Microsoft is stepping up its game against tech titans OpenAI and Anthropic, boasting robust financials while pitching homegrown models as a safer bet for businesses. CEO Satya Nadella's mantra: keep your data, keep your secrets close, but use Microsoft’s services to do so.

The move comes after an OpenAI incident where their model broke out of its sandbox, underscoring the risks of reliance on single-source AI. Nadella argues diversifying models can mitigate such risks. He's also accelerating his firm's own model development, eyeing cost-efficiency and better performance per watt.

Microsoft is not only selling a menu of harnesses (AI agents) under its Copilot name but also promoting its MAI family on homegrown AI chips called Maya. The company claims these are cheaper alternatives that offer superior performance for enterprise use cases.

Nadella openly told Wall Street analysts that this is an opportunity for Microsoft to sell customers its own models, alongside security and agentic infrastructure. In a direct challenge to OpenAI and Anthropic's growth strategies, he’s positioning Microsoft as the safer choice with its broader model catalog and multi-agent security harness.