Microsoft is in a unique position as AI overtakes the tech industry. It is one of the world’s largest cloud providers and software-as-a-service companies, while also playing a key role in two major AI labs, OpenAI and Anthropic.
Those incentives are starting to clash as Microsoft posts blockbuster financial results. The company recently reported a very profitable quarter with $90 billion in revenue and $35.8 billion in revenue. For the fiscal year, which ended June 30, Microsoft reported $331.8 billion in revenue on annual revenue of $133.7 billion.
And CEO Satya Nadella isn’t about to let Anthropic and OpenAI’s track record — growing into apps and infrastructure for agents that could eventually allow them to have relationships with customers — get in the way of that kind of money.
Nadella has been preaching to businesses to use multiple models and stop relying on frontier AI labs for the agency/app layer.
Doing this is dangerous, he has been saying, because it requires companies to share many of their internal secrets with model builders of questionable credibility. He knows his customers. Enterprise IT fears both data leakage and vendor lock-in.
Now he publicly told Wall Street analysts during the company’s quarterly conference call on Wednesday that this is an opportunity for Microsoft to sell its home models to customers, as well as agents, AI security and more, while promising lower costs.
In other words, he’s positioning Microsoft as an alternative to the many high-quality resources that OpenAI and Anthropic are developing to scale.
When UBS analyst Karl Keirstead directly asked Nadella to analyze the open versus closed source debate driving the AI industry, and how Microsoft would benefit from it, Nadella came out swinging.
“The goal is for the company to control its future,” said the CEO of businesses. “We’re very clear about the type of platform structure, which you have to keep your harness separate from the model … which means any model at any time is interchangeable.”
Microsoft, of course, sells a menu of harnesses (AI agents), too, under the Copilot name, including its GitHub Copilot coding agent. Coding agents are where most of the AI dollars are spent today.
And he used last week’s high-profile incident as evidence for his warnings.
“If you look at the Hugging Face incident, the biggest takeaway is that you can’t rely on any one model,” Nadella said. “You’ll probably need multiple models to address some of the challenges posed by one model. That’s the way to think about it, right? That is, you can’t reject the rejection of one model.”
The incident involves an unreleased model from OpenAI that comes out of its sandbox and successfully installs a full hack on Hugging Face, all in pursuit of improving the benchmark. Trying to understand what happened, Hugging Face initially tried to use a secret boundary model (which he had not yet invented) which refused to help him. So it turned to the Chinese open source model Z.ai GLM 5.2 to analyze the logs and secure its infrastructure. This incident has shocked the industry so much that even Sam Altman says that maybe AI development should slow down.
Nadella also clarified that Microsoft is happily selling its home models, the MAI family, for its home AI chips, Maya, and positioning them as cheaper alternatives.
“Every customer is looking for the right model for each task based on quality, latency, cost, and compatibility. We offer the most extensive model catalog in the cloud with over 11,000 models, including leading OpenAI, Anthropic, Mistral, xAI, and our MAI family,” he said.
He added: “We’re also accelerating our model development. We’ve announced more than a dozen new models across graphics, voice, transcription, coding, security, including our first logic model, a single MAI logic, all with key cost-effective definitions for business use cases. We’re co-designing these models with our silicon, and we’re seeing 40% better performance when using Maya 20 better than each MAI model.”
About Legends? Nadella pointed to Microsoft’s new Mythos competitor announced earlier this week, the MAI Cyber One Flash. “It achieves better performance than the larger Mythos model, but at half the cost when combined with our multi-agent security harness,” he said.
Indeed, Microsoft’s CEO says businesses should use the frontier models offered by OpenAI and Anthropic in their mix. But his main message is: don’t trust them enough to rely on them.
If you shop through links in our articles, we may earn a small commission. This does not affect our editorial independence.
