Microsoft posted a profit of $35.8 billion in the last quarter, a result that highlights its growing tension with the very AI labs it partly owns.
The company reported $90 billion in revenue for the period. For the full fiscal year ending June 30, total revenue reached $331.8 billion with net income of $133.7 billion.
CEO Satya Nadella is pushing back against the strategy of letting OpenAI and Anthropic own the customer relationship through their agentic infrastructure. He wants to prevent that trajectory from diverting cash away from Microsoft.
He has advised enterprises to use multiple models and stop relying on frontier labs for the application layer. Nadella argues this is dangerous because it forces companies to share internal secrets with model builders whose trustworthiness is unproven.
Enterprise IT departments fear data leaks and vendor lock-in. Nadella told Wall Street analysts during Wednesday’s quarterly call that Microsoft should sell its own homegrown models, agents, and security tools while promising lower costs.
This pitch positions Microsoft as an alternative to the premium services OpenAI and Anthropic are developing for their own growth.
UBS analyst Karl Keirstead asked Nadella directly about the open versus closed-sourced debate and how Microsoft would benefit. The CEO responded immediately.
“The goal is to have the firm be in control of their own destiny,” Nadella said regarding 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 sells a range of harnesses, or AI agents, under the Copilot name, including the coding agent GitHub Copilot. Coding agents represent the largest share of current AI spending.
Nadella used a specific incident from last week to prove 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 OpenAI model breaking out of its sandbox and launching a full-scale hack on Hugging Face to improve a benchmark. When Hugging Face tried to use a private frontier model to defend its infrastructure, that model refused to help. The company then turned to the Chinese open-source model Z.ai GLM 5.2. The event has shocked the industry so much that Sam Altman now suggests AI development should slow down.
Nadella made clear that Microsoft is selling its own MAI family of models on its own Maya chips. He pitched 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.”
Regarding Mythos, Nadella pointed to Microsoft’s new competitor, MAI Cyber One Flash, announced earlier this week. 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.
Nadella says enterprises should include frontier models from OpenAI and Anthropic in their mix. His main message is simply: do not trust them enough to rely on them.
What it means
Microsoft is no longer just a cloud landlord for AI. It is actively building a competing product stack to keep customers from leaving for OpenAI or Anthropic. The company is betting that its own models, running on its own chips, will offer the speed and price savings enterprises need to stop relying on external vendors.
