Microsoft has developed a significant business in China selling AI models to Chinese companies, including those that are blacklisted by the US, despite growing US-China tensions over artificial intelligence. Social media and AI giant ByteDance is reportedly Microsoft's largest AI customer in recent years, projected to spend over $1 billion annually on Microsoft AI and cloud services. Other major Chinese tech firms, such as Ant Group, Meituan, and Tencent Holdings, are also substantial spenders on AI models through Microsoft's Azure cloud service.
This commercial activity is controversial, particularly because American tech executives and lawmakers perceive China's AI advancements as a potential threat to the US industry. While companies like Anthropic and even OpenAI itself prohibit direct sales of their models to Chinese companies due to intellectual property theft concerns, Microsoft operates under a unique partnership with OpenAI that allows it to set its own policies for selling GPT series models in China. Microsoft states that it uses automated monitoring to prevent customers from building competing products and only sells to established companies, not individual developers, in China.
However, OpenAI has privately expressed concerns to Microsoft that it may not be doing enough to prevent Chinese companies from copying its models, a process known as "distillation." Meanwhile, Chinese AI groups are also bypassing US export controls on high-end chips like Nvidia's A100 by renting access through cloud providers and third-party arrangements, effectively achieving the same capabilities as direct ownership. This highlights the challenges the US faces in enforcing its technology restrictions, as Chinese companies leverage these access methods to further their own AI development. Microsoft's president, Brad Smith, testified that the company's China operation, though controversial, remains relatively small, accounting for about 1.5% of overall revenue in 2024.
