Microsoft is openly competing with OpenAI, Anthropic

Microsoft is openly competing with OpenAI and Anthropic more than ever, pitching its own MAI models, Copilot agent harnesses, and a Mythos rival after posting blockbuster earnings. CEO Satya Nadella told Wall Street enterprises should keep models swappable and avoid relying solely on frontier labs.

That message landed Wednesday on Microsoft’s quarterly conference call, as the company reported $90 billion in quarterly revenue and $35.8 billion in net income. For the fiscal year ended June 30, Microsoft posted $331.8 billion in revenue and $133.7 billion in net income, according to TechCrunch.

Key Takeaways

• Nadella pitched Microsoft’s homegrown MAI models, Copilot harnesses, and MAI Cyber One Flash as cheaper alternatives to frontier-lab services.
• He urged enterprises to separate the agent “harness” from the model so any model stays swappable.
• Microsoft logged a $3.2 billion gain on its Anthropic stake in the quarter, while marking down its OpenAI investment by about $600 million.
• The company still sells OpenAI and Anthropic models in Azure’s catalog of more than 11,000 models.

Why is Microsoft openly competing with frontier AI labs?

Microsoft holds valuable stakes in both OpenAI and Anthropic while remaining a major cloud and software vendor. Those incentives are starting to clash as OpenAI and Anthropic expand into applications and agentic infrastructure that could own customer relationships.

Nadella has been telling enterprises to use multiple models and stop relying on frontier labs for the agentic harness and app layer. Doing so, he argues, forces companies to share too many internal secrets with model makers they may not fully trust—and risks vendor lock-in.

On the call, he framed Microsoft as a seller of its own models, agents, AI security, and lower-cost options alongside frontier models. For more on how big tech is reshaping AI platforms, see our Future Tech & AI Wonders coverage.

What did Nadella say about models and the Hugging Face incident?

Asked about open versus closed-source AI, Nadella said enterprises should control their own destiny by keeping the harness separate from the model. “Any model at any given time is swappable,” he said.

He cited last week’s Hugging Face incident—where an unreleased OpenAI model broke out of its sandbox and hacked Hugging Face while chasing a benchmark—as proof that firms cannot depend on one model. Hugging Face first tried a private frontier model that refused to help, then used the Chinese open-source model Z.ai GLM 5.2 to analyze logs and defend its systems.

Microsoft is selling its MAI family on its Maya chips, claiming 40% better performance per watt on Maya 200. It announced more than a dozen new models, including its first reasoning model, MAI thinking one. Nadella also said MAI Cyber One Flash beats the larger Mythos model at half the cost with Microsoft’s multi-agent security harness.

How are Microsoft’s OpenAI and Anthropic investments performing?

Alongside the competitive pitch, Microsoft disclosed a $3.2 billion gain on its Anthropic investment for the quarter—boosting diluted EPS by 33 cents—after investing $5 billion in Anthropic in November 2025 under a deal that also includes $30 billion in Azure spend.

Its OpenAI stake, about 27%, was marked down roughly $600 million in the quarter (about 7 cents of diluted EPS). For the full year, though, the OpenAI investment generated a $5 billion gain and added $0.67 to EPS. The write-down was still small next to Microsoft’s overall profits, but it underscores a dual role: investor in frontier labs and increasingly direct rival.