Azure sales of OpenAI models to Chinese tech groups are surging even as OpenAI and Anthropic avoid the market over intellectual property and misuse risks.
Microsoft has carved out an unusual role in the U.S.-China AI rivalry by selling OpenAI’s GPT models to major Chinese technology companies through Azure, despite OpenAI and Anthropic choosing not to serve the market directly. ByteDance, Ant Group, Tencent and Meituan use Microsoft’s Azure AI services, with ByteDance described as Microsoft’s largest AI customer in recent years and on track to spend more than $1 billion annually on the company’s AI and cloud offerings. Azure’s AI revenue in China grew faster than in any other sales region, roughly tripling in the fiscal year ending June 2025 after a 400% increase the previous year, though Brad Smith said the China business accounted for about 1.5% of Microsoft’s total revenue in 2024. OpenAI has pressed Microsoft to strengthen safeguards against Chinese customers using GPT outputs for distillation, a technique for training smaller competing models from a larger model’s responses. Microsoft keeps OpenAI models off servers in China and serves customers from nearby data centers such as Singapore, while saying it uses automated monitoring and limits sales to established companies. At the same time, some U.S. developers are shifting to lower-cost Chinese models such as DeepSeek, Xiaomi’s MiMo and Minimax, and Microsoft is testing a fine-tuned version of DeepSeek-V4 for its Copilot Cowork enterprise agent.