Microsoft pushes open-weight AI to protect Azure dominance

Microsoft has signed an open letter on AI leadership alongside more than 20 companies including Meta, Nvidia, Hugging Face, and Mistral. The letter, titled "Open Weights and American AI Leadership," argues that U.S. leadership should be measured not by any single frontier AI model but by "whether the United States builds a strong, open ecosystem that diffuses into every sector." The signatories frame open-weight models as tools for spurring innovation and reducing dependence on a handful of providers. They also defend distillation—the practice in which a smaller AI model learns from a more capable model's output—as "a long tradition of learning from, building upon, and improving existing technologies, a tradition that has helped drive innovation since the rise of the open-source software movement." The timing reflects geopolitical concern: Chinese providers have faced criticism for using distillation techniques, and the Trump administration reportedly plans to take action against Chinese open-weight models.

Microsoft's push for open-weight AI is strategically inseparable from the company's cloud and operating system business. Microsoft makes revenue from Azure infrastructure and lock-in created by Windows and Office; more models available on Azure give customers fewer incentives to use rival AI platforms. By promoting a fragmented AI market rather than relying on expensive models from OpenAI and Anthropic, Microsoft lowers costs and improves margins while preventing any single AI lab from becoming powerful enough to threaten its core businesses. CEO Satya Nadella wrote on LinkedIn that the key is using "the right model for each task" while keeping costs down, having previously warned that a few AI models could absorb the value of entire industries.

The company is replacing OpenAI and Anthropic models in GitHub Copilot, Excel, and Outlook with its in-house MAI family. This shift carries clear cost benefits for Microsoft but is less clearly beneficial for customers. Independent benchmarks put MAI well behind OpenAI and Anthropic, with performance roughly matching Deepseek V3.2. Copilot users are receiving weaker AI for the same price while Microsoft improves its margins. Microsoft claims MAI performs "just as well or better," but its benchmarks compare MAI only against smaller, less powerful models like GPT-5.4 Mini, Anthropic's Haiku, and "GPT-5.6" without specifying whether that refers to the Sol, Terra, or Luna variant or which reasoning mode it runs at. Microsoft's own blog explicitly states that the smaller MAI model "significantly lowers the cost of deployment for Microsoft" because it runs on older Nvidia GPUs like the H100 and A100 instead of requiring the latest accelerators. The company has announced more MAI models are planned for Copilot Chat, Outlook, and PowerPoint.

Microsoft's embrace of open-weight AI is framed in the language of ecosystem building and American competitiveness, but the strategy is fundamentally rooted in the company's core business model. Rather than remaining dependent on expensive models from OpenAI and Anthropic—partnerships that limit Microsoft's control and margin—the company is consolidating around its own MAI family while simultaneously championing an open-weight philosophy that fragments the broader AI market. This fragmentation serves Microsoft's interests directly: it prevents any single AI lab from accumulating enough power to threaten Azure's dominance or the lock-in created by Windows and Office.

The timing and framing of the open letter also reflect geopolitical pressures. The Trump administration has reportedly planned action against Chinese open-weight models, and Chinese providers have faced criticism for using distillation techniques (where smaller models learn from larger ones' outputs). By signing the letter alongside industry peers, Microsoft positions itself and the open-weight movement as aligned with U.S. interests—while simultaneously defending the legitimacy of distillation as "a long tradition of learning from, building upon, and improving existing technologies."

However, the gap between Microsoft's stated commitment to openness and its product strategy is substantial. By replacing proven-capable models with MAI—which benchmarks show is materially weaker—across GitHub Copilot, Excel, Outlook, and planned for Copilot Chat, PowerPoint, and elsewhere, Microsoft is extracting cost savings that flow to the company's bottom line rather than to users. Customers receive lower-performing AI at the same price, a trade-off Microsoft does not explicitly acknowledge despite claiming MAI "performs just as well or better."

The company's benchmarking choices—comparing MAI only to smaller models and using vague naming for OpenAI's variants—further signal that Microsoft is aware the performance gap is real and wishes to obscure it. This pattern suggests that Microsoft's open-weight advocacy, while strategically coherent, is less about democratizing AI than about building a competitive moat around Azure that does not depend on any external AI provider.

No comments yet. Be the first to share your thoughts!