Microsoft pitches its own models against OpenAI and Anthropic
Nadella told analysts Microsoft will sell enterprises its own MAI models on its Maya chips — including first reasoning model MAI Thinking One and MAI-Cyber-1-Flash, which he says beats the larger Mythos at half the cost — while keeping stakes in OpenAI and Anthropic and 11,000 models on Azure. His pitch is architectural: keep the harness separate from the model so any model is swappable, which reframes the frontier labs as interchangeable suppliers rather than platform partners. Same quarter: $90B in revenue, $35.8B net income, and a $3.2B gain on the Anthropic stake — partner and competitor economics now sit in one income statement.
Source: techcrunch.com ↗
The goal is to have the firm be in control of their own destiny. We are very, very clear about the architectural sort of design of the platform, which is you got to keep your harness separate from the model … that means any model at any given time is swappable.
Why this matters
- → Microsoft now competes directly with its portfolio companies in frontier AI.
- → Architectural shift (swappable models) weakens OpenAI/Anthropic's platform lock.
- → Enterprise data security concerns give Microsoft leverage against its own partners.