Satya Nadella criticizes AI labs for restricting distillation, calling it hypocritical and urging companies to own their AI infrastructure.

Microsoft’s chief executive Satya Nadella delivered a pointed critique of AI research labs that limit the sharing of model distillation techniques, labeling the practice as “hypocritical” and urging firms to take full ownership of their AI infrastructure.

Nadella’s Main Argument

In a recent interview, Nadella argued that restricting model distillation—where large models are compressed into smaller, more efficient versions—undermines the broader goal of democratizing AI. He emphasized that companies should open up their pipelines rather than gatekeep valuable research outcomes.

According to Nadella, “If you build a powerful model and then prevent others from creating lighter versions, you’re stifling innovation and creating an uneven playing field.”

Why Distillation Matters

Distillation enables smaller firms and startups to run advanced AI capabilities on modest hardware, lowering costs and expanding access. It also accelerates deployment in edge devices, where computational resources are limited.

  • Reduces inference costs","Broadens AI accessibility","Speeds up product iteration

Responses from the AI Community

Some AI labs, including Anthropic, have defended their policies, citing concerns over misuse and intellectual property. They argue that unrestricted sharing could lead to the rapid proliferation of powerful models without adequate safety measures.

Industry observers note that the tension reflects a broader debate: balancing open research with responsible deployment.

We must protect the integrity of our models while still fostering collaboration, said an Anthropic spokesperson.

Nadella’s remarks come at a time when Microsoft is deepening its partnership with OpenAI and investing heavily in its own Azure AI infrastructure, positioning the company to offer end‑to‑end solutions that bypass third‑party restrictions.

He concluded by urging AI developers to “own the stack” and make their technologies available for downstream innovation.

Read the full report Source coverage.