Microsoft's biggest cloud customer is also its 24%-owned affiliate

1. Summary: what this piece answers

This piece answers one question for a Microsoft holder: after the September quarter, is the company still a broad toll gate on the superintelligence buildout, or is it becoming a capital-heavy supplier to one related party? The answer matters because the stock is priced for the buildout to pay back, and the filing now gives enough data to test how concentrated that payback is.

Here is what you will learn. First, how much of Microsoft's revenue now comes from OpenAI, an equity-method affiliate in which Microsoft holds about 24% on an as-converted basis.

The 10-Q puts September-quarter OpenAI revenue at $10.0B, or 10.84% of the $92.28B total, our calculation ^[1]. The 10-K puts fiscal 2026 OpenAI revenue at $24.1B, or 7.26% of that year's $331.84B, our calculation ^[3].

The share is rising. Second, what the capex surge is doing to cash: cash capex was $39.37B in the quarter, against $19.39B a year earlier, and quarterly free cash flow fell 27% to $18.63B, our calculation ^[1].

Third, what the price assumes. On our numbers, enterprise value is about $3.96T, 83.63x trailing owner free cash flow, and our base case lands at 1.01x today's value in five years.

The headline still holds. OpenAI is Microsoft's largest disclosed customer, and it is a related party under ASC 850 ^[1].

Azure grew 48% to $33.17B, which is the strongest growth line in the company ^[2]. The filing does not say how much of that Azure growth came from OpenAI.

2. Thesis and the toll-gate test

The thesis is that Microsoft owns the connection from the enterprise seat to compute. A company that runs on Microsoft 365, Entra identity, GitHub and Azure routes its agents, data and model calls through Microsoft's stack.

Every new superintelligence workload that lands on that stack pays a toll, either per seat, per token or per hour of compute. The September quarter shows the toll is growing: Microsoft Cloud revenue was $63.9B, up 30%, and commercial remaining performance obligation rose 73% to $677B ^[1][2].

The counterweight is concentration. One affiliate, OpenAI, paid $10.0B in the quarter, and receivables from it rose to $7.2B from $6.0B at June 30 ^[1].

A toll gate is most valuable when many travelers pay it. It is less valuable when one traveler pays a tenth of the tolls and the gate owner also owns a quarter of that traveler.

The gate result is three passes and one mixed, which is a passing gate under our method.

What invalidates the thesis: OpenAI keeps rising as a share of revenue while Azure growth slows below the pace of Microsoft Cloud overall, which would show the gate narrowing to one buyer. A second break would be receivables from OpenAI rising faster than OpenAI revenue for two more quarters, since that would mean Microsoft is financing its own customer.

3. Azure and the superintelligence buildout

Azure is now disclosed in dollars. Revenue was $33.17B in the September quarter, against $22.38B a year earlier, growth of 48.16%, our calculation ^[2].

That is the fastest-growing product line Microsoft reports, and at 35.94% of total revenue it is also the largest, our calculation ^[2]. Microsoft Cloud, which includes Azure, Microsoft 365 commercial cloud and other cloud services, reached $63.9B and grew 30%, or 31% in constant currency ^[2].

The new segment structure puts Azure inside Agents and Infra, which combines applications, agents, Microsoft's multi-model system and Azure infrastructure ^[1]. Agents and Infra revenue rose 25% to $76.80B, and its operating income rose to $40.56B from $32.72B ^[2].

Cost of revenue in that segment rose faster than revenue, to $23.89B from $17.70B, which is the margin cost of the buildout showing up ^[2].

The 10-Q says Microsoft Cloud gross margin percentage fell to 66%, "driven by sales mix shift to Azure, offset in part by product-level margin improvements, primarily in Azure" ^[1]. That sentence is the core of the Azure story.

Each dollar of Azure carries more depreciation and power than a dollar of Microsoft 365, so the faster Azure grows, the lower the blended margin, even as Azure itself becomes more efficient.

4. Capex and capacity

Cash capex, reported as additions to property and equipment, was $39.37B in the September quarter, up from $19.39B a year earlier ^[1]. Over the trailing 12 months it was $135.92B, our calculation from the 10-K and the 10-Q ^[1][3].

Operating cash flow over the same period was $195.88B, so capex took 69.39% of operating cash, our calculation.

Quarter by quarter, capex has climbed steadily: $19.39B in September 2025, $29.88B in December, $30.88B in March, $35.80B in June and $39.37B in September 2026 ^[1][3][4][5]. Free cash flow moved the other way, from $25.66B a year ago to $18.63B, a fall of 27% ^[1].

Capex is only part of the bill. Microsoft also takes capacity through finance leases, which add assets and liabilities without running through the cash capex line.

Finance lease liabilities reached $70.97B at September 30, from $66.59B at June 30 and $54.43B a year earlier ^[1][6]

5. The affiliate relationship and concentration

The 10-Q describes Microsoft's stake in OpenAI as "an investment accounted for under the equity method that represents an approximate 24% interest on an as-converted basis" ^[1]. Because OpenAI is an equity-method investee, it is a related party under ASC 850, and Microsoft must disclose revenue and receivables from it ^[1].

