Microsoft is considering including a Chinese AI model within its Copilot company, and the reason is money.
The company told Axios that it is exploring a streamlined, self-hosted version of DeepSeek V4, or another open source model, as a cheaper option to power Copilot Cowork, the agent assistant in its Microsoft 365 suite. It hopes to offer a lower-cost model within weeks.
At the same time, Microsoft is moving Copilot Cowork to usage-based pricing, charging companies for the compute they actually burn rather than a flat rate.
Why can’t even Microsoft foot the bill?
The change is a window into the economics of agent AI. Tools like Copilot Cowork, Anthropic’s Claude Code, and OpenAI’s Codex call a model over and over while performing a task, which is powerful and, it turns out, expensive.
“We have users doing hundreds of tasks a week, which is great, they are very productive, but the consequence is that the costs can be very high,” said Charles Lamanna, executive vice president of Copilot, agents and platform at Microsoft.
Copilot Cowork currently runs on Anthropic and OpenAI models, both of which have raised prices and removed from all-you-can-eat plans. Microsoft already measured GitHub Copilot for the same reason. A cheaper open source engine is the next obvious lever.
The cheapest option turns out to be the Chinese one.
That’s where the calculation gets complicated. DeepSeek V4, released in April, is open source, popular with developers and much cheaper to run, which is precisely why it is on Microsoft’s list.
He’s also Chinese, and the political moment couldn’t be worse. Washington has proposed banning DeepSeek, threatened Chinese AI companies, and simply forced Anthropic to cut its best models for users outside the US.a dispute that escalated to crisis talks with the Commerce Department.
Microsoft is clearly aware of what this looks like. It says any DeepSeek option would be optional for customers and would be fully hosted in Azure, keeping data within Microsoft’s cloud under its security, compliance and data residency controls, and that it has fine-tuned the model and added safeguards, including changes aimed at reducing bias.
Coverage against your own suppliers
The bigger picture is that Microsoft no longer wants to rely on a single lab. Freed from its tight and often tense exclusivity with OpenAI, it is pushing a multi-model approach, mixing and matching engines under its own roof.
For now, this is an evaluation, not a decision, and Microsoft says it will confirm its choice when the cheapest tier is released. But the fact that he’s even willing to name DeepSeek as a candidate, in this climate, says a lot about how hard it’s started to hit the cost of running agents.
