Azure accelerates to 43%, and the capex bill arrives with it
Azure at 43% and an 11.8% EPS beat, landing in the name that had been the laggard of the Mag 7 — which is exactly why it moved sentiment as much as it did. The business is in good shape. After a run from the $350s to a shade under $500, the entry price no longer is. Moving to Hold.
4 Aug 2026 · Q4 FY26 earnings · 4 min read
Sources
Summary
June-quarter revenue of $90.0bn, up 18%, and normalised EPS of $4.74 against a $4.24 consensus — an 11.8% beat, the fourth in a row. Intelligent Cloud did the work at $39.3bn, up 32%, with Azure up 43%; Copilot seat additions carried Productivity to +14%; More Personal Computing fell 4%, the only segment going backwards. The figure worth most attention sits outside the income statement, though: commercial RPO rose 84% to $678bn, and that is contracted revenue rather than pipeline. Serving it cost $35.8bn of property and equipment in the quarter and $115.9bn across a year that closed with revenue of $331.8bn, up 18%.
| Metric | Reported | YoY |
|---|---|---|
| Revenue | $90.0bn | +18% |
| Operating income | $40.6bn | +18% |
| Diluted EPS (GAAP) | $4.81 | +32% |
| Intelligent Cloud | $39.3bn | +32% |
| Azure and other cloud | — | +43% |
| Microsoft Cloud | $59.3bn | +27% |
| Productivity and Business Processes | $37.8bn | +14% |
| More Personal Computing | $12.9bn | -4% |
| Commercial RPO | $678bn | +84% |
| Additions to property and equipment | $35.8bn | — |
Against the Street, this was not a one-off. Microsoft has now beaten normalised EPS consensus for four consecutive quarters, and the June beat was the largest of them bar one:
| Metric | Reported | Consensus | Surprise |
|---|---|---|---|
| Q1 FY26 (Sep 2025) | $4.13 | $3.66 | +12.7% |
| Q2 FY26 (Dec 2025) | $4.14 | $3.92 | +5.7% |
| Q3 FY26 (Mar 2026) | $4.27 | $4.07 | +4.9% |
| Q4 FY26 (Jun 2026) | $4.74 | $4.24 | +11.8% |
Positives and negatives
Positives
- The demand is contracted, not forecast. Commercial RPO up 84% to $678bn is signed business. Azure accelerating to 43% at this revenue base is the harder achievement — the law of large numbers should be working against it and isn't.
- Operating leverage held through the build. Operating income grew 18%, in line with revenue, and 21% for the full year against 18% revenue growth. The capex has not yet shown up as margin compression, which is the thing most feared about this cycle.
- The constraint is supply, not customers. Management added roughly a gigawatt of capacity in the quarter and 31 datacentres, 88 across the year, and expects to roughly double total capacity within two years. Companies do not build into weak demand.
- Copilot is turning into seats. Seat additions were strong enough to show up in Productivity and Business Processes growth, and that matters beyond the revenue: it gives the AI story a recurring, per-seat shape rather than one funded entirely by capital expenditure.
Negatives
- $115.9bn of full-year capex, two thirds of it short-lived. Management was explicit that most of the spend is CPUs and GPUs. That is a depreciation schedule measured in a few years, not the decades a datacentre shell earns over, and it recurs.
- More Personal Computing is shrinking, down 4% (−5% in constant currency). It is the one part of the business with no AI story attached, and nothing in the quarter suggests a floor.
- The RPO figure does most of the work in the bull case, and we cannot see inside it. Duration and customer concentration are undisclosed. An 84% increase built on a small number of very large multi-year AI commitments is a materially different asset from broad-based enterprise renewal.
Conclusion
Context matters here. Microsoft spent most of this cycle as the laggard of the Mag 7, and a beat this size, in the segment actually doing the work, lands very differently in a name the market had quietly written off as dull. Azure at 43% would have been well received whenever it arrived; arriving where expectations had already been lowered, it compounded. The Copilot seat additions completed the shift in sentiment.
The business is in good shape. The price is the problem. From the $350s to a shade under $500, the asymmetry that made this worth owning has gone, and the capex question remains open — $115.9bn a year, most of it in short-lived assets, has to earn a return nobody can yet verify. I would rather have that argument from a lower entry. The Street now has September at $90.6bn and $4.71, so another quarter like this one is the base case rather than the upside.
So the rating moves to Hold and I am staying put. I like the name and expect to own it again — nearer $400 the risk/reward comes back and I would look to put something on. That is not a bearish call. If the tape keeps rallying, Microsoft has a clean path back towards its highs, and missing the first leg of that beats paying up for the last. What would change it sooner is Azure itself: in the forties the spending is demand-led; drifting to the mid-thirties with capex still climbing makes it a bet.