That disclosure is what makes this piece possible.

The numbers: September-quarter revenue from commercial arrangements with OpenAI, inclusive of revenue-sharing payments, was $10.0B ^[1]. Accounts receivable from OpenAI were $7.2B at September 30, up from $6.0B at June 30 ^[1].

Microsoft has committed $13.0B of funding to the investment and had funded $12.0B by September 30, up from $11.9B at June 30 ^[1][3]. For fiscal 2026, OpenAI revenue was $24.1B ^[3].

That puts OpenAI at 7.26% of fiscal 2026 revenue and 10.84% of September-quarter revenue, our calculations ^[1][3]. A trailing 12-month figure cannot be built, because Microsoft did not disclose OpenAI revenue for the September 2025 quarter on its own.

6. Microsoft 365, Copilot and seat pricing

Microsoft 365 cloud revenue was $27.78B in the quarter, up 17.24% from $23.69B, our calculation ^[2]. The 10-Q says Microsoft 365 commercial cloud growth was "driven by higher revenue per user, reflecting continued upsell to premium seat offerings, including Microsoft 365 Copilot, E5, and E7, as well as increased consumption of usage-based services, primarily in GitHub Copilot" ^[1].

Commercial seats grew 5%, driven mainly by small and medium business and frontline worker offerings ^[1].

This is where the toll-gate pricing test is mixed. Most of Microsoft 365 is still sold per seat, and seat growth is 5%.

The growth above that comes from selling a more expensive seat, not from charging for outcomes. In a world where superintelligence agents do more of the work, a per-seat model has a ceiling, because the number of human seats does not grow with the number of agents.

Microsoft has not disclosed a paid Copilot seat count in the filing.

GitHub Copilot points the other way. The 10-Q names it as the main source of "usage-based services" consumption inside Microsoft 365 commercial cloud ^[1].

Usage pricing ties revenue to work done, which is closer to outcome pricing. If Microsoft moves more of Microsoft 365 Copilot toward consumption, the pricing gate would move from mixed toward pass.

Two public posts from Satya Nadella show where management is pushing. They are context only, not a source for any figure.

Satya Nadella (@satyanadella), Oct 22, 2026, on X: "Super to see Volkswagen Group putting Copilot to work across the company!" https://x.com/satyanadella/status/2113182815189942357

7. Gaming, LinkedIn, Windows and the rest

The Devices and Consumer segment shows the other side of the mix shift. Segment revenue fell 3% to $15.48B, while operating income rose to $5.36B from $5.24B on lower cost of revenue and operating expenses ^[2].

The segment is now a cash contributor rather than a growth engine.

Windows OEM and devices revenue fell 21.97% to $3.55B from $4.55B, our calculation, the steepest decline of any line ^[2]. The 10-Q defines the line as Windows licenses sold through the OEM channel plus Surface and PC accessories ^[1].

A decline of this size points to a weak PC channel, and the filing does not suggest a near-term turn.

XBOX revenue was $5.38B, down 2.34% from $5.51B, our calculation ^[2]. XBOX content and services, which covers first- and third-party games, Game Pass and cloud gaming, fell 2%, or 1% in constant currency ^[2].

Gaming is no longer adding growth after the Activision years.

Search and advertising revenue was $6.55B, up 10.30%, our calculation ^[2]. This line now includes Bing, Edge, LinkedIn Marketing Solutions and LinkedIn Premium Subscriptions, MSN and third-party affiliates ^[1].

8. Cash flow and owner free cash flow

We follow our approved method: free cash flow is operating cash flow minus all capex, and owner free cash flow is free cash flow minus stock pay. All figures are trailing 12 months to September 30, 2026, built from the fiscal 2026 10-K plus the September 2026 quarter minus the September 2025 quarter ^[1][3].

Operating cash flow was $195.88B. Cash capex was $135.92B.

Free cash flow was therefore $59.96B, a margin of 17.31% on trailing revenue of $346.45B, our calculations ^[1][3]. Stock-based compensation from the cash flow statement was $12.56B, or 3.62% of revenue and 20.94% of free cash flow, our calculations ^[1][3].

Owner free cash flow was $47.40B, a 13.68% margin, our calculation.

In the quarter alone, operating cash flow rose to $58.00B from $45.06B, helped by a $17.10B swing in accounts receivable as collections came in after the June quarter ^[1]. Capex rose faster, so free cash flow fell to $18.63B from $25.66B ^[1].

Stock pay was $3.13B against $2.98B

9. The balance sheet and capital returns

Cash and cash equivalents were $20.28B at September 30, and short-term investments were $44.23B, for a total of $64.51B in cash, equivalents and short-term investments, the filing's wider figure ^[1]. Under our method only the $20.28B of cash and cash equivalents is netted in enterprise value.

