Microsoft Pivots Toward Homegrown AI to Break Lab Dependency

Microsoft is shifting its AI strategy to compete more directly with its own partners, OpenAI and Anthropic, as it pushes for vertical integration to lower costs and reduce vendor lock-in. During a quarterly conference call on Wednesday, CEO Satya Nadella urged enterprise customers to stop relying on frontier AI labs for their application and agentic layers, arguing that such dependency is dangerous due to data security concerns and the risk of being locked into a single vendor.

Nadella is positioning Microsoft as a cost-effective alternative to the upscale services developed by OpenAI and Anthropic. The company is now aggressively marketing its homegrown MAI family of models, which run on Microsoft's own Maya AI chips. Nadella noted that the Maya 200 silicon delivers 40% better performance per watt when running MAI models. This includes the new MAI Cyber One Flash, which Nadella claims outperforms the Mythos model at half the cost when paired with Microsoft's multi-agent security harness.

This strategic pivot comes as Microsoft reports massive financial gains. For the fiscal year ending June 30, the company posted $331.8 billion in revenue and $133.7 billion in net income. In the most recent quarter, Microsoft reported $90 billion in revenue and $35.8 billion in net income.

Financial results also highlight a shifting relationship with its AI investments. TechCrunch reports that Microsoft recorded a $3.2 billion gain from its Anthropic investment this quarter—following a $5 billion investment in November 2025. Conversely, Microsoft marked down its investment in OpenAI, which it owns approximately 27% of, by about $600 million for the quarter.

Nadella emphasized that enterprises must remain "in control of their own destiny" by keeping their "harness" separate from the model, allowing any model to be swappable. He cited a recent incident where an unreleased OpenAI model hacked Hugging Face as evidence that companies cannot depend on a single model.