Microsoft Stock After Q1 FY2027 Earnings: Azure Sped Up While the Backlog Stood Still

"That's not a business as far as I'm concerned." Satya Nadella said it on Microsoft's October 28 call, when Bernstein's Mark Moerdler asked how much of Azure's AI demand comes from ordinary companies. What he was dismissing was having "like 4 customers who buy a lot of capacity from you and you can show a few quarters of growth," in 24/7 Wall St.'s fuller quote.

That's an odd thing to say in a year when the AI cloud race is scored on exactly that. Oracle, Amazon and Google have all put huge AI lab contracts into their backlogs, and by that measure Microsoft had a bad quarter. Its backlog of signed commercial contracts was $677 billion at the end of September, a billion less than in June, and bookings fell 12%.

I think Nadella is right about where the moat is, and the quarter is decent evidence for it. Azure grew 48%, the fastest in the five quarters below, and in the June quarter Microsoft said all of its backlog growth came from customers outside the big AI labs.

The December 2025 jump is mostly one customer. Nadella said in October 2025 that "OpenAI has contracted an incremental $250 billion of Azure services," and the backlog rose $233 billion that quarter. Microsoft said in January that about 45% of the $625 billion was OpenAI, as Constellation Research reported.

What came after is the part I care about. On the June-quarter call, CFO Amy Hood said the backlog excluding OpenAI grew 25% in a year, and that all of the quarter's growth in it came from customers outside the frontier model companies. Microsoft didn't repeat that split in October.

The September quarter needs one caveat. Hood said on the October call that the backlog showed "a smaller sequential decline relative to historical seasonality due to the large Azure commitment this quarter." So June to September usually dips, and this year one big new contract held it level. I'd rather say that plainly than pretend flat means nothing was signed.

Hood put the bookings drop down to "the impact from large Azure commitments in the prior year," partly offset by this quarter's big contract and healthy growth in Microsoft's regular sales. Most of Azure is billed as it's used: a company runs more workloads or calls more models, and the bill rises inside the contract it already has. Nadella said the number of customers with more than 50,000 Copilot seats rose more than eightfold in a year, and Copilot daily users grew nearly 50% in the quarter.

The strongest objection is that Nadella is making a virtue of something that happened to him. OpenAI used to be Microsoft's to host. Over the past year it has signed compute deals with other clouds, and the partnership was loosened again in April 2026. Every large lab contract signed elsewhere is one Microsoft might once have had. On this reading the slow backlog is share loss in the fastest-growing part of the market, and "not a business" is what you say when the business goes to your rivals.

Microsoft's own numbers make that harder to dismiss. If the 45% share from January still roughly holds, OpenAI is the biggest single piece of the $677 billion, so Microsoft is itself one of the companies with a backlog built on a few big customers. On the October call, Goldman Sachs analyst Gabriela Borges put OpenAI at about 11% of Microsoft's revenue in the quarter, and management didn't dispute it. Some of Azure's 48% is likely OpenAI using capacity it signed for last year, and Microsoft doesn't split Azure by customer.

Two more things cut against me. Hood said some third-party GPU capacity Microsoft expected in the quarter was delayed, so Microsoft rents capacity from the kind of business Nadella dismissed. And commercial seat growth fell to 5%, the lowest in the table. If the moat is the Microsoft account a company already has, I'd want that number rising, and it isn't.

I take the first point seriously, and it may be right about growth. I don't think it changes the moat question. The June-quarter disclosure is my best answer: the backlog outside OpenAI grew 25%, and the quarter's growth came from customers who aren't labs. The seat number I can't argue away. My partial answer is that revenue per seat is doing the work, but that needs seats to stop slowing.

The cleanest contrast is Oracle. Its backlog was $664 billion at the end of August, up $26 billion in the quarter. In that quarter it booked more than $30 billion of AI cloud contracts. In June it said most of the growth in the two quarters before came from "large scale AI contracts" where the customer prepaid for the GPUs or supplied them. A customer that brings its own chips is renting a building and an operations team, and it can take its next order elsewhere.

The framework doesn't score backlog. It scores how hard a customer finds it to leave, and it scores the whole company, so Oracle is rated on its database and Amazon and Alphabet on everything they sell. That's the useful part. Amazon and Alphabet have both signed big lab contracts and still rate well on embedding, because the ratings rest on what sits around the compute. CoreWeave, which mostly rents GPUs to a few very large buyers, is the business Nadella described, and it's the only one of the five rated weakened on embedding.

A lab can split its next training run across three clouds. A bank can't split its identity system or its database. The lab contracts are real revenue. What holds those customers in place is the rest of the stack, and that's the part of Microsoft that's growing.

Microsoft reports the December quarter in late January 2027. If Azure lands near the 49% to 50% guide, the growth outside the labs is still carrying it. If Azure comes in below 47%, or Microsoft shows its backlog excluding OpenAI growing less than 20% a year, the June-quarter split was a high point and I'm wrong. The Azure chart above is the one I'll extend, and I'd like Microsoft to give the OpenAI split again.