Microsoft claims 20M paid Copilot users—and claims they actually use it

Microsoft dropped some numbers this week that made me stop scrolling: 20 million paid Copilot users. That’s up from 10 million six months ago. And before you roll your eyes and mutter “but how many actually use it?”, Microsoft says they’ve got data on that too.

The company claims daily active users are up 150% year over year, and the average paid user is interacting with Copilot across multiple Microsoft 365 apps—Word, Excel, Teams, Outlook. Not just opening it once and forgetting about it.

I’ve been skeptical about the “AI assistant” hype since day one. We’ve seen this movie before: big launch, early adopters, then a slow fade into irrelevance. Remember Clippy? Yeah. But something feels different here. Microsoft is embedding Copilot into workflows people already have, not asking them to learn a new tool. That matters.

Let’s talk about the 20 million number for a second. That’s paid seats, not free trial users or people who accidentally clicked “try Copilot” once. These are organizations and individuals paying $30/month per user for Copilot for Microsoft 365. At scale, that’s serious revenue—roughly $7.2 billion annually if all users are on the full enterprise plan. Microsoft didn’t break down the mix between enterprise, business, and consumer, but even if half are on cheaper plans, we’re still talking billions.

Compare that to where we were a year ago. In early 2025, analysts were openly questioning whether Copilot would ever gain traction. The narrative was that employees tried it, got bored, and went back to their old workflows. Microsoft’s own internal data apparently showed that at one point. So what changed?

I think it’s a combination of things. First, the product got better. Copilot now handles context across documents, emails, and meetings in ways it couldn’t before. Second, Microsoft started forcing the issue—Copilot is increasingly baked into the default experience, not an optional add-on. Third, and this is the cynical take, companies that already paid for Microsoft 365 saw Copilot as a relatively cheap upsell and bought it without a clear plan. Some of those seats are probably collecting dust.

But the engagement data suggests more than dust collection. Microsoft says Copilot is used for specific, repeatable tasks: summarizing email threads, drafting responses, pulling data from Excel into PowerPoint, generating meeting recaps. These aren’t flashy use cases, but they’re practical. They save time in small ways that add up.

I still have questions. How does Microsoft define “daily active user”? Is it someone who does one Copilot query per day, or multiple? What’s the churn rate on those 20 million paid seats? And how much of this growth is driven by bundling versus genuine demand? Microsoft didn’t share those details, and I doubt they will.

Still, 20 million is a real number. It’s higher than I expected, and the growth trajectory is faster than what we’ve seen from standalone AI tools like ChatGPT‘s enterprise tier or Google’s Gemini for Workspace. Microsoft’s advantage is distribution: they already have hundreds of millions of Office users. Getting even a fraction of them to pay extra for AI is a massive business.

The “nobody uses Copilot” meme is probably overblown at this point. Are there still seats going unused? Sure. Is every organization getting value? Definitely not. But 20 million people (or their employers) are paying for it, and enough of them are using it that Microsoft felt confident enough to brag about engagement. That’s not nothing.

I’ll be watching the next earnings call closely. If Microsoft breaks out Copilot revenue separately, we’ll have a clearer picture. Until then, I’m taking the 20 million number at face value—with a healthy dose of skepticism about what “using it” really means.