Microsoft AI model business in China grows fast as OpenAI stays out

Microsoft is expanding its AI model business in China by selling access to OpenAI’s GPT series and other advanced models via Azure cloud. The move comes as OpenAI and Anthropic have largely avoided China due to concerns over intellectual property theft and model misuse. Key figures and numbers: Microsoft’s Azure AI revenue in China tripled in the fiscal year ended June 2025, after growing 400% the year prior. ByteDance (TikTok’s parent) is projected to spend over $1 billion annually on AI and cloud services through Azure. Ant Group, Meituan, and Tencent are also customers. Risk controls: Microsoft sells access through Azure with “guardrails.” It uses automated monitoring and restricts availability to established companies rather than open consumer access. While Azure has data centers in China, the article says OpenAI models are not hosted there. Instead, Chinese users access models remotely through data centers outside China, a routing approach aimed at making model-weight theft harder. Policy and industry context: Microsoft co-founded the Frontier Model Forum (2023), which focuses on tackling model distillation and improving AI safety coordination. The article notes that Chinese firms like DeepSeek can build competitive models. Why it matters for traders: The Microsoft AI model business in China could face regulatory pressure if concerns about distillation or misuse escalate, especially if access restrictions tighten after broader U.S. chip export limits to China. The near-term impact is likely more sentiment- and policy-driven than directly tied to crypto fundamentals.