Microsoft eyes DeepSeek to lower AI costs while US restricts top AI models over China fears

Published 17 June 2026 · tech

The artificial intelligence boom is creating strange alliances and even stranger contradictions. For years, American AI companies have accused Chinese firms of copying or "distilling"

The artificial intelligence boom is creating strange alliances and even stranger contradictions. For years, American AI companies have accused Chinese firms of copying or "distilling" their cutting-edge AI models. Washington has repeatedly warned that China could use advanced AI to gain an economic and strategic edge. Yet now, one of America's biggest tech companies may end up relying on a Chinese AI model to keep its own costs under control. Read Full Story According to a report by Axios, Microsoft is exploring the possibility of using DeepSeek, a Chinese open-source AI model, to power parts of its enterprise AI assistant, Copilot Cowork. The reason is simple: AI is becoming expensive. As companies like OpenAI and Anthropic increasingly move towards usage-based pricing, the cost of running AI assistants for heavy users is rising sharply. The more employees use AI, the more providers have to spend on computing power. "We have users who do hundreds of tasks a week, which is great — they're way productive — but the consequence is the costs can go very high," Charles Lamanna, Microsoft's executive vice president for Copilot, agents and platform, told Axios.

Today, Copilot Cowork runs primarily on Anthropic's AI models and also offers compatibility with OpenAI's technologies. But as Microsoft has ended its relationship with OpenAI, the company is now free to look for cheaper and more flexible alternatives. Microsoft's search for a cheaper AI model That search may now lead it to DeepSeek. Microsoft told Axios that it is evaluating a fine-tuned version of DeepSeek V4, or another open-source model, as a lower-cost option. The company expects to announce its choice in the coming weeks. If DeepSeek is selected, Microsoft says customers will have a choice whether to use it. The model would run entirely on Microsoft's Azure cloud platform, meaning customer data would remain inside Microsoft's infrastructure and continue to be protected by Azure's security, compliance and data-residency safeguards. Microsoft also says it has modified the model and added safeguards, including measures aimed at reducing bias. Still, the move could prove controversial.

Why the move could trigger controversy The US government has become increasingly cautious about sharing advanced AI technology. Recently, reports suggested that American authorities asked Anthropic to restrict access to its latest AI models, Fable 5 and Mythos 5, for non-US citizens. The concern, according to reports, was that the models could potentially be jailbroken and exploited by rival nations, including China. Against this backdrop, Microsoft's reported interest in a Chinese AI model underscores how economics is beginning to reshape the AI race. The irony is hard to miss. American companies once worried that Chinese firms were learning from their AI models. Now, one of America's biggest technology companies may be considering a Chinese model because it is cheaper to run. Satya Nadella's warning The development also comes just days after Microsoft CEO Satya Nadella warned about another risk facing the AI industry: dependence on a handful of powerful AI providers.