#How is Microsoft Expanding Its AI Business in China?
Microsoft is making significant strides in expanding its artificial intelligence operations in China. The tech giant is incorporating models from DeepSeek, a local firm, into its Azure cloud platform. This move occurs in a complex environment where other AI leaders, like OpenAI and Anthropic, express concerns over potential technology theft by Chinese companies.
Microsoft partnered with OpenAI, Anthropic, and Google to establish the Frontier Model Forum. This initiative, which began in April 2026, aims to combat practices that involve training smaller models based on outputs from larger, more expensive AI systems. As this unfolds, Microsoft is actively testing DeepSeek’s latest model for integration into its enterprise products.
#What Are DeepSeek Models and Why Are They Cost-Effective?
DeepSeek's R1 model was added to Azure AI Foundry and GitHub in January 2025. With over 1,800 AI models now available to business clients, DeepSeek’s models stand out due to their significantly lower training and operational costs compared to similar Western technologies. This economic advantage is particularly important as Microsoft evaluates its options for enterprise product development.
As of June 2026, Microsoft is testing DeepSeek-V4 specifically for its Copilot Cowork applications. The reason behind this is straightforward—rising inference costs associated with models from OpenAI and Anthropic have made it imperative for Microsoft to find economically viable alternatives for tasks that do not necessitate the highest-end models.
#How Does Microsoft Navigate Chinese Market Regulations?
It's important to note that Microsoft does not directly sell its proprietary AI models in China. Instead, it leverages Azure to use Chinese-developed models, enabling the company to provide solutions to international clients without breaching U.S. export regulations on sensitive technologies. Furthermore, Microsoft plans to establish an incubator for Chinese tech startups in 2026, fostering innovation while navigating regulatory challenges.
#What Concerns Exist About Model Distillation?
Concerns regarding model distillation practices have been brought to the forefront by OpenAI and Anthropic. The industry recognizes it as a method whereby smaller, less expensive models are trained using outputs from larger models. Allegations have emerged that some Chinese laboratories have reverse-engineered Western AI models, creating cheaper alternatives. These practices have gained significant scrutiny, especially in light of tightening U.S. restrictions targeting Chinese AI companies.
#What Are the Implications for Investors?
Investors need to understand the shifting landscape of AI technologies. Companies are sorting their AI demands into different tiers, opting for expensive models for complex tasks while relying on cost-effective alternatives like DeepSeek for daily operations. If Microsoft seeks affordable options, it is likely that other enterprises will follow suit. This shift could complicate OpenAI's profitability as customers may increasingly choose cheaper models available through Azure.
Ultimately, tighter U.S. regulations on Chinese AI could necessitate the removal of DeepSeek models from Azure, which would not only eliminate a competitive cost advantage but also ease pricing pressures on models from other partners like OpenAI and Anthropic.