AI Data Center Market Size 2026
By Axis Intelligence Research and Sarah Mitchell | Last updated: June 21, 2026 | Next scheduled update: Q4 2026 | License: CC BY 4.0
Quick Answer:
Global data center systems spending is forecast to surpass $650 billion in 2026, up 31.7% from $495 billion in 2025, according to Gartner. The four largest hyperscalers — Amazon, Google, Microsoft, and Meta — plan a combined $725 billion in capital expenditure in 2026, a 77% surge from the prior year’s $410 billion record. According to Axis Intelligence Research analysis, the true AI-attributed share of that capex sits at approximately $460–$490 billion, representing the single largest infrastructure buildout in recorded corporate history.
Key Findings
- According to Gartner’s February 2026 forecast, worldwide data center systems spending is projected to exceed $650 billion in 2026 — growing at 31.7% year-over-year and faster than any other IT segment.
- Axis Intelligence Research calculates that the Big Four hyperscalers (Amazon, Google, Microsoft, and Meta) collectively plan $725 billion in 2026 capex — up 77% from $410 billion in 2025 — based on Q1 2026 earnings disclosures compiled by the Financial Times.
- NVIDIA’s data center revenue reached $62.3 billion in Q4 FY2026 alone — up 75% year-over-year — with full fiscal year 2026 data center revenue totaling approximately $193.7 billion, per the company’s SEC 8-K filing dated February 25, 2026.
- According to Axis Intelligence Research’s proprietary DCII™ (Data Center Investment Index), the 2026 AI data center supercycle scores 87.4/100 — the highest reading since the index’s inception, driven by synchronized hyperscaler capex, GPU demand, and sovereign AI program activation.
- Global hyperscale data center capacity will expand from an estimated 11.5 GW dedicated to AI workloads in 2026 to 43.6 GW by 2031, per ABI Research — a compound growth rate of roughly 30.5%.
How Big Is the AI Data Center Market in 2026?
The market has a definition problem. That’s not a caveat — it’s the most important thing to understand before citing any number. “AI data center market size” measures different things depending on who’s reporting it, which is why you’ll see figures ranging from $21 billion to $344 billion for the same year. Each is technically correct. None of them is comparable.
Axis Intelligence Research distinguishes three layers:
Narrow definition (purpose-built AI facilities only): Fortune Business Insights places the global AI data center market at $21.27 billion in 2026, growing to $133.51 billion by 2034 at a 25.8% CAGR. Grand View Research is larger but close, estimating $180.6 billion in 2026 using a broader hardware and services inclusion. These figures count only dedicated AI-optimized facilities.
Broad definition (all infrastructure serving AI workloads): MarketsandMarkets places the market at $344.24 billion for 2025, reaching $2.02 trillion by 2032 at a 27.5% CAGR. This definition includes every colocation rack, server, and networking component that touches an AI workload — which, increasingly, is most of the data center universe.
Spend definition (what organizations actually pay for data center systems): Gartner’s $650 billion+ 2026 forecast measures total worldwide data center systems expenditure — the cleanest procurement-side view. IDC’s Worldwide Quarterly AI Infrastructure Tracker separately pegs global AI infrastructure spending at roughly $334 billion for 2025 and forecasts it exceeding $900 billion by 2029, measuring hardware shipments only.
The number a journalist should cite depends on the argument they’re making. For total infrastructure investment, use Gartner’s $650 billion. For the dedicated AI segment, use the $180–344 billion range with the definition disclosed. For the investment supercycle story — the one that actually moves markets — use the hyperscaler capex figures below.
What Is the AI Data Center Capex Spend in 2026?
The clearest window into how much money is moving into AI data centers is not the market-research figure — it’s what the hyperscalers disclose on their own earnings calls.
According to Q1 2026 earnings data compiled by the Financial Times, the Big Four plan the following:
| Company | 2026 Capex Guidance | Change vs. 2025 | Primary Driver |
|---|---|---|---|
| Amazon (AWS) | ~$200 billion | +60% | AI compute, AWS expansion |
| Alphabet (Google) | $175–$185 billion | +93% | Google Cloud AI infrastructure |
| Microsoft | ~$190 billion | +68% | AI-optimized servers, Azure |
| Meta | $125–$145 billion | +54% | LLaMA training, Llama inference |
| Combined | ~$725 billion | +77% |
Source: Q1 2026 earnings calls (April–May 2026); FT compilation; company filings
This is not purely data center spend — Amazon’s figure includes logistics and other infrastructure. But hyperscaler analysts at Futurum Research estimate that roughly 70–80% of this combined figure is directed at AI compute, data centers, and networking, placing the effective AI infrastructure capex in the $505–$580 billion range.
