Microsoft’s Copilot revenue sounds big. The adoption number tells a very different story.

• Microsoft revealed it has 15 million paying users of Microsoft 365 Copilot, priced at $30 per user per month. On its face, that’s several billion dollars a year in high-margin software revenue.
• But Microsoft also has more than 400 million Office 365 users worldwide. Do the math: 15 million is roughly 4% of the installed base. And that’s paying, not actively using.
• Internally, Microsoft executives have already acknowledged the harder problem: many customers who pay for Copilot barely use it at all. Revenue booked does not equal product pull.
• This makes Copilot less a breakout adoption story and more a pricing experiment riding on enterprise inertia: bundle pressure, pilots, and executive mandates rather than bottom-up demand.
• The comparison set reinforces the point:
  • Google has about 8 million Gemini Enterprise subscribers, also a small slice of Workspace users.
  • OpenAI reports 35 million ChatGPT subscribers, plus 7 million business users, suggesting far stronger voluntary pull.
  • Anthropic won’t disclose subscriber counts but admits Claude subscription revenue is growing fast from a smaller base.
• Microsoft’s disclosure mattered for another reason: Azure growth is slowing relative to AI capex. On the earnings call, CFO Amy Hood admitted that if all new Nvidia-equipped capacity had gone to Azure, growth would have been higher than the reported 39%. Instead, capacity was reserved for Copilot and internal AI work.
• That’s the real signal. Microsoft is diverting scarce, expensive compute away from cloud rentals to support Copilot, even though only ~4% of Office users are paying and fewer are meaningfully engaged.
• Nadella framed this as portfolio optimization: software margins beat infrastructure margins. That’s true. But it also exposes the bet: Copilot must eventually justify the compute it consumes by converting hundreds of millions of reluctant Office users, not by extracting $30 from a thin early-adopter layer.
• Microsoft says it has reduced Copilot’s compute costs. It hasn’t disclosed margins. Until usage rises materially, efficiency gains are compensating for weak adoption, not amplifying strong demand.

Bottom line: Copilot revenue is real, but it’s not proof of mass-market AI adoption inside Office. At ~4% penetration—and with usage lagging payments—it looks less like a runaway product and more like a bridge strategy to keep AI capex defensible while Microsoft searches for a version of Copilot people actually want to use.