Microsoft Isn’t Abandoning OpenAI, But It’s Definitely Cozying Up to DeepSeek

By Steven Orlowski, CFP, CNPR

Microsoft has been synonymous with OpenAI for the better part of the AI boom, investing billions in the startup that brought us ChatGPT, integrating its models into its products, and positioning itself as the de facto enterprise pipeline for OpenAI’s innovations. But in recent weeks, Microsoft’s actions have made one thing clear: while the OpenAI alliance remains intact, Redmond is increasingly hedging its bets—and one of the names coming up more often in those conversations is DeepSeek.

A Subtle Shift in Strategy

On the surface, nothing has changed. Microsoft continues to embed OpenAI’s GPT models into its Office suite, Azure cloud platform, and Windows Copilot. CEO Satya Nadella has reaffirmed the company’s commitment to OpenAI. But under the radar, Microsoft has been quietly engaging with a new player in the frontier model space: DeepSeek, a Chinese startup that has made waves with its open-weight large language models, most notably DeepSeek-V2.

Microsoft recently began showcasing DeepSeek models in its Azure Model Catalog, a move that reflects growing industry interest in alternatives to OpenAI’s tightly controlled APIs. These models are not just available—they're highlighted, promoted, and increasingly being used in benchmarking tests and enterprise demonstrations. For developers and businesses seeking flexibility, this signals a new era of optionality.

Why DeepSeek?

DeepSeek has rapidly emerged as one of the most competent open-source AI labs globally. Its latest models, trained on massive multilingual corpora and equipped with competitive reasoning and coding capabilities, have drawn favorable comparisons to GPT-4 and Claude 3. While not yet at the top of the leaderboard in all categories, DeepSeek has positioned itself as a credible and increasingly attractive alternative—especially for enterprises that want more transparency and control.

By courting DeepSeek, Microsoft accomplishes a few strategic goals:

• Diversification of AI suppliers: Relying solely on OpenAI creates business risk. Microsoft is smart to broaden its AI partner ecosystem.

• Appealing to open-source advocates: DeepSeek’s open-weight models fit better with Microsoft's GitHub- and developer-focused branding.

• Global AI relevance: DeepSeek is one of the few top-tier labs based in Asia. Collaborating with them signals global reach and cross-border collaboration.

OpenAI Still Reigns—For Now

None of this suggests Microsoft is turning away from OpenAI. The two are still deeply integrated at the infrastructure and product levels. GPT-4-turbo and DALL·E remain embedded in Azure and Office365, and many of Microsoft’s AI ambitions are still tied to OpenAI’s roadmap.

But Microsoft’s moves indicate a broader vision: one in which multiple model providers—including open-source players like Meta, Mistral, and now DeepSeek—compete, coexist, and complement each other. It’s an acknowledgment that the AI space is moving too fast, and too globally, for any one lab to dominate unchallenged.

What’s Next?

Expect Microsoft to continue balancing its AI portfolio, especially as the geopolitics of tech and the demand for model transparency heat up. DeepSeek, meanwhile, is likely to gain more visibility in the West through this partnership—and might soon become a regular name alongside Anthropic, Google DeepMind, and OpenAI.

Microsoft isn’t abandoning OpenAI. But it’s no longer betting the whole farm on one partner either. In the fast-evolving world of AI, even loyalty has a shelf life—and DeepSeek is fast becoming a partner worth watching.