Cloud hyperscalers — Amazon (AWS), Microsoft (Azure), Alphabet (Google Cloud), Meta, and Oracle (OCI) — own and operate the largest data center fleets on Earth. Together they spent $384B on capital expenditure in their most recent fiscal years and have guided $740–770B for 2026. The three pure cloud providers (AWS, Azure, GCP) generated $92B of combined cloud revenue in Q1 2026, growing 28–63% year-over-year. Meta does not sell cloud services; it consumes its own infrastructure to serve 3.56 billion daily users and train its Llama AI models. Oracle's cloud infrastructure (OCI) is growing 55–68% year-over-year off a smaller base, with an RPO backlog that ballooned to $553B.
Combined 2026 capex guidance of ~$750B is roughly double the ~$384B spent in 2025. Nearly all incremental spend targets AI: GPU clusters, liquid-cooled data centers, and the power to run them. Cloud GPU instance rental prices have been rising, not falling. All three pure-cloud providers have stated they are capacity-constrained.
AWS, Alphabet, Meta Q1 2026 earnings releases (Apr 2026); Microsoft FY26 Q1 earnings (Oct 2025); Oracle Q3 FY2026 earnings (Mar 2026); capex guidance from Platformonomics Q1 2026 scoreboard.
The product is rented compute, storage, and AI infrastructure. Customers pay to use servers, GPUs, databases, and AI services housed in data centers they do not own. Revenue comes in two forms:
Meta does not sell cloud services externally. Its infrastructure serves its advertising platform (3.56B daily users) and trains/runs its Llama AI models. Revenue is entirely from advertising ($196B in 2025). At $72B of capex in 2025 and $125–145B guided for 2026, it builds at hyperscaler scale with data centers as a cost center, not a revenue segment.
AWS earned $14.2B operating income on $37.6B revenue in Q1 2026 (38% margin). Google Cloud earned $6.6B on $20.0B (33% margin), up from 18% a year earlier. Microsoft's Intelligent Cloud segment operating income grew 27% YoY. Oracle carries $100B+ of debt to fund its buildout while OCI grows 55–68%. Meta's infrastructure spend is absorbed into its 53% company-wide operating margin.
AWS, Alphabet Q1 2026 earnings; Microsoft FY26 Q1 earnings; Oracle Q1–Q3 FY2026 earnings; Meta Q1 2026 and FY2025 earnings.
| Company | Backlog metric | Amount | Date | YoY change |
|---|---|---|---|---|
| Oracle | RPO | $553B | Feb 2026 | +325% |
| Google Cloud | Backlog | $460B+ | Mar 2026 | ~2× QoQ |
| AWS | Not separately disclosed | — | — | — |
| Microsoft | Not separately disclosed | — | — | — |
| Meta | N/A (internal consumption) | — | — | — |
contracted Oracle's $553B RPO is backed by "four multi-billion-dollar contracts with three different customers" (CEO Safra Catz). Google Cloud's backlog "nearly doubled quarter on quarter to over $460 billion" (CEO Sundar Pichai). AWS and Microsoft do not break out cloud-specific RPO. Microsoft's CFO has stated the company is "capacity-constrained, not demand-constrained."
Oracle Q3 FY2026 and Q1 FY2026 earnings; Alphabet Q1 2026 earnings call; Microsoft FY26 Q1 earnings; Synergy Research via CRN (Q1 2026 cloud market); Holori cloud market analysis 2026.
Synergy Research (Q3 2025 hyperscale count); Data Center Knowledge hyperscaler 2026 survey; datacenters.com 2026 project tracker.
| Measure | Demand signal | Supply signal |
|---|---|---|
| Cloud revenue growth (Q1 2026) | AWS +28%, GCP +63%, Azure +40% | All three report capacity constraints |
| Contracted backlogs | Oracle $553B, Google $460B+ | Multi-year delivery timelines (2027–2031) |
| Capex ramp (2025 → 2026 guidance) | $740–770B guided spend | Power, transformers, labor limit deployment speed |
| GPU instance availability | Waitlists across providers | NVIDIA allocation + data center energization |
| GPU rental pricing direction | Prices rising, not falling | H100 rental: $1.49–$6.98/hr across 15+ providers |
| Hyperscale DC pipeline | 770 facilities in planning/construction | 36+ projects ($162B) blocked/delayed mid-2025 |
When could the gap close? The bottleneck is shifting from chips to powered, energized data center capacity. Grid interconnection timelines of 3–5 years put meaningful supply relief at 2028+. Inference demand scales with users and model capability, so the demand side is open-ended.