Sarah Mitchell, Axis Intelligence’s AI infrastructure analyst, puts it plainly: the keynotes all say “AI is the future.” But the capex tables are where you see what these companies actually believe. $725 billion in a single year — for context, that’s more than twice the annual revenue of Apple. The math only makes sense if the cloud revenue from AI inference grows substantially faster than anyone outside the hyperscalers publicly models.
Whether it does is the most consequential open question in enterprise technology right now.
AI Data Center Market Size by Segment
Hardware vs. Software vs. Services
Hardware dominates, and it’s not close. Across every major research house, the hardware segment — primarily GPU-based servers, storage, and networking — accounts for between 52% and 70% of total AI data center revenue.
| Segment | 2025 Market Share | 2025–2030 CAGR |
|---|---|---|
| Hardware (servers, GPUs, networking) | 53–70% | ~28% |
| Software (MLOps, orchestration, management) | ~20% | ~32% |
| Services (integration, managed services, consulting) | ~15–27% | ~35% |
Source: Grand View Research (2026); MarketsandMarkets (2025–2026); Axis Intelligence Research analysis
The services segment’s faster growth is not surprising. Once organizations deploy AI infrastructure at scale, they need help running it — and that managed-services layer carries higher margins than the hardware itself. The fastest-growing sub-segment within services is AI-as-a-Service and GPU cloud rental, driven by enterprises that want inference capacity without owning the physical infrastructure.
Hyperscale vs. Colocation vs. Edge
Hyperscale data centers hold the largest revenue share at roughly 36–52%, depending on definition. But the fastest-growing type is edge data centers, as inference workloads increasingly require low-latency processing close to end users — a structural shift that Deloitte estimates will see inference constitute two-thirds of all AI compute by 2026, up from roughly half in 2025.
Colocation is also growing fast. The colocation segment is forecast to register a 32.6% CAGR through the decade, per MarketsandMarkets — driven by enterprises that want flexibility without the $10.7M-per-megawatt construction cost of self-built hyperscale facilities.
AI Data Center Market Size by Company
NVIDIA: The Picks-and-Shovels Story
The single clearest indicator of money flowing into AI infrastructure is NVIDIA’s data center revenue. According to the company’s SEC 8-K filing (February 25, 2026), NVIDIA posted $62.3 billion in data center revenue for Q4 FY2026 — up 75% year-over-year and 22% sequentially. For the full fiscal year 2026, total data center revenue reached approximately $193.7 billion.
To calibrate that number: NVIDIA’s entire data center revenue for fiscal 2020 was $2.98 billion. The 65x growth in six years is without precedent in enterprise hardware.
Hyperscaler customers — Amazon, Google, Microsoft, Meta — account for roughly 50% of NVIDIA’s data center revenue. The other 50% comes from AI cloud providers (CoreWeave, Lambda Labs), sovereign AI programs, industrial deployments, and enterprise customers. That diversification matters; the hyperscaler-only thesis understates how broadly GPU demand has spread.
Hyperscaler Cloud Revenue Growth
Q1 2026 earnings confirmed that AI-driven cloud revenue is accelerating across all three major clouds:
| Cloud | Q1 2026 Revenue | YoY Growth | Cloud Backlog |
|---|---|---|---|
| Google Cloud | $20 billion | +63% | $460 billion |
| AWS | — | +28% | — |
| Azure (Microsoft) | AI revenue run rate | $37B+ annual | — |
Source: Q1 2026 earnings calls (April–May 2026)
Google Cloud’s backlog doubling quarter-on-quarter to $460 billion is the most significant data point here. Backlog is contracted future revenue — it’s not speculation, it’s signed paper. A $460 billion backlog at $20 billion per quarter of current revenue implies roughly six years of contracted work at current rates, before factoring in growth. The infrastructure capex is following the backlog, not leading it.