Debt at carrying value was $35.48B: $4.39B current and $31.09B long-term ^[1]. Microsoft repaid $4.85B of debt in the quarter, taking it down from $40.29B at June 30 ^[1][3].

Against operating cash flow of $195.88B, the bond debt is small.

The larger obligation is leases. Finance lease liabilities were $70.97B, split $5.01B current and $65.97B long-term, with a weighted average remaining term of 13 years and a 4.5% discount rate ^[1].

Operating lease liabilities on the balance sheet rose to $21.72B long-term from $16.53B ^[1]. Our method excludes operating leases and follows the filing's debt total for finance leases, which the filing does not include in long-term debt.

Counting finance leases as debt would add $70.97B to enterprise value, about 1.79%, our calculation.

Capital returns continue but are modest next to capex. Microsoft paid $6.76B of dividends and repurchased $4.38B of stock in the quarter, and returned $10.2B to shareholders in total

10. Valuation and the five-year cases

Enterprise value is about $3.96T, our calculation. It uses 7,426,194,552 basic cover-page shares at the prior close, plus $35.48B of debt at carrying value, minus $20.28B of cash and cash equivalents ^[1].

On trailing figures that is 11.44x revenue of $346.45B, 19.31x EBITDA of $205.32B, 66.11x free cash flow of $59.96B and 83.63x owner free cash flow of $47.40B, our calculations ^[1][3]. Stock pay takes 20.94% of free cash flow.

Netting the $44.23B of short-term investments would lower enterprise value to about $3.92T, and adding finance leases would raise it to about $4.03T, our calculations.

Five-year view, our calculations. We start from trailing revenue of $346.45B, a free cash flow margin of 17.31% and stock pay at 3.62% of revenue ^[1][3]. Growth rates, margins and exit multiples below are our assumptions, not company guidance.

The bear case keeps capital intensity near today's level, so free cash flow margin stays at 17% while growth slows to 10%. The base case needs 15% growth and a recovery in free cash flow margin to 24%, which requires capex to grow slower than revenue from fiscal 2028. The bull case assumes Azure keeps compounding, backlog converts and margin reaches 28%, close to Microsoft's pre-buildout level.

11. Risks

Customer concentration in a related party. OpenAI paid $10.0B in the quarter, 10.84% of revenue, up from 7.26% across fiscal 2026 ^[1][3]. The 10-Q risk factors warn that the expected benefits from strategic relationships, "including through commercial arrangements, technology access, and Azure consumption, may not be realized or sustained," and that partners "may likewise adjust their purchasing decisions" ^[1].

Because Microsoft does not split OpenAI's share of Azure or of the $677B backlog, an investor cannot size the exposure precisely. If OpenAI moved workloads to other clouds or slowed spending, the hit would land on the fastest-growing line.

Receivables and financing of the customer. Receivables from OpenAI rose to $7.2B from $6.0B in one quarter, while Microsoft has funded $12.0B of its $13.0B commitment to the same company ^[1]. Microsoft both funds and bills the affiliate.

If receivables keep growing faster than revenue, more of the reported revenue would depend on OpenAI's ability to raise outside capital.

Capital intensity and return on capital. Trailing capex of $135.92B took 69.39% of operating cash flow, and quarterly free cash flow fell 27% ^[1][3]. Depreciation is rising, with depreciation, amortization and other up to $11.74B from $8.15B ^[1]

12. Catalysts and conclusion

The next catalyst is the fiscal second-quarter 10-Q, covering the December 2026 quarter. Three numbers in it decide whether this hold becomes a buy or a watch.

First, OpenAI revenue as a share of the total: a share above 10.84% would confirm rising concentration, and a share below it would show Azure broadening ^[1]. Second, receivables from OpenAI against the $7.2B at September 30 ^[1].

Third, capex against the $39.37B run rate and backlog against $677B: capex that catches up while backlog keeps growing above 30% would show the September shortfall was timing, not a slowdown.

Any disclosure of OpenAI's share of commercial remaining performance obligation would settle the largest open question in the story. Microsoft has never given it, so we do not assume it.

Stock position: hold. Microsoft owns the record and the connection, Azure grew 48%, and backlog is $677B. Against that, a tenth of revenue comes from a 24%-owned affiliate, owner free cash flow is less than a quarter of EBITDA, and our base case only matches today's value.

The call fits the hold band: base case 1.01x with a passing gate of three passes and one mixed. We would move to buy only if concentration falls while Azure growth holds, and to watch if OpenAI's share keeps climbing alongside its receivables.

What it means for the future of software. The superintelligence buildout is being sold by the gatekeepers that own compute, and the first large buyers are the model labs those gatekeepers part-own. Software economics now depend on whether demand for compute broadens from a few model labs to the long tail of enterprises running their own agents.

If it does, the owners of the connection, Microsoft first among them, earn a toll on all of it. If it does not, the gatekeepers have turned into capital-heavy suppliers to a handful of related customers, and software multiples will compress to match.