Microsoft aims to reduce dependency on OpenAI, as it pushes for "AI self-sufficiency"

Microsoft is looking to significantly reduce its dependency on OpenAI, said the company’s AI chief, Mustafa Suleyman. In a recent interview with the Financial Times, Suleymain stated the company is pursuing "true AI self-sufficiency" by developing its own frontier-level foundation models.

These “AI self-sufficiency” claims fit into Microsoft’s October 2025 deal with OpenAI. The revised partnership now allows Microsoft to develop its own frontier models and eventually pursue its own AGI (artificial general intelligence), while keeping the 27% stake in OpenAI and retaining long-term access to OpenAI's models and IP through 2032.

OpenAI models still power core features of Microsoft’s main consumer-facing AI, Copilot. However, dependence on a single supplier poses risks, as any issue OpenAI potentially faces could affect Microsoft’s products that rely on it. Developing in-house frontier models that will be optimized for products like Copilot, Office, and Azure gives Microsoft significant flexibility. Both when it comes to strategic independence and financial efficiency.

In the FT interview, Suleyman explained the shift in strategy:

“Three or four months ago, having renegotiated our partnership, we also decided that this was a moment when we have to set about delivering on true AI self-sufficiency. I mean, this after all is the most important technology of our time.”

This isn't a full break from OpenAI, as Microsoft will continue to rely on it to be its primary partner for many applications. But a diversified approach guards Microsoft from any undesirable scenarios.

Microsoft is already executing this diversified approach. It’s already paying Anthropic to host Claude models on Azure, along with additional models like Meta's Llama, Mistral, and others. At the same time, the company is also investing heavily in developing its own models, most notably the MAI lineup.

All of this requires massive infrastructure and Microsoft is moving quickly to build it. The company has already presented its own AI-accelerator chip, Maia 200, and is currently building the Fairwater network of AI data centers, which includes some of the biggest supercomputers in the world.

Despite hundreds of billions of dollars worth of investments across the industry, nobody can predict in which way will the entire AI timeline eventually develop. Microsoft deliberately chooses not to put all of its eggs in one basket, so it can emerge victorious regardless of the outcome.

Source

Hope you enjoyed this news post. Feedback welcome.

Posted Friday 13 February 2026 at 1:01 pm AEST (my time).

News posts: 2023 5,800+ | 2024 5,700+ | 2025 5,700+ | 2026 (to end of January) 461

RIP Matrix