Microsoft has swiftly integrated DeepSeek’s R1 reasoning model onto its Azure cloud platform and GitHub, a move that comes shortly after legal scrutiny of the China-based AI company. This integration signifies a notable step in expanding AI accessibility and potentially diversifying Microsoft’s AI partnerships.
Microsoft plans to offer “NPU-optimized” versions of the DeepSeek AI model, specifically designed for enhanced compatibility with Windows 11 Copilot+ PCs. The rollout will begin with a version for Qualcomm Snapdragon X devices, followed by Intel Lunar Lake PCs, and finally, a variant for AMD Ryzen AI 9 processors. Furthermore, DeepSeek-R1-Distill-Qwen-1.5B, along with 7B and 14B versions, will be incorporated into the Microsoft AI Toolkit, empowering developers to build and deploy AI-driven applications leveraging the power of NPUs in Copilot+ PCs. Microsoft highlighted these advancements in a recent blog post.
This strategic move aligns with Windows 11 Copilot+ PC requirements for AI model processing, including a minimum of 256GB storage, 16GB RAM, and an NPU capable of 40 TOPS output, as outlined by Windows Central. Older NPUs may lack the capacity to run these models locally.
Reuters suggests Microsoft’s embrace of DeepSeek could signify a move to reduce reliance on OpenAI for its AI needs. This strategy involves developing proprietary models and integrating third-party models like DeepSeek to enhance Microsoft 365 Copilot. This partnership may also address privacy and data-sharing concerns, particularly given DeepSeek’s data servers being located in China, a potential challenge for U.S. users, according to Reuters.
However, Microsoft is also investigating allegations of DeepSeek employing illegal practices to train its models, specifically concerning the potential misuse of OpenAI’s intellectual property, as reported by a White House official. This concern stems from DeepSeek’s alleged use of “distillation,” a teacher-student model dynamic, to extract data from OpenAI’s code, a claim supported by prior research. DeepSeek markets itself as an open-source model with low operating costs, particularly on lower-powered Nvidia chips.
Despite controversies surrounding censorship and potential jailbreaking by users, DeepSeek is rapidly gaining prominence in the tech world. This latest collaboration with Microsoft positions DeepSeek for further growth and innovation in the AI landscape.
