What should have been a banner second quarter for Microsoft was met with tepid apprehension on Wall Street on Wednesday, sending its share price down by 6 percent in after-hours trading.
It could have been a celebratory quarter, as Redmond saw profits surge 60 percent year over over (YoY) to $38.5 billion on revenues of $81.3 billion. But in question after question, analysts politely pressed CEO Satya Nadella and CFO Amy Hood for assurances that Microsoft’s lopsided reliance on AI model makers – and the massive capital expenditures required to serve them – wasn’t going to come back to haunt the cloud giant.
Of particular concern was the fact that 45 percent of Microsoft’s $625 billion backlog was directly attributable to OpenAI.
As part of OpenAI’s restructuring as a public benefit corporation last fall, Microsoft revealed OpenAI had contracted to purchase an incremental $250 billion in Azure services. In exchange, Redmond would give up the right of first refusal to be OpenAI’s compute provider.
Hood attempted to assuage investors that the other 55 percent of Microsoft’s remaining performance obligations were broad-based and growing at a healthy pace, up 28 percent during the quarter.
“The reason we talked about that number is because 55% or roughly $350 billion is related to the breadth of our portfolio, the breadth of customers, across solutions, across Azure, across industries, across geographies,” she said. “That is a significant RPO balance, larger than most peers, more diversified than most peers. And frankly, I think we have super high confidence in it.”
Left out of Hood’s analyst commentary, but conspicuously noted in Microsoft’s earnings slides, was the fair portion of that growth driven by OpenAI rival Anthropic.
As you may recall, back in November, Anthropic committed to purchasing $30 billion of Azure compute capacity from Microsoft with an additional commitment to support the deployment of up to a gigawatt of compute capacity. But the deal didn’t come cheap. In exchange, Nvidia and Microsoft agreed to invest half that ($10 billion and $5 billion) respectively back into the AI model dev.
In other words, roughly half of Redmond’s backlog is tied up in AI startups that haven’t yet shown they can turn a profit.
To realize those revenues, Microsoft will have to invest heavily in AI infrastructure, and it won’t all be its in-house silicon, like the Maia 200 revealed this week. That gigawatt of compute capacity we mentioned earlier is tied specifically to deployments of Nvidia Grace Blackwell and Vera Rubin systems.
Speaking with analysts on Wednesday, Hood attempted to dispel Wall Street’s jitters over Microsoft’s rampant capex spending, which topped $37.5 billion during the quarter. Of that, two-thirds is wrapped up in fast depreciating assets like GPUs and CPUs, which have just six years to generate a profit.
“I think it’s probably better to think about the Azure guidance that we give as an allocated capacity guide about what we can deliver in Azure revenue,” she said.
“The majority of the capital that we’re spending today and a lot of the GPUs that we’re buying are already contracted for most of their useful life.” she added when pressed on Microsoft’s ability to achieve a meaningful return on investment the chips aged out. “Much of that risk that I think you’re pointing to isn’t there because they’re already sold for the entirety of their useful life.”
But as any landlord knows, just because someone signs a lease doesn’t mean they’re good for the rent when the going gets tough.
Looking ahead to Q3, Microsoft expects to continue riding the AI wave. The company is forecasting revenues of $80.65 to $81.75 billion, up 15 to 17 percent year-over-year.
At the same time, Microsoft doesn’t expect to spend quite so lavishly during the third quarter, with capex down compared to Q2, due to “normal variability from cloud infrastructure build outs and the timing of delivery of finance leases,” Hood explained. ®