Hyperscaler growth is accelerating, so cloud capacity needs planning like a budget
Azure grew 48%, AWS 39.3% and Google Cloud 103% last quarter. For cloud capacity planning, treat AI-driven demand as a risk to your costs, not only a vendor success story.
What happened
In the last week of October 2026, the three largest cloud providers reported quarterly results. Cloud growth was the main story in each.
Microsoft published its first-quarter results for fiscal 2027 on 28 October. Azure revenue rose 48%, to $33,165 million from $22,384 million a year earlier. Microsoft Cloud revenue was $63.9 billion, up 30%. The release lists additions to property and equipment of $39,367 million for the quarter, against $19,394 million a year earlier, roughly double.
Alphabet reported on 28 October, and the Quartr summary of its third quarter is dated 29 October. It shows Google Cloud revenue of $30.709 billion, up 103% year on year. Total revenue backlog was $567.7 billion, of which $560.8 billion is Google Cloud. Capital spending on property and equipment in the quarter was $51.355 billion, with full-year 2026 guidance of $195 billion to $205 billion. Free cash flow for the quarter was negative $9.065 billion.
Amazon's figures, reported by Anadolu Agency on 29 October, show AWS net sales of $46 billion, up 39.3%, the fastest growth in nearly five years. AWS operating income was $17.8 billion, up 55% from $11.4 billion a year earlier. Amazon repeated that its full-year 2026 cash capital spending will be about $220 billion, mainly for AI infrastructure.
We have taken figures for Microsoft from its own release, and for Alphabet and Amazon from secondary summaries of their announcements. Anyone relying on a specific number should check the companies' filings.
Why it matters
Demand for AI capacity is shaping what suppliers build and sell. Spending on this scale, at companies whose cloud revenue is growing at these rates, tells buyers where supplier attention is going. That is good news for availability over time. It also means the products, regions and commercial terms you rely on are being shaped by the largest AI customers, not only by you.
Supplier results are now tied to one theme. The Alphabet summary shows a quarter with record cloud revenue and negative free cash flow in the same period. We draw no conclusion about any provider's health from that. The point for customers is narrower: your suppliers have committed to very large investment, and their incentives will favour workloads that fill it.
Cost exposure is moving to the workload level. With AI usage, a small change in a prompt, a model choice or a retry loop can change a bill. Overall growth rates say nothing about your own unit cost. Only measurement does.
Concentration is a business risk, not a technical one. All three growth stories sit on a few providers. If your critical systems and your AI workloads share one supplier, a commercial or capacity problem there becomes yours quickly.
Our view is that most organisations plan cloud capacity as a procurement exercise once a year. It should be a standing management process, with an owner and a forecast.
What we would do
- Forecast demand by workload. For each major application and AI workload, estimate growth over the next twelve months in terms the business recognises, such as users, transactions or documents processed, then translate that into resources.
- Measure unit cost. Track cost per transaction, per customer or per task, not just the monthly total. Set an alert when the unit cost moves.
- Separate baseline from experiment. Commit to spend only for the usage you can predict, and keep experimental and bursty work on flexible terms. Revisit the split every quarter.
- Find your scarce-resource dependencies. List the workloads that need specific accelerators or regions, and check what happens if capacity is limited or priced differently. Have a fallback for the important ones.
- Keep a credible alternative. For critical systems, know how you would move or run elsewhere, even if you never do. Test one workload's portability, not just document it.
- Review supplier terms with your advisers. Check price-change, notice and capacity clauses ahead of renewal, and take legal and commercial advice on what you negotiate.
Nothing in these results tells you to move cloud. They tell you that planning is worth more than it was last year.
Questions we are asked
Do these results mean cloud prices will rise?
The results do not say so, and we would not predict it. They show very fast demand growth and very large investment in capacity. What buyers can control is exposure. Understand which of your workloads depend on scarce resources such as GPUs, set budgets and alerts at the workload level, and keep the contractual and technical option to move or scale back.
Should we commit to a long spend agreement now?
Only if your usage is predictable enough to justify it. Commitments can lower unit cost but remove flexibility, and AI workloads change quickly. A sensible approach is to commit for the stable baseline, leave experimental and bursty workloads on flexible terms, and review the split each quarter against measured usage. Take commercial and legal advice before signing.
Why does supplier capital spending matter to a customer?
It signals how much capacity suppliers are building and how dependent their results are on AI demand. For a customer it affects availability, negotiating leverage and long-term supplier risk. It is one input among several, alongside your own forecast, your resilience needs and the terms you can get, and it should inform planning rather than replace it.