Nadella framed Microsoft’s approach as a mix of frontier and proprietary models, warning that exclusive reliance on outside labs could expose data and limit control.
Microsoft holds a special position in the AI era: as one of the most powerful providers of cloud services and software as a service, the company also owns substantial stakes in the two largest AI research labs – OpenAI and Anthropic.
These incentives are starting to clash amid the release of impressive financial results: the quarter delivered strong profits – revenue of $90 billion, net income of $35.8 billion. For the fiscal year ended June 30, Microsoft reported revenue of $331.8 billion and net income of $133.7 billion.
The head of the company, Satya Nadella, does not allow the trajectory of Anthropic and OpenAI – which are expanding into applications and infrastructure with agents that could in the future have direct interactions with customers – to derail Microsoft from maintaining its current financial tempo and cash-flow momentum.
Nadella urges enterprises to deploy multiple models and not rely exclusively on frontier AI labs to manage the agent component and other critical functions. He emphasizes that such an approach is dangerous because it may require companies to disclose internal data to model providers with questionable trust. He knows his customers well: among the enterprise IT community, fears of data leaks and vendor lock-in prevail.
The goal is for the company to control its own destiny and be able to guide its future, keeping control over the platform’s architecture.
– Satya Nadella
Microsoft’s Strategy in the Context of Competition
Nadella noted that the company offers a broad portfolio of AI agents under the Copilot brand, including the coding agent GitHub Copilot. It is the agent-based component that is emerging as one of the main investment directions in the integrated AI ecosystem.
Microsoft is also continuing to develop its own MAI model line and Maya chips, positioning them as economical and secure solutions for customers who support their ecosystem.
“Each client wants to pick the optimal model for a task based on quality, latency, cost, and compliance. We have the widest catalog of models in the cloud – over 11,000 options, including leaders OpenAI, Anthropic, Mistral, xAI, and our own MAI line,” he said.
40% better performance per watt-hour when using MAI models on Maya 200
– Satya Nadella
In the context of competitive dynamics, Nadella mentioned the recently announced MAI Cyber One Flash as one of the solutions that could compete with Mythos thanks to more efficient integration with a multifaceted security system and reduced costs.
Describing the overall strategy, Nadella believes that enterprises should use a combination of frontier models from OpenAI and Anthropic with Microsoft’s own solutions, but should not fully rely on them for critical data and risk management aspects.
Following the remarks, the executive emphasized Microsoft’s aspiration to offer clients the broadest possible access to different models and to integrate them with its own solutions, which helps reduce costs and increase data security in corporate environments.