Cloud and AI Platforms Industry Explained: The U.S. View
The three largest U.S. cloud providers, Amazon Web Services, Microsoft Azure and Google Cloud, are where AI demand turns into revenue. They spend on chips, data centers and power, then rent that capacity to AI labs, sell AI models and tools to businesses, and keep running ordinary cloud workloads. In the second quarter of 2026 all three accelerated. AWS revenue rose 37% to $42.2 billion, its fastest growth in 18 quarters, with operating income of $16.6 billion. Azure grew 43% in constant currency and passed $100 billion in annual revenue. Google Cloud revenue jumped 82% to $24.8 billion. Contracted backlog is huge: Microsoft's commercial backlog was $678 billion and Google Cloud's $514 billion. The spending is even larger. AWS raised its 2026 capital spending plan to about $220 billion, Alphabet to $195–205 billion, and Microsoft guided $255–260 billion for its fiscal 2027. Demand is running ahead of capacity: AWS said its 2027 data center capacity was already largely reserved. The question for investors is whether revenue keeps growing fast enough to earn a return on these budgets.
- Growth
- Q2 2026: AWS +37%, Azure +43%, Google Cloud +82%
- Backlog
- Microsoft commercial backlog $678B; Google Cloud $514B
- Capex
- AWS raised 2026 capex to about $220B; Alphabet to $195–205B
How the chain works
The three U.S. clouds, Q2 2026
| AWS | Microsoft Azure | Google Cloud | |
|---|---|---|---|
| Revenue growth | 37%, to $42.2B | 43% in constant currency | 82%, to $24.8B |
| Profitability | Operating income $16.6B, 39.4% margin | Azure passed $100B annual revenue | Operating margin 35.6%, from 20.7% |
| Backlog | Not stated in this source | Commercial backlog $678B | Backlog $514B, from $460B in Q1 |
| Capex plan | About $220B for 2026 | $255–260B for fiscal 2027 | $195–205B for 2026 |
Where AI money becomes revenue
AI labs need huge amounts of computing to train and run models, and most rent it from cloud providers. Businesses buy AI through the same clouds, either as raw computing or as model access and tools. That makes cloud revenue the first place AI demand shows up in profits, ahead of most AI applications.
In the second quarter of 2026, according to UpperEdge's summary of the companies' results, AWS revenue rose 37% to $42.2 billion, the fastest in 18 quarters and above the 31% analysts had modeled. Azure grew 43% in constant currency, up from 40% the prior quarter, and Microsoft guided about 45% for the September quarter. Google Cloud revenue rose 82% to $24.8 billion, well above the 64% analysts expected.
Backlog: demand already signed
Long-term contracts give clouds visibility. Microsoft's commercial remaining performance obligation, its contracted backlog, reached $678 billion at the end of its fiscal year on June 30, 2026, and Google Cloud's backlog rose to $514 billion from $460 billion in the first quarter.
Capacity is the constraint. AWS said 2027 data center capacity is already largely reserved, with some committed as far out as 2028, and UpperEdge describes Azure's capacity constraints as persistent.
Capex: the size of the bet
To meet that demand, budgets keep rising. AWS raised its 2026 capital spending guidance to about $220 billion from $200 billion. Alphabet raised its 2026 range to $195–205 billion from $180–190 billion. Microsoft spent $41 billion in its fiscal fourth quarter and guided $255–260 billion for fiscal 2027.
Margins are holding up for now: AWS earned a 39.4% operating margin and Google Cloud 35.6%, up from 20.7% a year earlier. But depreciation on these investments will rise for years, so revenue has to keep growing quickly to keep returns attractive.
How the U.S. view differs from the China view
The Chinese version of this page looks at Chinese clouds such as Alibaba Cloud, which build AI platforms with domestic chips and limited access to the most advanced U.S. accelerators, and compete on price in a market growing more slowly.
From the U.S. side the three largest clouds are American, and their combined capital spending sets demand for chips, memory, networking and power worldwide. Their results are the best single read on whether the AI build-out is paying off.
2026: growth accelerates with spending
In 2026 the three U.S. clouds raised capital spending again, to a combined level above $600 billion a year.
Revenue followed: in the second quarter, AWS grew 37%, Azure 43% and Google Cloud 82%, and backlogs reached hundreds of billions of dollars.
AWS said its 2027 capacity was already largely reserved. For now, demand is outrunning supply; the risk is that spending keeps rising after demand slows.
- Cloud revenue is where AI demand first becomes profit.
- Backlog shows how much demand is already contracted.
- Capex commitments raise the bar for future revenue growth.
Common misreadings
MythAI is not making money for anyone yet.
RealityGoogle Cloud's operating margin rose to 35.6% as revenue grew 82% in Q2 2026.
MythCloud growth is slowing as the market matures.
RealityAWS grew 37% in Q2 2026, its fastest in 18 quarters.
MythClouds can add capacity whenever demand rises.
RealityAWS said its 2027 data center capacity was already largely reserved.
MythCapex budgets are falling after the first wave.
RealityAWS raised its 2026 capex plan to about $220B from $200B.
Demand, capacity and returns: scenarios
| Scenario | Main effect | How it passes through |
|---|---|---|
| AI demand keeps outrunning capacity | Cloud revenue | Growth stays high, capex rises again |
| AI lab customers cut spending | Backlog quality | Slower growth, idle capacity |
| Power and chip supply loosen | Capacity | Faster revenue conversion |
| Depreciation rises faster than revenue | Margins | Profitability falls |
| Custom chips cut costs | Cloud margins | Better economics per workload |
| Price competition for AI models | Revenue per customer | Lower prices, more volume |
Supply and demand by segment
| Segment | Supply rigidity | Concentration | New demand | Policy role | What to watch |
|---|---|---|---|---|---|
| Data center capacity | Very high (power, sites) | High | Yes | High (permits, power) | Capacity reserved |
| AI compute rental | High (chips) | High (three clouds) | Yes | Medium (export controls) | Revenue growth |
| Enterprise AI services | Low | High | Yes | Medium | Adoption, pricing |
| Traditional cloud | Low | High | Low | Low | Steady growth |
Supply rigidity: how long new capacity takes and whether money can speed it up. Concentration: whether output sits in a few countries or companies. The more rigid and concentrated, the more prone to shortages.
How to track it
| Indicator | What it tells you | Where to find it |
|---|---|---|
| Cloud revenue growth and margins | AI demand turning into revenue | Amazon, Microsoft, Alphabet quarterly results, free |
| Backlog (remaining performance obligations) | Contracted future demand | Company 10-Q filings, free |
| Capex guidance | Size of the build-out | Earnings calls, free |
| AI compute | What the capex buys | U.S. view: AI compute |
| AI applications | Who uses the capacity | U.S. view: AI applications |
Sources
Last updated 2026-10-18.