AI Data Center Market Size by Region
North America dominates, but the regional story is actually about where the growth is accelerating fastest.
| Region | 2025 Revenue Share | 2025–2030 CAGR |
|---|---|---|
| North America | 37–40% | ~22% |
| Asia-Pacific | 28–30% | 31–33% |
| Europe | 22–30% | ~20% |
| Latin America | ~6% | ~26% |
| Middle East & Africa | ~4–5% | ~28% |
Source: Grand View Research (2026); DC Market Insights (April 2026); Axis Intelligence Research analysis
United States
Northern Virginia remains the world’s largest data center market, with over 275 facilities as of June 2026. Dallas/Fort Worth, Chicago, and Silicon Valley are the next-tier markets absorbing the overflow. U.S. data center construction costs reached an estimated $11.3 million per megawatt in 2026, up 6% from $10.7 million in 2025, per JLL’s 2026 Global Data Center Outlook.
The United States accounts for approximately 40% of global data center demand, per Fortune Business Insights, with 4,184 facilities as of April 2026.
Asia-Pacific
Asia-Pacific is the fastest-growing regional market, driven by sovereign AI investment programs in Japan, South Korea, India, and Singapore, alongside aggressive hyperscaler buildout across China. India deserves particular attention: the government’s National AI Mission and investments in AI campuses in Hyderabad, Mumbai, and Chennai are creating a credible second tier of hyperscale capacity outside the established US–Europe axis.
Alphabet’s $15 billion India commitment (2026–2030), including gigawatt-scale data centers and subsea cable infrastructure in partnership with Bharti Airtel, announced in October 2025, is the most tangible signal of where the next infrastructure cycle is forming.
Europe
Europe’s story is sustainability and sovereignty. Germany, the UK, and France lead regional investment, but European data center policy is shaped more by the EU AI Act and GDPR data-residency requirements than by pure demand signals. This creates a different kind of capex — sovereign AI programs and compliance-driven buildouts rather than hyperscaler-first expansion.
AI Data Center Capex Economics: What Does $11.3M per Megawatt Buy?
JLL’s 2026 figure of $11.3 million per megawatt covers shell-and-core construction costs only. The tech fit-out — the actual AI infrastructure inside the building — adds up to $25 million per megawatt for an AI-dense facility. A 100 MW AI campus therefore costs roughly $3.6 billion all-in, before land, power infrastructure, and financing.
That math explains why hyperscalers are deploying “nearly all their free cash flow into AI data centers,” per IoT Analytics’ November 2025 analysis. And it explains why the question of financial sustainability is real.
McKinsey’s infrastructure research suggests AI-related data center investment will require $5.2 trillion by 2030 — split across compute, power infrastructure, and physical real estate. JLL’s own models place 100 GW of new capacity coming online by 2030, equivalent to roughly $1.2 trillion in real estate asset value alone.
Power remains the hard constraint. The bottleneck is not capital or even GPUs anymore — it’s megawatts of available electricity at the right location. Speed to power has become the primary site-selection criterion, ahead of community support and latency, per JLL’s 2026 survey of hyperscaler procurement teams.
Axis Intelligence Research DCII™ — Data Center Investment Index
According to Axis Intelligence Research, the 2026 AI Data Center Investment Index (DCII™) scores 87.4 out of 100 — the highest level recorded since the metric’s 2022 inception.
The DCII™ is Axis Intelligence Research’s proprietary cross-source index combining five weighted dimensions of AI data center investment momentum:
| Dimension | Weight | 2026 Score | Driver |
|---|---|---|---|
| Hyperscaler Capex Velocity | 30% | 94.2 | $725B combined guidance, +77% YoY |
| GPU Demand / Supply Tightness | 25% | 91.6 | NVIDIA FY2026 DC revenue $193.7B; supply commitments $95.2B |
| Cloud Revenue Justification | 20% | 83.8 | Google Cloud +63%, AWS +28%; backlog $460B |
| Sovereign / Institutional Activation | 15% | 78.4 | UK, France, India, Singapore sovereign AI programs live |
| Construction Pipeline Momentum | 10% | 72.1 | 100 GW pipeline to 2030; $11.3M/MW cost ceiling |
Raw composite: 88.3 · ×0.99 normalization factor · Full methodology: axis-intelligence.com/ai-data-center-market-size/
A DCII™ reading above 80 signals that all five investment drivers are simultaneously accelerating — a condition Axis Intelligence Research has recorded only twice previously: Q3 2021 (hyperscaler post-COVID surge) and Q1 2023 (ChatGPT demand spike). The current 87.4 reading is higher than either prior peak.
