Microsoft sells OpenAI models to Chinese firms via Azure despite tensions

Microsoft has quietly built a substantial business selling OpenAI’s GPT models to Chinese companies via its Azure cloud platform, even as US-China tensions over AI escalate. ByteDance, the parent company of TikTok, is Microsoft’s biggest AI customer and is on track to spend over $1 billion annually on Microsoft’s AI and cloud services. Other major Chinese tech firms—including Ant Group, Meituan, and Tencent—are also significant spenders on Azure AI, using models for software development, customer service automation, and more. Microsoft’s China operation remains small (about 1.5% of 2024 revenue) but is growing rapidly: Azure AI revenue in China tripled in the fiscal year ending June 2025 and surged 400% in 2024. This growth has drawn criticism from US lawmakers and rivals like Anthropic and OpenAI, who refuse to sell directly to Chinese companies over IP theft and national security fears.

Microsoft’s unique partnership with OpenAI allows it to set its own policies for selling GPT models in China, bypassing OpenAI’s direct restrictions. The company markets a range of AI models, excluding Anthropic’s, and employs automated monitoring to prevent customers from building competing products. However, it’s impossible to fully stop “distillation”—using OpenAI models to train rival models. OpenAI has privately complained to Microsoft about insufficient safeguards. Meanwhile, Chinese clients like ByteDance and Ant Group develop their own AI models (e.g., ByteDance’s Doubao chatbot) and use Azure primarily for non-China expansion. Microsoft argues its China presence is necessary to stay abreast of local innovation and serve multinational customers, but the arrangement highlights the complex, profitable entanglement of Western AI with Chinese tech giants.