TL;DR — Microsoft’s commercial remaining performance obligation, its contracted backlog, hit $678 billion, up 84% year over year. Azure crossed $100 billion in revenue, but the backlog number is the one that actually predicts the next three years.
On the Q4 call, Satya Nadella spent maybe forty seconds on the number that should have led the transcript. Azure crossed $100 billion in annual revenue for the first time. Good headline, easy clip for the wires. Buried three questions later, almost as an aside, was the figure that actually tells you what the next three years look like: commercial remaining performance obligation hit $678 billion, up 84% from a year earlier. That’s not a growth metric. That’s a contract book. And contract books don’t lie the way guidance does.
The Number That Doesn’t Fit in a Soundbite
Remaining performance obligation, RPO for short, is signed, committed revenue that hasn’t been recognized yet. Think of it as the backlog a contractor would show you before breaking ground: money already promised, work not yet done. Microsoft’s RPO went from $627 billion to $678 billion in a single quarter — a $51 billion jump in thirteen weeks, more than most companies in the S&P 500 will book in annual revenue this year.
Wall Street loves quarterly revenue because it’s clean and comparable. RPO is messier — it’s lumpy, it’s disclosure-dependent, and companies can juice the optics by signing long-duration deals that won’t convert to cash for years. Which is exactly why you have to read past the topline number before getting excited.
Strip Out OpenAI and the Story Still Holds
Here’s the objection every skeptic raises the second they see that 84% figure: it’s all OpenAI, isn’t it? Microsoft’s largest AI partner has been signing compute commitments that could single-handedly inflate a backlog number and make the rest of the business look stronger than it is. It’s a fair question, and Microsoft actually answered it this quarter, which is more transparency than most hyperscalers offer.
Excluding OpenAI, commercial RPO still grew 25% year over year. That’s the number that matters more than the headline 84%. It means the demand signal isn’t a single counterparty’s balance sheet decision — it’s coming from the rest of the commercial customer base, the banks and retailers and manufacturers signing multi-year Azure commitments that have nothing to do with frontier model training runs. One customer can inflate a number. One customer can’t sustain a quarter-over-quarter acceleration across a base that size.
Duration Math: 2.3 Years and Counting
The weighted average duration on that $678 billion book is 2.3 years. Roughly 30% converts to revenue in the next twelve months — itself up 37% year over year — which means the other 70% sits further out, and that portion grew 112% year over year. Translation: Microsoft isn’t just selling more. It’s selling longer. Customers are locking in multi-year Azure capacity at a pace that’s accelerating, not plateauing.
That duration extension is the part the market tends to misprice. A backlog that converts fast is nice for next quarter’s revenue line. A backlog that’s getting longer, at an accelerating rate, is a statement about how customers view their own multi-year compute needs. Nobody signs a three-year cloud commitment on a whim. They sign it because the finance team already ran the model on what workloads are coming, and Azure is where they decided to park that bet.
What This Actually Means for the Multiple
None of this shows up cleanly in a trailing P/E. You can’t screen for backlog duration on most retail platforms, and even institutional models tend to treat RPO as a footnote rather than a forward indicator. But if you’re trying to underwrite Microsoft’s next three years of cloud revenue, the backlog is a better map than the last four quarters of reported growth, because it’s already been signed. It’s not a forecast. It’s a commitment.
The risk, and there is one, is capital intensity. A $678 billion book doesn’t fulfill itself — it requires data centers, power contracts, and chips, all of which Microsoft has to build or buy ahead of the revenue recognition. Backlog growth without capex discipline is how a good story turns into a margin problem. Watch the capex-to-RPO ratio over the next two quarters, not just the backlog headline, before deciding this is an unambiguous win.
The Comparison Nobody’s Making
We’ve written before about Alphabet’s own RPO conversion mechanics, where a $240 billion backlog tells a similar story with a different growth rate. Set the two side by side and the hyperscaler race stops looking like a market-share argument and starts looking like a capital-commitment argument — whoever can fund the buildout without breaking the balance sheet wins the next leg, not whoever has the better model this quarter. Azure’s backlog is bigger in absolute terms; Google Cloud’s is smaller but carries a cleaner margin profile, which we broke down in our piece on Google Cloud’s margin inflection.
Neither backlog guarantees anything. Contracts get renegotiated, capacity gets delayed, and a signed commitment isn’t cash in the bank. But if you’re trying to figure out which hyperscaler has the deepest, longest-duration demand signal sitting on its books right now, $678 billion with 2.3 years of average duration is the number to anchor to — not the $100 billion Azure revenue headline that made the wires.
Sources: Microsoft FY2026 Q4 earnings call and investor materials (quarter ended June 30, 2026, reported July 29, 2026); CNBC earnings coverage. Third Pole Markets holds no position in Microsoft as of publication. This is not investment advice — see our About page for our full disclosure policy.
The backlog figures are disclosed in Microsoft’s investor relations materials and its 10-K filings on SEC EDGAR.
