Satya Nadella's $37B AI Run Rate: The July 29 Earnings Breakdown

Microsoft's AI business crossed $37 billion in annual revenue run rate during the quarter ending March 31, 2026, up 123 percent year on year, per the company's Q3 FY2026 earnings release. Azure grew 40 percent year on year in the same period. With Microsoft reporting Q4 FY2026 results after market close on July 29, this is a breakdown of how CEO Satya Nadella has structured that revenue across Azure, Copilot, and the OpenAI partnership.

Azure at 40 Percent Growth and the $190 Billion Infrastructure Bet

The 40 percent year-on-year growth in Azure during Q3 FY2026 was the company's highest quarterly growth rate in seven quarters, per the Microsoft Q3 FY2026 earnings conference call. The company's earnings transcript attributed a meaningful portion of the acceleration to Azure AI services consumption, with enterprise customers running AI workloads at a pace that has outstripped Microsoft's own initial capacity planning.

Behind that growth sits a capital commitment of unusual scale. Microsoft spent $31.9 billion on capital expenditures in Q3 FY2026 alone and guided Q4 capex to exceed $40 billion, putting total calendar-year 2026 spending on track for roughly $190 billion, per management guidance cited by TradingKey. On the earnings call, Microsoft confirmed that AI capacity is allocated first to its own products, M365 Copilot and GitHub Copilot, before becoming available to external Azure enterprise customers, which helps explain why Azure AI Foundry customer wait times have been a recurring topic in enterprise IT circles.

The platform underpinning that capacity is Azure AI Foundry, which now hosts more than 1,900 curated AI models from Microsoft and third-party providers alongside over 10,000 open-source models from Hugging Face, per Microsoft's Build 2026 announcements. The model breadth is designed to let enterprises run proprietary AI workloads on Microsoft infrastructure without locking into a single model provider. The implied Q3 FY2025 AI ARR, derived from the 123 percent year-on-year growth figure, is approximately $16.6 billion, meaning the business roughly doubled in 12 months.

Copilot's 20 Million Seats and the Revenue Gap

Microsoft 365 Copilot crossed 20 million paid enterprise seats in Q3 FY2026, up 250 percent year on year, with 5 million seats added in that single quarter, the fastest sequential seat growth since launch, per Microsoft's official Q3 results post on X. Approximately 70 percent of Fortune 500 companies now use M365 Copilot in some capacity, according to Microsoft's own disclosures. GitHub Copilot separately serves more than 20 million users across roughly 140,000 organizations, with enterprise customers growing 55 percent quarter on quarter per Microsoft Build 2026.

The largest single deployments show where enterprise adoption is concentrating. Consulting firm Accenture holds 740,000 M365 Copilot seats, which Nadella described as Microsoft's largest Copilot win to date at Build 2026. The UK's National Health Service has deployed M365 Copilot across more than 500,000 NHS staff, a scale that required a bespoke data-governance arrangement under NHS privacy rules. Both deployments were cited in Microsoft's Build 2026 announcements.

The seat counts are large but the monetization math narrows them. At the M365 Copilot list price of $30 per seat per month, 20 million seats implies a $7.2 billion nominal annual run rate. Analysts at Citi and J.P. Morgan, cited by Vaasblock, have documented enterprise discounting in the 40 to 60 percent range for competitive Copilot deals, putting real annual M365 Copilot revenues closer to $2 to $4 billion. Microsoft does not break Copilot revenue out separately from its $37 billion AI ARR figure, which also includes Azure AI consumption. The April 2026 move to usage-based billing for GitHub Copilot, adding a token-metered consumption tier on top of flat-rate seats, indicates the company sees enough demand to charge for incremental usage rather than simply maximising seat count.

The $228 Billion OpenAI Position and the "Token Capital" Thesis

Microsoft has invested approximately $13 billion in OpenAI across three rounds between 2019 and 2023, securing a roughly 26.79 percent economic interest, per AiFundingTracker.com. At OpenAI's latest implied valuation of $852 billion, that stake is worth an estimated $228 billion, a 17.6x return on the invested capital. The financial impact is already flowing through Microsoft's income statement: the company's 10-Q for the nine months ending March 31, 2026 disclosed $5.9 billion in net investment gains from the OpenAI position, a sharp reversal from $2.7 billion in net losses during the same period a year earlier, as detailed by Om Malik citing the filing directly.

The OpenAI relationship has also shaped how Nadella frames Microsoft's long-term AI strategy. In a June 2026 interview with Stratechery's Ben Thompson, Nadella laid out what he described as a "token capital" framework: every company, he argued, will need to build human capital (knowledge, judgment, and relationships) alongside token capital, the AI capability a firm builds and owns rather than rents via API calls from an outside provider. The essay version of that argument, published on his X account, positions Azure AI Foundry as the enterprise layer for building proprietary AI systems. The practical implication is that Microsoft is selling not just access to OpenAI's models through Azure, but the infrastructure for enterprises to build AI capabilities that would be theirs to own and refine independently of any single model vendor.

Microsoft reports Q4 FY2026 results after market close on July 29, 2026. Management's Q3 guidance language pointed to FY2027 double-digit revenue and operating income growth. The central question for analysts on the call will be whether the $190 billion in annual capex is translating into Azure AI revenue at a rate that justifies the spending, or whether the conversion timeline is extending as enterprise AI projects move from pilot to production more slowly than initially modelled. Sam Altman's OpenAI portfolio breakdown published here earlier this month showed the $852 billion valuation context from the other side of the partnership.

What It Means

The $37 billion AI ARR, a 123 percent increase in 12 months, positions Microsoft as the largest disclosed enterprise AI infrastructure vendor by revenue run rate. Nadella has built a three-layer stack: Azure as the compute layer, OpenAI models as the capability anchor, and Copilot as the enterprise application surface. The $190 billion in annual capital expenditure is the bet that AI infrastructure supply will remain the primary enterprise bottleneck through at least 2027 or 2028. The July 29 Q4 FY2026 earnings will show whether that conversion rate held through the final quarter of the fiscal year, and whether Azure AI growth acceleration continued past the 40 percent mark set in Q3. For context on how a peer is managing the same infrastructure trade-off, see the Demis Hassabis / Alphabet breakdown published yesterday.

Sources

• Microsoft: Cloud and AI strength fuels Q3 FY2026 results
• Microsoft: FY2026 Q3 Earnings Conference Call transcript
• Microsoft on X: Q3 FY2026 results summary
• Stratechery: An Interview with Microsoft CEO Satya Nadella About Finding Core Competencies
• Om Malik: What Microsoft's 10-Q Says About OpenAI
• AiFundingTracker: Who Owns OpenAI? Complete Ownership Breakdown (2026)
• Vaasblock: Microsoft Copilot 3.3pct Penetration vs $190B AI Capex
• Fortune: The 33-year-old executive Satya Nadella is trusting to fix Microsoft's Copilot AI assistant

Editorial standards: every claim is sourced. Tips: [email protected]