CSC: North America's Top 5 Cloud Giants Expand Computing Power in Q3 FY26, Three See Free Cash Flow Turn Negative
nashnova research
The five largest North American cloud providers spent a combined $560 billion in capex through the first three quarters of 2026, nearly doubling year-on-year — but Amazon, Google and Meta have all tipped into negative free cash flow, pushing their legacy businesses to the limit.
How much are they spending — and how fast is it growing?
Microsoft, Amazon, Google, Meta and Oracle spent a combined $560 billion in capex through 26Q1–Q3, up 98% year-on-year. Full-year 2025 totaled roughly $412 billion, up 75%.
CICC Research estimates 2027 capex at $1.12 trillion in the base case (+38%) and $1.38 trillion in the bull case (+70%).
This means → compute investment has shifted from "accelerating" to "doubling," with the trillion-dollar threshold likely breached in 2027.
Can cloud revenue keep pace with spending?
In 26Q3, Google Cloud revenue grew 103% year-on-year, Azure 48%, and AWS 39% — driven by both new capacity coming online and better utilization of existing resources.
However, Google's and Microsoft's order backlogs are growing more slowly, and large multi-year contracts are concentrated among a handful of leading model developers.
This means → revenue growth hinges on a few big customers. If those model builders cut usage, top-line growth could decelerate quickly.
Why are purchase commitments still surging?
In 26Q3, Google, Amazon and Meta reported purchase commitments of $924 billion, $254 billion and $455 billion respectively — up roughly 887%, 249% and 460% year-on-year. Combined: about $1.63 trillion.
In plain terms = purchase commitments — contracts signed but not yet fully paid — are "bills" that must be honored over the coming years, far exceeding current-period spending.
This reflects the big cloud firms' willingness to keep raising their bets on long-term compute demand, not pulling back.
What does negative free cash flow actually mean?
As infrastructure spending keeps rising, Amazon, Google and Meta have all seen free-cash-flow margins turn negative. As of 26Q3, only Microsoft still generates positive free cash flow.
This means → the other three are spending more on AI than their operations produce in cash. The gap is covered by borrowing or drawing down reserves.
In plain terms = their legacy businesses — search ads, e-commerce, social ads — are the "blood bank" funding the AI buildout. The blood bank is now being drained faster than it can replenish.
Whose legacy business is holding up best?
Google faces dilution from AI Mode — a search feature that gives AI-generated answers directly — which monetizes at lower rates than traditional search ads.
Amazon needs to see whether ad revenue growth can recover once the distortion from cross-quarter promotional shifts fades.
Microsoft's M365 commercial cloud revenue grew 18% with seat growth of 5%; premium tiers and AI monetization are lifting per-customer revenue, and the stickiness of enterprise productivity tools provides relative resilience. Meta's social graph and content-consumption habits are hard for general-purpose models to displace, and AI-driven recommendation and ad optimization also support monetization.
What should investors watch next?
Custom chip sales: whether Google's TPU — tensor processing unit, Google's in-house AI chip — and Amazon's Trainium can win external customers; also, cost-reduction progress on Microsoft's Maia and Meta's MTIA chips used internally.
AI gateway monetization: whether AI Search, Copilot, Muse and similar products can convert gateway advantages into incremental revenue; and the demand scale and commercial potential of Agents — AI programs that execute tasks autonomously.
The sustainability of compute demand through 2027 and the pace of supply delivery will be the critical window for judging whether this trillion-dollar capex cycle pays off.
市场有风险,内容仅供研究参考,不构成投资建议。
