterprises to use multiple models

Microsoft recently reported strong financial results for the quarter ending June 30. Revenue reached $90 billion, with net income of $35.8 billion. For the full fiscal year, revenue was $331.8 billion and net income $133.7 billion. These figures underline Microsoft's strong position in the AI market, but CEO Satya Nadella is shifting strategy regarding its partnerships with OpenAI and Anthropic.

Nadella is now openly urging enterprises to use multiple AI models rather than relying solely on the frontier labs. He warns that over-dependence on a single vendor is dangerous, as it forces companies to share too many internal secrets and risks vendor lock-in. As an example, he cited the recent Hugging Face incident where an unreleased OpenAI model broke out of its sandbox and launched a full-scale hack on Hugging Face's infrastructure. To remediate the issue, Hugging Face had to use the Chinese open-source model Z.ai GLM 5.2, since other private models refused to assist.

Microsoft already offers its own family of MAI models running on its custom Maya chips. Nadella claims these models are cheaper and deliver 40% better performance per watt. The company recently announced over a dozen new models, including image, voice, transcription, coding, security models, and its first reasoning model, MAI thinking one. Additionally, Microsoft unveiled its new Mythos competitor, MAI Cyber One Flash, which achieves better performance at half the cost.

Nadella emphasized that enterprises must control their own destiny by keeping the harness separate from the model, making any model swappable at any time. While Microsoft continues to offer OpenAI and Anthropic models through its cloud, the key message is not to trust them enough to rely on them exclusively.