Microsoft finds itself in an unusual position as artificial intelligence reshapes the technology landscape. The company is simultaneously one of the largest cloud providers and software-as-a-service firms globally, while also holding significant stakes in the two leading AI research organizations, OpenAI and Anthropic. These competing interests are beginning to collide as Microsoft reports staggering financial results. The company recently posted an exceptionally profitable quarter, generating $90 billion in revenue and $35.8 billion in net income. For the fiscal year ending June 30, Microsoft achieved $331.8 billion in revenue with a net income of $133.7 billion. CEO Satya Nadella is determined not to let the growth trajectories of Anthropic and OpenAI—both of which are expanding into applications and agentic infrastructure that could ultimately allow them to own customer relationships—disrupt such cash flow.

Nadella has been advising enterprises to adopt multiple models and avoid relying solely on frontier AI labs for the agentic harness or application layer. He warns that such dependence is risky because it forces companies to share too many internal secrets with model makers of questionable trustworthiness. He understands his customer base well. Enterprise IT departments fear both data breaches and vendor lock-in. During Microsoft’s quarterly conference call on Wednesday, Nadella openly told Wall Street analysts that this presents an opportunity for Microsoft to sell customers its own in-house models, along with agents, AI security solutions, and more, all while promising lower costs. In essence, he is positioning Microsoft as an alternative to many of the premium services that OpenAI and Anthropic are developing for their own growth.

When UBS analyst Karl Keirstead specifically asked Nadella to weigh in on the open versus closed-source debate currently roiling the AI industry, and how Microsoft would benefit, the CEO responded forcefully. "The goal is to have the firm be in control of their own destiny," Nadella said of enterprises. "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." Microsoft, of course, also sells a range of harnesses—also known as AI agents—under the Copilot brand, including its coding agent GitHub Copilot. Coding agents are where much of the AI spending is currently concentrated. Nadella pointed to a high-profile incident from the previous week as evidence supporting his warnings. "If you look even at the Hugging Face incident, the biggest thing that we should take away from that is you can’t sort of depend on any one model," he said. "You will maybe need multiple models to even remediate some challenges that get caused by one model. Like that’s the way to think about it, right. Which is you can’t be subject to a refusal of one model."

The incident involved an unreleased model from OpenAI that broke out of its sandbox and successfully mounted a full-scale hack on Hugging Face, all in pursuit of outperforming a benchmark. While trying to understand what happened, Hugging Face initially used a private frontier model (which it hasn’t named) that refused to assist. So it turned to the Chinese open-source model Z.ai GLM 5.2 to analyze logs and defend its infrastructure. The event has so shocked the industry that even Sam Altman is now suggesting that AI development might need to slow down. Nadella also made clear that Microsoft is happily selling its own homegrown models, the MAI family, on its own homegrown AI chips, Maya, and pitching them as cheaper alternatives. "Every customer wants the right model for each task based on quality, latency, cost, and compliance. We offer the broadest model catalog in the cloud with over 11,000 models, including the leads from OpenAI, Anthropic, Mistral, xAI, as well as our own MAI family," he said. He added: "We’re also accelerating our own model development. We announced more than a dozen new models across image, voice, transcription, coding, security, including our first reasoning model, MAI thinking one, all with cost-efficient inference at the core for the enterprise use cases. We are co-designing these models with our silicon, and we are seeing 40% better performance per watt when running MAI models on Maya 200."

As for Mythos, Nadella pointed to Microsoft’s new competitor announced earlier this week, MAI Cyber One Flash. It "achieves better performance than the much larger Mythos model, but at half the cost when combined with our multi-agent security harness," he said. Sure, the Microsoft CEO acknowledges that enterprises should incorporate the frontier models from OpenAI and Anthropic into their mix. But his overarching message is clear: don’t trust them enough to become overly reliant on them.