Microsoft Champions Open-Weight AI as an Azure Strategy

Microsoft has concluded it is better positioned as an AI orchestrator than as an OpenAI reseller, and is now actively promoting open-weight models — a notable strategic posture for a company that has invested an estimated $13 billion in OpenAI.

In an open letter titled "Open Weights and American AI Leadership," Microsoft joined Meta, Nvidia, Hugging Face, Mistral, and more than 20 other companies in arguing that "leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector." The letter contends that open-weight models — AI systems whose trained parameters are publicly released for download and modification, though typically without full training data or code — spur innovation and reduce dependence on a handful of providers.

The letter also frames "distillation" — a process in which a smaller AI model learns from a more capable model's output — as a legitimate development method. "It reflects 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 letter states.

Chinese providers have drawn particular criticism for employing distillation, and the Trump administration reportedly plans to take action against Chinese open-weight models. That geopolitical backdrop is likely the primary reason Microsoft and other signatories chose to speak out at this time.

Open AI Advocacy Aligned With Cloud Business Model

Microsoft's enthusiasm for openness aligns closely with its business model. The company generates revenue from cloud infrastructure on Azure and the lock-in created by its Windows and Office ecosystem. Hosting more models on Azure gives customers fewer reasons to turn to rival AI platforms. Relying less on costly models from OpenAI and Anthropic also improves Microsoft's margins. This mirrors a broader industry pattern: Amazon Web Services offers dozens of models through its Bedrock service, and Google Cloud's Vertex AI similarly positions itself as a neutral model host.

A fragmented model market additionally 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 using "the right model for each task" while keeping costs down. He had previously warned that a few AI models could absorb the value of entire industries.

In-House MAI Models Replace OpenAI and Anthropic in Copilot Products

Microsoft is replacing OpenAI and Anthropic models in GitHub Copilot, Excel, and Outlook with its in-house MAI family. Whether customers benefit from that swap is questionable. Independent benchmarks place MAI well behind OpenAI and Anthropic, with performance roughly matching Deepseek V3.2.

Copilot customers are effectively receiving weaker AI for the same price while Microsoft improves its margins. The company claims MAI performs equally well or better, but its comparisons are limited to smaller, less powerful models such as GPT-5.4 Mini, Anthropic's Haiku, and "GPT-5.6" — without specifying whether that refers to Sol, Terra, Luna, or which reasoning mode it runs at. Microsoft's "superintelligence team," which produced the comparison, would be aware that such benchmarking methodology is inadequate.

Microsoft does not conceal that cost reduction is the objective. On its own blog, the company states that the smaller MAI model runs on older Nvidia GPUs like the H100 and A100 rather than requiring the latest accelerators, which "significantly lowers the cost of deployment for Microsoft." Additional MAI models are planned for Copilot Chat, Outlook, and PowerPoint. Whether enterprise customers paying per-seat Copilot licensing will accept a quiet model downgrade — or migrate to alternatives — will be a key indicator of how much leverage Microsoft's ecosystem lock-in actually provides.