As AI overtakes the technology industry, Microsoft is in a unique position. The company is one of the world’s largest cloud providers and software-as-a-service companies, as well as holding valuable stakes in two major AI labs, OpenAI and Anthropic.
Those incentives are starting to collide as Microsoft releases blockbuster financial results. The company just reported a very profitable quarter with $90 billion in revenue and $35.8 billion in net income. For the fiscal year that ended June 30, Microsoft reported revenue of $331.8 billion and net income of $133.7 billion.
And CEO Satya Nadella has no intention of letting Anthropic and OpenAI’s trajectory (which ultimately expands into the applications and agent infrastructure that enables its own customer relationships) derail those funds.
Nadella has been advocating for companies to use multiple models and stop relying on Frontier AI Labs for the agent harness/app layer.
Doing so is dangerous, he said, because it would require companies to share too many internal secrets with model makers of questionable credibility. He knows his customers very well. Corporate IT departments are concerned about both data breaches and vendor lock-in.
He openly told Wall Street analysts on the company’s quarterly conference call Wednesday that this is an opportunity for Microsoft to sell its homegrown models to customers alongside things like agents and AI security, while promising cost savings.
In other words, he’s pitching Microsoft as an alternative to many of the premium services that OpenAI and Anthropic are developing to grow their companies.
When UBS analyst Karl Keilstedt specifically asked Nadella to consider the open source vs. closed source debate roiling the AI industry and how Microsoft stands to benefit from it, Nadella seemed perturbed.
“The goal is to empower companies to control their own destiny,” the CEO said of companies. “We’re very specific about the architectural kind of design of the platform, which means we have to keep the harness separate from the model, which means it can be swapped out on any model at any time.”
Of course, Microsoft also sells a menu of harnesses (aka AI agents) under the Copilot name, including the coding agent GitHub Copilot. Currently, much of the AI spend is on coding agents.
And he used last week’s high-profile incident as evidence for his own warning.
“If you look at the face-hugging incident, the biggest thing we have to learn from it is that we shouldn’t rely on one model,” Nadella said. “You might need multiple models even to fix some challenges posed by one model. That’s the way you think about it, right? It’s not subject to rejection of one model.”
The incident involved OpenAI’s unreleased model breaking out of the sandbox and a successful full-scale hack into Hugging Face, all with the goal of exceeding benchmarks. In an attempt to understand what happened, Hugging Face initially tried using the Private Frontier model (which it has not named), but it refused to help. So the company turned to China’s open source model Z.ai GLM 5.2 to analyze logs and defend its infrastructure. The incident shocked the industry so much that even Sam Altman suggested that AI development should slow down a bit.
Nadella also revealed that Microsoft is willing to sell its MAI family of models powered by its Maya AI chip, touting it as a cheaper alternative.
“Every customer wants the right model for each task based on quality, latency, cost, and compliance. We offer the most extensive model catalog in the cloud with over 11,000 models, including leading models from OpenAI, Anthropic, Mistral, and xAI, as well as our own MAI family,” he said.
He added, “We are also accelerating our own model development. We have announced more than a dozen new models across images, audio, transcription, coding, and security, including our first inference model, MAI Thinking One, all with cost-effective inference for enterprise use cases at their core. We are co-designing these models with our silicon and Maya. “Running the MAI model at 200 shows a 40% increase in performance per watt.”
What about Mythos? Nadella pointed to Microsoft’s new Mythos competitor, MAI Cyber One Flash, which was announced earlier this week. “It delivers better performance than the much larger Mythos model, but when combined with our multi-agent security harness, it costs half as much,” he said.
Indeed, the Microsoft CEO has said that companies should use a combination of OpenAI and the frontier model provided by Anthropic. But his larger message is: “Don’t trust them enough to rely on them.”
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
