Microsoft’s open-weight push is an Azure play, and Indian users should take note

Microsoft has signed an open letter urging the United States to embrace open-weight AI models, a move that the company’s own business model suggests is less about openness and more about driving customers to its Azure cloud. The letter, titled “Open Weights and American AI Leadership,” was co-signed by Meta, Nvidia, Hugging Face, Mistral, and over 20 other companies. It argues that US leadership in AI depends on building a strong, open ecosystem rather than relying on a few frontier models. The letter also defends “distillation” — training a smaller AI model on the output of a more capable one — as a legitimate innovation method. This comes amid growing US criticism of Chinese AI providers for using distillation, with the Trump administration reportedly planning to take action against Chinese open-weight models. Microsoft’s timing appears designed to influence that policy debate.

Open letter and the Azure play

Microsoft’s enthusiasm for open-weight models makes sense when viewed through its business model. The company profits from Azure cloud infrastructure and the lock-in created by Windows and Office. More AI models running on Azure give customers fewer reasons to use rival platforms. Relying less on expensive models from OpenAI and Anthropic also improves Microsoft’s margins. A fragmented model market prevents any single 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. The open letter aligns with that vision: by promoting open-weight models, Microsoft positions itself as an AI orchestrator rather than a reseller of OpenAI’s technology.

MAI models and performance concerns

Microsoft is simultaneously replacing OpenAI and Anthropic models in GitHub Copilot, Excel, and Outlook with its in-house MAI family. Independent benchmarks place MAI well behind OpenAI and Anthropic, with performance roughly matching Deepseek V3.2. Copilot customers are receiving weaker AI for the same price, while Microsoft improves its margins.

The company claims MAI performs just as well, but only compares it against smaller, less powerful models like GPT-5.4 Mini, Anthropic’s Haiku, and an unspecified “GPT-5.6.” Microsoft’s own blog admits that the smaller MAI model runs on older Nvidia H100 and A100 GPUs instead of requiring the latest accelerators, which “significantly lowers the cost of deployment for Microsoft.” More MAI models are planned for Copilot Chat, Outlook, and PowerPoint.

Key facts at a glance

Aspect	Detail
Open letter signatories	Microsoft, Meta, Nvidia, Hugging Face, Mistral, and 20+ others
MAI performance vs competitors	Roughly matches Deepseek V3.2; behind GPT-5.4 Mini, Haiku
Impact on Copilot users	Same price, weaker AI; Microsoft gets better margins
Planned MAI expansion	Copilot Chat, Outlook, PowerPoint
GPU cost reduction	Runs on older H100/A100 GPUs instead of latest accelerators

Implications for Indian users

Indian enterprises and startups that rely on Azure for AI workloads should pay attention. Open-weight models could reduce dependence on foreign API providers and lower costs, but Microsoft’s in-house models may not deliver the same quality as the OpenAI or Anthropic models they replace. For developers using GitHub Copilot, the change could mean less accurate code suggestions. Indian AI startups building on Azure may also face a fragmented ecosystem where Microsoft favours its own models.

The open letter, while focused on US policy, has global consequences. If the US restricts Chinese open-weight models, Indian developers who rely on those models for research or commercial projects could be affected. The debate over distillation also touches on intellectual property norms that are still evolving.

What remains unclear

It is not yet known how quickly Indian enterprises will notice the performance difference, or whether Microsoft will offer pricing tiers that reflect the lower quality of MAI models. The company’s benchmark comparisons are disputed by independent evaluators, and no third-party audit of MAI performance on Indian workloads is available. The actual impact on Indian users will depend on the scale of MAI deployment and the availability of alternative models on Azure.

Source: The Decoder – Microsoft’s open-weight AI push is so obviously an Azure play it hurts (https://the-decoder.com/microsofts-open-weight-ai-push-is-so-obviously-an-azure-play-it-hurts/)