Microsoft Copilot Daily Users Up Fivefold as Enterprises Scale AI

What Microsoft reported

Microsoft opened fiscal 2027 with a quarter that beat Wall Street on both revenue and earnings. The company reported revenue of $92.3 billion, up 19% year over year.21 Earnings per share were $4.89 against a consensus estimate of $4.73, and revenue came in above the roughly $90.6 billion analysts expected.30 Shares rose more than 4% in early trading the morning after the report.30

Copilot was at the center of the story Microsoft told investors. Daily Copilot users grew nearly 50% from the prior quarter and more than fivefold from a year earlier.21 The number of customers with more than 50,000 Copilot seats rose more than eightfold year over year.23 Chief executive Satya Nadella named Capgemini, HCLTech, Hitachi Group, Truist Financial and the U.K. tax authority, HM Revenue and Customs, as examples.23 He also said Volkswagen plans to roll Copilot out to more than 330,000 employees.2 Automated tasks nearly doubled after July, and Nadella called Copilot the "new OS for work."23

The wider cloud business picked up speed as well. Microsoft Cloud revenue passed $63 billion, and growth accelerated to 30% year over year.23 Remaining commercial performance obligations, meaning contracted revenue not yet recognized, jumped 73% to $677 billion.30

From productivity tool to operational infrastructure

The most notable change on the call was how Microsoft described its customers. Nadella pointed to Unilever, Rolls-Royce and Raiffeisen Bank International as companies applying AI to operational problems instead of isolated productivity tasks.21 Unilever is automating order processing and trade disputes, and Rolls-Royce is using AI for aircraft engine maintenance and supply chain planning.21

This is a deliberate repositioning. Earlier in the year, analysts argued that the real enterprise value might come from automation more than from augmentation. They noted that Copilot Studio users had built hundreds of thousands of custom agents within a single quarter.29 Evercore ISI's Kirk Materne described the move from Copilot to agents as Microsoft's most important strategic shift since its cloud pivot, provided agents deliver real process automation.29 The Unilever and Rolls-Royce examples are Microsoft's answer to that test, though they are anecdotes chosen by the company, not audited results.

The agent story is clearest in software development. Agents now open more than a third of all pull requests on GitHub, and agent-generated pull requests rose sixfold over six months.23 Microsoft also wants to show that governance is keeping pace. Customers have used Microsoft Purview to audit more than 7.6 billion agent activities, a fivefold increase from the prior quarter.23

The quiet shift from seats to consumption

The least flashy part of the call matters most for anyone trying to read Copilot's commercial path. The number of Copilot customers actively consuming credits through usage-based billing grew more than 75% in the quarter.23 Nadella said Microsoft's business model is expanding from per-seat subscriptions to include consumption, which widens both the addressable market and revenue per user.2

One analysis noted that Nadella's prepared remarks gave five usage or consumption growth figures and no total Copilot seat count.2 The last disclosed figure is still the more than 30 million paid seats reported for the fiscal year that ended in June.2 Microsoft also does not disclose consumption revenue or the number of customers on usage billing, so the 75% figure is a growth rate with no published base.2

Chief financial officer Amy Hood offered a measured caveat: she expects usage-based pricing to catch on more slowly outside software development.21 The underlying numbers support a cautious reading. Microsoft 365 commercial cloud revenue grew 17%, while commercial seats grew only 5%.2 Hood attributed part of that seat softness to weaker sales execution in Microsoft 365 E3 and standalone products.23 My reading is that Microsoft is now growing mainly by charging existing customers more, through Copilot, E5, the new E7 bundle (which Hood said already has millions of paid seats) and metered consumption, rather than by adding many new users.23 Hood's guidance that December-quarter growth will come mainly from average revenue per user, with a smaller contribution from seats, points the same way.2

How fast adoption has actually moved

The seat trajectory before this quarter was steep. Microsoft disclosed 15 million paid Microsoft 365 Copilot seats in January, more than 20 million in April, and more than 30 million at the end of July.27 The jump from April to July added about 10 million seats in one quarter, the fastest growth since launch.1 Large commitments drove much of it. Accenture's more than 740,000 seats are Microsoft's largest Copilot win to date, and Bayer, Johnson & Johnson, Mercedes and Roche each committed to 90,000 or more seats.25 HSBC committed to 200,000 seats.3

Skeptics have a fair counterpoint about penetration. Thirty million seats against a commercial Microsoft 365 base of about 464 million works out to roughly 6.5%, a third-party estimate that Microsoft has not published itself.28 One independent tracker estimates paid seats at about 36.2 million, or 7.7% of commercial seats, but that is a modeled figure, not a company disclosure.8 Barclays' Raimo Lenschow previously called the Copilot numbers "the elephant in the room" given Microsoft's infrastructure spending.29 Research cited by critics found that Copilot's share of paid AI subscribers fell from 18.8% to 11.5% between July 2025 and January 2026, as ChatGPT and Gemini gained.7

The two sides largely agree on the facts and differ on how to frame them. The bullish view is that large enterprises are moving from pilots to company-wide licenses. The bearish view is that more than nine in ten commercial seats still lack a paid Copilot license. Both are accurate. The fivefold rise in daily users matters because it speaks to the strongest objection to seat counts: that a paid license says nothing about whether anyone uses it.28

What the ROI evidence shows, and what it doesn't

Return on investment remains the weakest part of the enterprise AI story. Commissioned studies do exist. A Forrester Total Economic Impact study estimated a three-year net present value of $19.7 million and a 116% ROI for a composite enterprise.13 A separate Forrester study of GitHub Enterprise Cloud calculated a 376% three-year ROI. That model, however, combines Copilot with Actions, Advanced Security and platform consolidation, so it is not a standalone Copilot figure.15

The customer stories are more concrete. PwC says its employees carried out more than 20 million Copilot actions in April 2026 and freed up more than one million hours of capacity, with 54% of its global workforce using AI tools weekly.20 Bristol Myers Squibb reported 96% adoption across a 30,000-person rollout and 25,000 agents in use.28 Other published results include DWF cutting a seven-day contract process to seven hours and Commercial Bank of Dubai saving 39,000 hours a year.13

Commentators have pushed back hard on these numbers. One analysis put it plainly: Copilot's seat count proves purchasing momentum, not satisfaction, and vendor case studies do not add up to comparable, audited ROI.28 Another described how hours-saved figures can be produced by multiplying headcount by an estimate nobody checks.16 Consultancies selling deployment services promote first-year ROI of 150% to 400%, but those projections depend on their own adoption assumptions.11

The most useful finding from that same consulting literature is about activation. One firm reports that organizations without a funded adoption program tend to plateau at 15% to 30% of licensed seats active, while those that invest in enablement reach 55% to 70%.18 Whatever the exact figures, the pattern fits Microsoft's own emphasis this quarter on daily usage and consumption instead of raw license counts.

The bottom line

The quarter supports the view that enterprise AI adoption is accelerating, especially among the largest companies. Daily users, giant deployments and metered consumption all rose sharply.2123 Still, the evidence that this adoption pays off is mostly anecdotal and commissioned, and Microsoft has stopped giving the headline seat count that once anchored the story.2

Microsoft is betting that agents running real business processes, such as order handling, engine maintenance and code review, will produce returns that are easier to measure than minutes saved on email. The capital at stake is large. Capital expenditures reached $45.2 billion this quarter, about three-quarters of it on short-lived assets such as chips.30 Management says AI demand still exceeds supply and expects capex to rise above $50 billion next quarter.23 The next few quarters will show whether consumption revenue, which Microsoft has not yet disclosed, becomes large enough to report as its own figure.