Despite the escalating technological rivalry between Washington and Beijing, Microsoft has quietly established a highly lucrative business selling advanced artificial intelligence models to some of China’s largest technology companies. A new report reveals that ByteDance, Ant Group, and Meituan are among Microsoft’s biggest AI customers, utilizing the American tech giant’s enterprise services to access cutting-edge capabilities that their domestic alternatives cannot yet match for global deployments.
The arrangement highlights a complex loophole in the current geopolitical landscape. While leading AI labs like OpenAI and Anthropic have explicitly refused to offer their services in China, Microsoft provides access to OpenAI’s models through its Azure cloud computing platform, which operates globally. This allows Chinese firms to leverage American AI technology for their international operations and internal development, according to a Bloomberg report cited by The Next Web.
The Azure Loophole and Its Commercial Scale
ByteDance is generally Microsoft’s single biggest AI customer globally and is on track to spend more than $1 billion a year on Microsoft’s AI and cloud services. The striking aspect of the relationship is what ByteDance is primarily buying: OpenAI models, sold through Azure, in a market that OpenAI itself will not serve directly. Both OpenAI and Anthropic decline to sell their models to companies in China, citing concerns about intellectual property theft and potential harmful use. Microsoft, by virtue of its unusual partnership with OpenAI, sets its own China policy and still sells the GPT series there.
ByteDance is not alone. Ant Group, Meituan, and Tencent are also significant buyers of AI models through Azure. Internally, Microsoft has treated this as a strategic win. At a July 2025 sales meeting, then chief commercial officer Judson Althoff told staff that Azure’s AI revenue was growing faster in China than in any other territory, roughly tripling in the financial year to June 2025 after a 400 percent surge the year before.
(Related: When Americans Choose Chinese AI: Cost, Quality, and the Quiet Shift in Model Adoption)
Strategic Implications for the Tech War
The revelation of Microsoft’s extensive business with Chinese tech giants underscores the difficulty of enforcing a complete technological decoupling between the US and China. While Washington has focused heavily on restricting the flow of physical hardware, specifically advanced semiconductors, the fluid nature of cloud computing and software services presents a more complex regulatory challenge. China accounted for about 1.5 percent of Microsoft’s overall revenue in 2024, a figure that understates the strategic sensitivity of the relationship given the nature of what is being sold.
There are limits Microsoft does observe. Under its OpenAI agreements, it does not host the models in its Chinese data centers near Beijing and Shanghai, out of concern for intellectual property theft. Customers instead access them online from facilities in other countries, such as Singapore. OpenAI has privately complained that Microsoft is not doing enough to prevent Chinese firms from copying its models through a process known as distillation. Microsoft says it uses automated monitoring and restricts sales to established companies rather than individual developers, but Chinese customers are not subject to heightened scrutiny, and synthetic-data training is difficult to prevent. The situation exemplifies the structural tension at the heart of the US-China tech war: the same companies that lobby for export controls on hardware continue to profit from software and services that flow across the same geopolitical boundary.
The political durability of this arrangement is uncertain. Congressional scrutiny of US technology companies’ China relationships has intensified, and the revelation that ByteDance, the parent of TikTok, which Congress attempted to ban, is Microsoft’s single-largest AI customer globally will inevitably draw attention on Capitol Hill. Microsoft’s position is that it is operating within the law and that restricting cloud services to Chinese companies would simply push them toward domestic alternatives, weakening US influence rather than strengthening it. That argument has merit, but it is unlikely to satisfy lawmakers who view any commercial relationship with Chinese technology companies as a national security risk. The coming months will test whether Microsoft’s China AI business can survive the political environment it has helped create.
For Chinese technology companies, the Microsoft relationship is also a strategic vulnerability. Dependence on American cloud infrastructure for global operations creates a single point of failure that could be severed by executive order or legislative action. ByteDance in particular has reason to be concerned: its TikTok operations have already been subject to forced divestiture legislation, and a future administration could extend similar logic to its Azure dependency. The urgency with which Chinese companies are building domestic AI capabilities is partly a response to exactly this risk, the recognition that any technology stack with an American component is potentially subject to the same geopolitical disruption that has already reshaped the semiconductor supply chain.