CRN Q1 2026 cloud earnings face-off; Oracle Q3 FY2026 earnings; Alphabet Q1 2026 earnings call; intuitionlabs.ai H100 pricing comparison; Data Center Knowledge 2026.
| Ticker | Cloud segment | Cloud rev (latest Q, ann.) | Cloud growth (YoY) | Cloud op. margin | 2026 capex guidance | Market cap |
|---|---|---|---|---|---|---|
| AMZN | AWS | $150B | 28% | 38% | ~$200B est. | $2.29T |
| MSFT | Intelligent Cloud (Azure) | $139B | 28% (Azure 40%) | ~39% est. | ~$190B est. | $3.00T |
| GOOGL | Google Cloud (GCP) | $80B | 63% | 33% | $175–185B | $3.74T |
| META | Internal (no cloud sales) | N/A | N/A | N/A | $125–145B | $1.65T |
| ORCL | OCI + Cloud Apps | $36B | 44% (OCI 55–68%) | ~29% est. | ~$50B est. | $0.70T |
Cloud rev annualized = latest quarter × 4. AMZN: Q1 2026 $37.6B × 4. MSFT: FY26 Q1 $34.7B × 4. GOOGL: Q1 2026 $20.0B × 4. ORCL: Q3 FY26 $8.9B × 4. Market caps from stockanalysis.com / westmountfundamentals.com (late May 2026). Op margins from segment filings.
| Ticker | Price | Market cap | P/E (trailing) | TTM net income | Cloud as % of revenue | 2026E capex / mkt cap |
|---|---|---|---|---|---|---|
| AMZN | $213 | $2.29T | ~30× | $77.7B (FY2025) | ~21% (AWS) | 8.7% |
| MSFT | $404 | $3.00T | ~25× | $101B (FY2025) | ~51% (IC) | 6.3% |
| GOOGL | $309 | $3.74T | ~29× | ~$130B (CY2025) est. | ~15% (GCP) | 4.8% |
| META | $652 | $1.65T | ~28× | $60.5B (CY2025) | 0% (internal) | 8.2% |
| ORCL | $245 | $0.70T | ~57× | $12.4B (FY2025) | ~43% (cloud) | 7.1% |
Dilution: All five have modest dilution (1–2% annually from stock compensation). Oracle raised $30B in bonds and mandatory convertible preferred stock in FY2026 to fund its buildout; total debt exceeds $100B.
Revenue mix: AMZN and GOOGL are primarily advertising/e-commerce businesses where cloud is 21% and 15% of total revenue respectively. MSFT has the highest cloud concentration (51%). META has zero cloud revenue — it is an advertising business spending at hyperscaler scale. ORCL has the most aggressive revenue targets ($67B → $144B over five years) and the most debt.
Capex intensity: All five are spending 5–9% of their market cap on capex in 2026, roughly 3–5× the capex intensity of a typical S&P 500 company. est.
Prices and market caps from stockanalysis.com, westmountfundamentals.com (late May 2026). P/E from trailing net income per filings. Capex guidance per company earnings releases and Platformonomics. Amazon FY2025 net income includes $16.8B Anthropic gains.
Confidence notes: Company financials are from primary earnings releases and SEC filings. Cloud backlog figures (Oracle RPO, Google Cloud backlog) are directly from earnings releases. Market share percentages are from Synergy Research. GPU pricing trends and GPU-as-a-Service growth rates are from industry commentary and are directional, not verified from primary sources — marked est. throughout. Capex guidance is from company earnings calls and may be revised. Fiscal year-ends: Microsoft June (FY2025 = Jul 2024–Jun 2025); Oracle May (FY2025 = Jun 2024–May 2025); others calendar year. Meta FY2025 net income of $60.5B reflects a one-time tax impact from the One Big Beautiful Bill Act (effective tax rate 30% vs. normalized ~13%). Alphabet CY2025 net income of ~$130B is annualized from reported quarters. est.