$280 Billion in Two Years
The five largest cloud and AI companies — Microsoft, Amazon (AWS), Google (Alphabet), Meta, and Oracle — have committed approximately $280 billion in combined capital expenditure for 2025-2026. The overwhelming majority is directed at data centre infrastructure: land, buildings, power, cooling, networking, and GPU clusters.
To contextualise this figure: $280 billion exceeds the GDP of 150+ countries. It is roughly 3x the annual capital budget of the entire US Department of Defence. It represents the largest concentrated infrastructure build in commercial history.
Company-by-Company Breakdown
### Microsoft — $80+ billion (2025-2026)
**Reported CapEx:** $80 billion for fiscal year 2025 alone (ending June 2025). Projected to exceed $85 billion for FY2026.
**Where it is going:** - Azure cloud capacity expansion (40+ new regions announced or under development) - NVIDIA GB200 and custom Maia AI accelerator deployment - Stargate partnership with OpenAI ($100 billion joint venture for dedicated AI infrastructure) - Nuclear energy investment (Three Mile Island restart via Constellation Energy)
**Key data centre investments:** - $13.5 billion in Wisconsin (Mount Pleasant campus) - $10 billion in Indiana - $7.5 billion in Sweden - $5.5 billion in Japan - Aligned Data Centers acquisition ($40 billion consortium with NVIDIA and xAI)
**Strategic posture:** Microsoft is spending more aggressively than any competitor, reflecting the belief that Azure + OpenAI capture disproportionate AI workload share. CEO Satya Nadella has publicly stated that "CapEx is the new moat."
### Amazon (AWS) — $75+ billion (2025)
**Reported CapEx:** $75 billion guided for 2025 (AWS-specific, excluding Amazon retail and logistics).
**Where it is going:** - AWS region expansion (33 regions, 105 availability zones) - Custom silicon: Graviton (CPU), Trainium (AI training), Inferentia (AI inference) - NVIDIA GPU clusters for third-party AI workloads - Nuclear energy procurement (960 MW from Talen Energy's Susquehanna plant)
**Key data centre investments:** - $35 billion in Virginia (multi-campus expansion) - $11 billion in Indiana - $10 billion in Mississippi - $9 billion in Saudi Arabia (new region) - Significant expansion in Ohio, Oregon, and Ireland
**Strategic posture:** AWS maintains the largest global data centre footprint and is investing to defend market share against Azure's AI-driven growth. Trainium custom silicon is a bet on reducing dependency on NVIDIA and improving margins.
### Google (Alphabet) — $60+ billion (2025)
**Reported CapEx:** $60 billion guided for 2025, up from $50 billion in 2024.
**Where it is going:** - Google Cloud Platform expansion - Gemini AI model training infrastructure (TPU v5p and v6) - Custom TPU deployment (sixth generation in development) - Fibre and submarine cable investment
**Key data centre investments:** - $10 billion in Indiana - $5 billion in South Carolina - $3 billion in Malaysia - $2 billion in Thailand - Kairos Power SMR partnership for nuclear-powered data centres (projected 2030+)
**Strategic posture:** Google differentiates on custom silicon (TPUs) and sustainability. The company operates the only carbon-free energy programme that matches consumption on an hourly (not annual) basis, campus by campus.
### Meta — $60+ billion (2025)
**Reported CapEx:** $60-65 billion guided for 2025, the most dramatic increase of any hyperscaler (up from $37 billion in 2024).
**Where it is going:** - Llama AI model training (largest open-source AI programme) - 2 GW+ AI training cluster (reportedly the largest single AI installation ever planned) - NVIDIA H100/H200/GB200 deployment at massive scale - New campus development across US and international markets
**Key data centre investments:** - $10 billion in Louisiana (multiple campuses) - $10 billion in Richland Parish, Louisiana (single campus) - Multi-billion investments in Ohio, Indiana, and international markets
**Strategic posture:** Meta is betting that AI-generated content, AI assistants, and AI-powered advertising will drive revenue growth that justifies the unprecedented CapEx. The company has no cloud infrastructure business — all spending serves Meta's own products.
### Oracle — $20+ billion (2025-2026)
**Reported CapEx:** Approximately $20 billion across 2025-2026, representing a dramatic acceleration for a company historically characterised as CapEx-light.
**Where it is going:** - Oracle Cloud Infrastructure (OCI) expansion - GPU Supercluster deployments (largest single OCI clusters exceed 32,000 GPUs) - 160+ planned cloud regions globally - Partnership with NVIDIA, xAI, and OpenAI for dedicated AI infrastructure
**Key data centre investments:** - 1 GW+ campus reportedly planned adjacent to nuclear generation - Multiple new OCI regions across US, Europe, and Middle East - Joint venture partnerships for build-to-suit campuses
**Strategic posture:** Oracle is the fastest-growing cloud infrastructure provider by percentage (revenue up 50%+ YoY). OCI's GPU supercluster architecture and competitive pricing are winning AI training workloads from AWS and Azure.
Market Impact
**Power demand:** Combined hyperscaler power demand is adding 5-10 GW annually to the US data centre fleet. This is equivalent to the residential electricity consumption of 5-10 million homes per year.
**Construction and labour:** The construction pipeline has created a nationwide shortage of electrical contractors, mechanical engineers, and data centre commissioning specialists. Labour costs have increased 15-25% in primary markets since 2023.
**Land prices:** Hyperscaler land acquisitions have driven prices from $50-100K/acre to $200-500K/acre in primary markets, and created secondary market interest in locations previously considered too remote.
**Supply chain:** Lead times for critical electrical equipment (transformers, switchgear, generators) have extended to 18-36 months, up from 6-12 months pre-2022. Hyperscalers are placing orders 3-4 years ahead of need.
**Implications for operators and investors:** The CapEx wave creates demand for colocation, powered shell, and build-to-suit capacity across all major markets. Operators with secured power and permitted sites are in the strongest negotiating position in industry history.
Track hyperscaler activity across markets on our market pages and provider directory. For strategic advisory on hyperscaler trends, contact our team.