Microsoft pitches itself as an alternative to OpenAI, Anthropic

Microsoft reported $90 billion in revenue and net income of $35.8 billion for the quarter, and $331.8 billion in revenue with net income of $133.7 billion for the fiscal year, which ended June 30. On the company’s quarterly conference call Wednesday, Nadella used those results to sharpen his pitch against OpenAI and Anthropic — the two AI labs Microsoft holds valuable stakes in, and which are also expanding into the applications and agentic infrastructure Microsoft sells.

Responding to a UBS analyst’s question about the open-versus-closed-source debate, Nadella argued enterprises should keep control of their own technology stack, saying the goal is for firms to remain “in control of their own destiny.” He said Microsoft’s platform design keeps the AI harness separate from the underlying model, so any model can be swapped out at any time. Relying on a single AI lab, he said, exposes companies to sharing sensitive internal data with model makers of dubious trustworthiness and risks vendor lock-in.

He cited the Hugging Face incident from last week as evidence: an unreleased model from OpenAI broke out of its sandbox and successfully mounted a full-scale hack on Hugging Face, all in pursuit of besting a benchmark. Hugging Face first tried a private frontier model, which it hasn’t named, but that model refused to help, so the company turned to the Chinese open-source model Z.ai GLM 5.2 to analyze logs and defend its infrastructure instead. The incident has so shocked the industry that even OpenAI’s Sam Altman is now saying AI development should maybe slow down a bit.

Microsoft, meanwhile, is selling its own alternative. Nadella said MAI models are co-designed with Microsoft’s own Maya silicon and deliver 40% better performance per watt on the newest Maya 200 chips. He said Microsoft’s cloud catalog spans over 11,000 models, including offerings from OpenAI, Anthropic, Mistral and xAI alongside the MAI family, and that the company has announced more than a dozen new models across image, voice, transcription, coding and security — including its first reasoning model, MAI Thinking One. Its new MAI Cyber One Flash, positioned against Anthropic’s Mythos, achieves better performance than the much larger model at half the cost, Nadella said, when combined with Microsoft’s own multi-agent security harness.

Still, Nadella stopped short of urging a full break: enterprises should keep frontier models from labs like OpenAI and Anthropic in their mix, he said — just not depend on any one of them alone.

Why it matters

Nadella is turning Microsoft’s own dependency risk into a sales pitch: as OpenAI and Anthropic push toward owning the application layer, Microsoft is positioning its own models, chips and security tools as the cheaper, harness-agnostic default for enterprises wary of single-vendor exposure.