- In an exclusive WSJ interview, Microsoft CEO Satya Nadella warned that the public won’t tolerate “a few models and companies doing all of the learning for the world,” and criticized AI companies that simultaneously predict catastrophic job losses and demand unlimited resources for data-center expansion.
- Microsoft is rolling out low-cost AI models, launching Copilot Cowork (an autonomous agent letting users choose across multiple AI models including cheaper ones), and is actively weighing whether to host DeepSeek on its Copilot platform — a move that would benefit the Chinese AI provider at the direct expense of OpenAI and Anthropic.
- Nadella skewered executives using AI purely as a cost-cutting tool: “No, how about we think about reorganizing the jobs?” — arguing companies of the future must have both “token capital” (in-house AI capability) and human capital, operating as a “continuous learning system.”
- The comments mark a striking strategic pivot: Microsoft invested billions making OpenAI what it is today and also has a multi-billion-dollar deal with Anthropic, yet is now positioning itself to commoditize the frontier models both companies built — with Copilot already losing subscribers to Google Gemini in H2 2025.
What Happened?
Satya Nadella sat for an exclusive WSJ interview to outline Microsoft’s new strategic posture in the AI race — one that amounts to a direct challenge to the frontier-model companies Microsoft helped build. Without naming OpenAI, Anthropic, or Google, Nadella attacked the internal logic of companies that warn of existential AI risks while demanding unlimited compute resources: “You can’t say, hey, all white-collar jobs are gone and this could even be a weapon and we will use all the power to build data centers.” Microsoft has already rolled out a suite of low-cost models and launched Copilot Cowork, an autonomous AI agent that lets enterprise users select among multiple models — including cheap alternatives — to complete long-running tasks. The company is also weighing hosting DeepSeek on Copilot, which Axios first reported, a step that would dramatically expand the Chinese model’s reach at OpenAI’s and Anthropic’s expense.
Why It Matters?
Microsoft trails OpenAI, Anthropic, and Google in frontier model development — and Copilot lost subscribers to Gemini in the latter half of 2025. Rather than compete head-on in a model race it appears to be losing, Microsoft is repositioning to commoditize frontier AI: drive prices down, enable model choice, and frame itself as the enterprise platform that connects workers to AI rather than the company that replaces them. The DeepSeek question is particularly pointed: hosting it would signal that Microsoft is willing to undercut its own OpenAI investment to win the platform war. Nadella’s framing — AI companies must “earn social permission” rather than impose change — is also a direct political shot at Anthropic CEO Dario Amodei’s prediction that AI could wipe out half of entry-level jobs by 2029, and at OpenAI’s history of safety-driven unlimited-scaling narratives.
What’s Next?
The DeepSeek hosting decision is the most immediate watch item — if Microsoft proceeds, it would mark a watershed in the AI platform wars and a direct provocation to OpenAI. Copilot Cowork’s enterprise rollout will test whether multi-model choice resonates with corporate buyers more than single-model depth. More broadly, Nadella’s June 14 essay and this interview signal that Microsoft is building a political and commercial case for “AI democratization” — a narrative likely to grow more powerful as AI’s labor-market effects become more visible ahead of midterm elections and regulatory battles over AI’s role in the economy.
Source: The Wall Street Journal
