Microsoft investors sweat cloud giant’s OpenAI exposure • The Register

What ought to have been a banner second quarter for Microsoft was met with tepid apprehension on Wall Road on Wednesday, sending its share worth by 6 p.c in after-hours buying and selling.

It may have been a celebratory quarter, as Redmond noticed income surge 60 p.c 12 months over over (YoY) to $38.5 billion on revenues of $81.3 billion. However in query after query, analysts politely pressed CEO Satya Nadella and CFO Amy Hood for assurances that Microsoft’s lopsided reliance on AI mannequin makers – and the large capital expenditures required to serve them – wasn’t going to return again to hang-out the cloud large.

Of explicit concern was the truth that 45 p.c of Microsoft’s $625 billion backlog was instantly attributable to OpenAI.

As a part of OpenAI’s restructuring as a public profit company final fall, Microsoft revealed OpenAI had contracted to buy an incremental $250 billion in Azure companies. In change, Redmond would surrender the proper of first refusal to be OpenAI’s compute supplier.

Hood tried to assuage traders that the opposite 55 p.c of Microsoft’s remaining efficiency obligations had been broad-based and rising at a wholesome tempo, up 28 p.c in the course of the quarter.

“The explanation we talked about that quantity is as a result of 55% or roughly $350 billion is said to the breadth of our portfolio, the breadth of shoppers, throughout options, throughout Azure, throughout industries, throughout geographies,” she mentioned. “That may be a vital RPO steadiness, bigger than most friends, extra diversified than most friends. And albeit, I believe we’ve tremendous excessive confidence in it.”

Omitted of Hood’s analyst commentary, however conspicuously famous in Microsoft’s earnings slides, was the honest portion of that development pushed by OpenAI rival Anthropic.

As chances are you’ll recall, again in November, Anthropic committed to buying $30 billion of Azure compute capability from Microsoft with a further dedication to assist the deployment of as much as a gigawatt of compute capability. However the deal did not come low-cost. In change, Nvidia and Microsoft agreed to speculate half that ($10 billion and $5 billion) respectively again into the AI mannequin dev.

In different phrases, roughly half of Redmond’s backlog is tied up in AI startups that have not but proven they’ll flip a revenue.

To comprehend these revenues, Microsoft must make investments closely in AI infrastructure, and it will not all be its in-house silicon, just like the Maia 200 revealed this week. That gigawatt of compute capability we talked about earlier is tied particularly to deployments of Nvidia Grace Blackwell and Vera Rubin programs.

Talking with analysts on Wednesday, Hood tried to dispel Wall Road’s jitters over Microsoft’s rampant capex spending, which topped $37.5 billion in the course of the quarter. Of that, two-thirds is wrapped up in quick depreciating belongings like GPUs and CPUs, which have simply six years to generate a revenue.

“I believe it is most likely higher to consider the Azure steerage that we give as an allotted capability information about what we are able to ship in Azure income,” she mentioned.

“The vast majority of the capital that we’re spending immediately and a whole lot of the GPUs that we’re shopping for are already contracted for many of their helpful life.” she added when pressed on Microsoft’s potential to realize a significant return on funding the chips aged out. “A lot of that danger that I believe you are pointing to is not there as a result of they’re already offered for the whole thing of their helpful life.”

However as any landlord is aware of, simply because somebody indicators a lease doesn’t suggest they’re good for the hire when the going will get powerful.

Waiting for Q3, Microsoft expects to proceed driving the AI wave. The corporate is forecasting revenues of $80.65 to $81.75 billion, up 15 to 17 p.c year-over-year.

On the similar time, Microsoft does not count on to spend fairly so lavishly in the course of the third quarter, with capex down in comparison with Q2, because of “regular variability from cloud infrastructure construct outs and the timing of supply of finance leases,” Hood defined. ®