Sponsored

A Quarter That Pulled the Curtain Back on AI Monetization

For nearly two years, Microsoft executives deflected the same question on every earnings call: how many people are actually paying for Copilot? The answer always came wrapped in deliberately vague language — “tens of thousands of customers,” “double-digit growth,” “strong adoption.” On Thursday evening, when Microsoft reported its fiscal third-quarter results for the period ending March 31, 2026, the company finally broke the seal. Microsoft 365 Copilot now has more than 25 million paid commercial seats, up from roughly 12 million at the start of the company’s fiscal year, according to remarks from CFO Amy Hood on the earnings call.

The disclosure transforms Copilot from a strategic narrative into a measurable line of business. At the published list price of $30 per seat per month — though enterprise discounts compress the realized average — the installed base now represents a run-rate north of $9 billion in annualized revenue from a single product line that did not exist three years ago.

The Numbers Behind the Headline

Microsoft’s overall results beat consensus on both lines. Revenue came in at $76.4 billion, up 16% year-over-year, against analyst estimates compiled by Refinitiv of $74.1 billion. Earnings per share landed at $3.62, ahead of the $3.41 expected. Intelligent Cloud, the segment that contains Azure, grew 31%, with Azure itself up 36% in constant currency — an acceleration from 33% in the prior quarter that CEO Satya Nadella explicitly credited to inference workloads.

Three details deserve scrutiny. First, capital expenditures hit $24.8 billion in the quarter, up from $22.6 billion the prior period, with Hood guiding to a similar figure for Q4. The company is now on track to spend roughly $95 billion on infrastructure in fiscal 2026, the vast majority going to GPU-dense data centers. Second, gross margin in the commercial cloud segment compressed by 180 basis points year-over-year to 68%, reflecting the upfront cost of bringing AI capacity online before revenue fully ramps. Third, Copilot attach rates inside the largest enterprise accounts — defined as customers with more than 10,000 seats of Microsoft 365 — now exceed 40%, according to Hood, up from 22% a year earlier.

What Changed in Six Months

The seat-count acceleration did not happen by accident. In late 2025, Microsoft restructured Copilot pricing into three tiers — a free Bing-anchored consumer version, the $30 commercial seat, and a new $50 “Copilot Studio Pro” tier targeted at users who build agents — and bundled the standard Copilot license into existing E5 enterprise agreements at a 30% discount during the renewal window. According to a Gartner client advisory dated April 8, that bundling strategy was the single largest driver of seat growth, accounting for an estimated 60% of net additions in the second half of fiscal 2026.

Microsoft also benefited from a competitive lull. Google’s Gemini for Workspace continued to grow but reached only an estimated 8 million paid seats by Q1 2026, according to a TD Cowen channel survey published last month. Salesforce’s Agentforce, while showing strong attach in customer service workflows, has not penetrated horizontal productivity in the way Copilot has. The result: Microsoft now controls roughly two-thirds of enterprise AI productivity spend.

The Read-Through

For investors, the report validates the core thesis that hyperscale infrastructure spending eventually monetizes through software seats and consumption revenue, not just compute resale. For competitors, it raises the bar: any vendor pitching enterprise AI productivity now has to explain why a CIO would not simply expand an existing Microsoft commitment. And for Microsoft itself, the next test is whether agent workloads — which carry higher per-task pricing but require Copilot Studio Pro adoption — can repeat the trick. Hood declined to break out Studio Pro seat numbers on the call, calling them “early but encouraging.” That answer will not survive another quarter.

L
Lois Vance

Contributing writer at Clarqo, covering technology, AI, and the digital economy.