Microsoft and Meta Push for Open Weight AI on Azure

Microsoft and Meta Push for Open Weights AI on Azure

Microsoft has realized that it is better positioned as an AI orchestrator than as a reseller of OpenAI. The company is now actively advocating for open weights models. In an open letter titled "Open Weights and American AI Leadership," which Microsoft signed alongside Meta, Nvidia, Hugging Face, Mistral, and over 20 other companies, the group argues that "leadership will be judged not by a cutting-edge AI model, but by the ability of the United States to build a robust and open ecosystem that permeates all sectors."

Open weights models are intended to stimulate innovation and reduce reliance on a few suppliers. The letter also mentions "distillation," where a smaller AI model learns from the outputs of a more powerful model, as a legitimate way to build AI. "This reflects a long tradition of learning, building, and improving upon existing technologies, a tradition that has helped drive innovation since the emergence of the free software movement," the letter states.

Chinese suppliers have been particularly criticized for their use of distillation, while the Trump administration apparently plans to take action against Chinese open weights models. This is likely the main reason why Microsoft and the other signatories are speaking out now.

The Company That Invented Software Lock-In Now Defends Open AI

Microsoft's enthusiasm for openness makes more sense when examining its business model. The company generates revenue through cloud infrastructure on Azure and the lock-in created by its Windows and Office ecosystem. More models on Azure give customers fewer reasons to use competing AI platforms. Relying less on expensive models from OpenAI and Anthropic also improves Microsoft's margins.

A fragmented model market also prevents any single AI lab from becoming powerful enough to threaten Microsoft's cloud and operating system businesses. CEO Satya Nadella wrote on LinkedIn that the key is to use "the right model for each task" while keeping costs low. He had previously warned that a few AI models could absorb the value of entire sectors.

Microsoft's Savings Could Come at the Expense of Customers

Microsoft is also replacing OpenAI and Anthropic models in GitHub Copilot, Excel, and Outlook with its own family of MAI models. It is debatable whether customers benefit from this exchange. Independent assessments place MAI well behind OpenAI and Anthropic, with performance roughly equivalent to that of Deepseek V3.2.

Copilot customers are getting a less capable AI at the same price, while Microsoft improves its margins. The company claims that MAI performs just as well, if not better, but only compares it to smaller, less powerful models like GPT-5.4 Mini, Anthropic's Haiku, and "GPT-5.6" without specifying whether it refers to Sol, Terra, Luna, or what reasoning mode it uses. This is not how benchmarks are conducted, and Microsoft's "superintelligence" team, which made the comparison, certainly knows that.

Microsoft isn't even trying to hide that this is about cost reduction. The company states on its own blog that the smaller MAI model runs on older Nvidia GPUs like the H100 and A100 instead of requiring the latest accelerators, which "significantly reduces deployment costs for Microsoft." Other MAI models are planned for Copilot Chat, Outlook, and PowerPoint.