- Microsoft is adding capacity faster than competitors, but it's not enough to keep up with cloud demand
- The company's backlog jumped 35% year on year
- Analysts seem to be wondering what comes next after the infrastructure rush dies down
Microsoft added a whopping 2 gigawatts (GW) of data center capacity over the past 12 months. But even that heaping helping wasn’t enough for Microsoft’s supply to catch up to insatiable demand for cloud and AI muscle.
The hyperscaler spent $24.2 billion on capex in its fiscal Q4 2025 (calendar Q2), bringing its full-year total to $88.2 billion. It now has more than 400 data centers across 70 cloud regions, CEO Satya Nadella said on an earnings call. French Bank BNP Paribas pointed out in a note to investors Microsoft is bringing new compute online faster than its peers.
And yet, Microsoft’s backlog of unfilled orders only continued to grow, reaching $368 billion. Only 35% of that will be recognized as revenue in the coming year, CFO Amy Hood noted.
The backlog is being driven by monster cloud demand, with Azure posting 39% growth in the quarter. Executives said Azure now has a $75 billion annual revenue run rate and BNP Paribas estimated that Microsoft also ended its fiscal 2025 “with $20bn+ of annualized AI revenues.”
Microsoft financial stats:
Consolidated revenue of $76.4 billion was up 18% year on year, with overall cloud revenue up 27% to $46.7 billion.
Azure and other cloud services revenue grew 39%.
Net income jumped 24% to $27.2 billion.
The company is scrambling to keep up. It is planning to spend even more on capex in fiscal 2026, starting with around $30 billion in FQ1. New Street Research estimated that all told Microsoft will spend $120 billion in capex in FY26.
But still, Hood said Microsoft expects to remain capacity constrained through the next six months.
To be fair, Microsoft isn’t the only hyperscaler in this position. Alphabet execs similarly said during their earnings call last week that Google Cloud will be up against supply constraints through the end of 2025.
But all of this raises the question: what happens once the dog finally catches its tail? In other words, what growth is there for hyperscalers once the infrastructure driver falls away?
Can it last?
Analysts seem to be turning their attention to this question. On the call, there were a lot of questions about monetizing AI and ROI. Nadella pointed to software and applications as differentiators for the company. Microsoft is also betting heavily on AI.
But given predictions that the AI bubble could burst, that answer almost seems risky.
In its downside evaluation of Microsoft’s position, BNP noted “dependency on OpenAI for its GPT model also creates a strategic risk, while the increased dependence on Generative AI for growth also creates risk as it is still an unproven technology in the enterprise.” Indeed, it's not yet clear how the dramatic standoff between Microsoft and OpenAI will be resolved.
David Linthicum, founder of Linthicum Research, noted on LinkedIn that while the ‘AI everywhere’ approach is “a common and effective strategy for driving growth” it “may only work once.”
“The challenge now is to sustain this momentum…The real test will come at the end of the year when the market begins to demand more than integrated AI and starts looking for what’s next,” he concluded.
We'll have to wait and see how Microsoft fares. For now, though, it seems the good times will continue to roll.