Microsoft Hits 20M Paid Copilot Seats — and the Engagement Numbers Match Outlook

Microsoft disclosed during its Q3 FY2026 earnings call on April 29 that it now has over 20 million paid seats for Microsoft 365 Copilot, up from 15 million in January 2026 — a 33% step-up in three months. More importantly, CEO Satya Nadella framed the engagement story directly: "Copilot queries per user were up nearly 20% quarter over quarter. Weekly engagement is now at the same level as Outlook." That last comparison — Copilot weekly active engagement matching Outlook, the most embedded productivity tool in enterprise IT — is the metric Microsoft has been waiting two years to disclose.

The deal-size disclosure that matters more than 20M

Buried under the headline number was a more telling figure: the count of Microsoft customers buying more than 50,000 Copilot seats has quadrupled year-over-year. That's a strategic signal, not a vanity stat. 50K-seat deployments aren't pilots — they're enterprise standards-setting commitments where Copilot becomes the default AI tooling for entire workforces. Named reference customers include Bayer, Johnson & Johnson, Mercedes-Benz and Roche, each at 90,000+ Copilot seats. The largest publicly disclosed deployment is Accenture, which is rolling Copilot out to approximately 743,000 employees — at full deployment, that single customer accounts for roughly 3.7% of Microsoft's entire reported paid Copilot base.

Why the Outlook engagement comparison is the line that matters

Three reasons. First, Outlook is the highest-frequency tool in Microsoft 365 — most knowledge workers open it within minutes of starting their day and return to it dozens of times. Saying Copilot weekly engagement matches Outlook means workers are returning to Copilot habitually, not opening it once per pilot week. Second, the comparison quietly answers the "shelfware" critique that has dogged enterprise AI rollouts since 2024 — the worry that companies were paying $30/seat/month for tools nobody actually used. Third, it gives Microsoft's salesforce a specific benchmark for upsell conversations: a CIO who has bought 10K seats but sees engagement well below the Microsoft-cited Outlook level can be told, plausibly, that the implementation rather than the product is the problem.

The number you should still be skeptical of

Paid seats are not the same as active users. Microsoft's disclosure is about who has bought access, not how many of those seats are being used in any given week. Reporting from Directions on Microsoft, The Motley Fool, and other industry analysts has flagged in prior quarters that some large enterprise Copilot purchases were partially-used sitting on the shelf, with utilization concentrated among 10–20% of the seat-holder population. The Outlook-engagement comparison is meant to put that critique to rest, but Microsoft notably did not disclose a weekly-active-user count, only the per-user query growth rate. Until Microsoft (or a credible third party) discloses an active-seat-to-paid-seat ratio, the headline 20M is best read as upper-bound usage, not realized usage.

BlockAI News' View

Two angles worth noting. First: this disclosure lands in the same week as Google Cloud's "compute constrained" admission and OpenAI's Stargate progress disclosure. Together they show the AI buyer market is bid-up in three places at once: enterprise productivity (Microsoft), enterprise platform (Google Cloud), and frontier model capacity (OpenAI). The supply side cannot keep up at any layer. Second: 20M paid seats × ~$30/seat/month implies roughly $7.2B in annualized Copilot subscription revenue, before usage-based or premium-tier overages. That's now meaningful enough to materially impact Microsoft's Productivity & Business Processes segment growth rate, and helps explain why Nadella was comfortable signaling "exploit" mode on the OpenAI rev-share restructure two days earlier — the Copilot economics work even with margin compression elsewhere.

Want every AI × Web3 signal the moment it breaks? Subscribe to the BlockAI News daily brief.

How we report: This article cites primary sources, regulatory filings, and on-chain data where available. BlockAI News uses AI tools to assist with research and first-draft generation; every article is reviewed and edited by a human editor before publication. Read our full How We Report page, Editorial Policy, AI Use Policy, and Corrections Policy.