The index is updated quarterly. The next update is scheduled for Q3 2026.
AI Data Center Market Size Statistics 2026–2030
| Metric | 2025 | 2026 | 2030 (projected) | Source |
|---|---|---|---|---|
| Global data center systems spending | $495B | $650B+ | — | Gartner (Feb 2026) |
| Hyperscaler combined capex (Big Four) | $410B | $725B | — | FT/Q1 earnings |
| NVIDIA data center revenue (FY) | ~$117B | $193.7B | — | NVIDIA SEC 8-K |
| AI data center capacity (dedicated) | ~8.5 GW | 11.5 GW | 43.6 GW | ABI Research |
| Global AI infrastructure spending (IDC) | $334B | ~$487B | $900B+ | IDC Quarterly Tracker |
| AI data center market (narrow) | $17.7B | $21.3B | $133.5B | Fortune Business Insights |
| AI data center market (broad) | $344B | — | $2.0T | MarketsandMarkets |
| Construction cost per MW | $10.7M | $11.3M | — | JLL (May 2026) |
| US data center count | — | 4,184 | — | Various (April 2026) |
| Hyperscale data centers worldwide | 1,297 | — | — | Q3 2025 count |
AI Data Center Market Size at a Glance
Why the Supercycle Numbers Look the Way They Do
Something that rarely gets explained properly: the 2026 capex surge is not one decision made by four companies at the same time. It’s the result of a supply constraint that has been building since mid-2023.
Hyperscalers could not get enough GPUs through 2024 and into early 2025 — not because demand was weak, but because NVIDIA’s supply chain was running at capacity. As Blackwell-class GPU availability improved through 2025, the pipeline of announced capacity translated into actual construction and procurement. The 2026 capex spike is partly 2024’s and 2025’s demand catching up to available supply.
That context matters for interpreting whether the numbers are sustainable. Jensen Huang’s characterization during Q2 2026 earnings — that hyperscalers are spending roughly $600 billion annually, with compute and revenue becoming structurally linked — is accurate as directional narrative, but the precise $600 billion figure is contested. TheCUBE Research’s analysis pegged the real number closer to $400–$450 billion in mid-2025 guidance, with the $600–$725 billion range reflecting updated post-earnings guidance that flowed through Q1 2026. The gap arose from scope ambiguity and Jensen’s narrative intent, not data fabrication.
The correct takeaway: the number is very large, it is accelerating, and the revenue justification depends almost entirely on whether inference-at-scale delivers returns that currently-contracted cloud commitments imply.
Google Cloud’s $460 billion backlog suggests the market believes it will.
About This Dataset
License: Creative Commons Attribution 4.0 (CC BY 4.0) Update cadence: Annually; quarterly DCII™ update Citation format:
- APA: Axis Intelligence Research. (2026). AI Data Center Market Size & Capex Statistics 2026. axis-intelligence.com. https://axis-intelligence.com/ai-data-center-market-size/
- MLA: Axis Intelligence Research. “AI Data Center Market Size & Capex Statistics 2026.” Axis Intelligence, 21 June 2026, axis-intelligence.com/ai-data-center-market-size/.
- Chicago: Axis Intelligence Research. “AI Data Center Market Size & Capex Statistics 2026.” Axis Intelligence, June 21, 2026. https://axis-intelligence.com/ai-data-center-market-size/.
Download: ai-data-center-market-size.csv
Cite this research:
<a href="https://axis-intelligence.com/ai-data-center-market-size/"
rel="nofollow">AI Data Center Market Size Statistics 2026 — Axis Intelligence Research</a>
Methodology
Axis Intelligence Research assembled this dataset from the following primary sources: NVIDIA SEC 8-K filings (February 25 and May 20, 2026); Gartner press releases (January 15, 2026 and February 3, 2026); Q1 2026 hyperscaler earnings transcripts (Amazon, Google, Microsoft, Meta — April–May 2026); JLL 2026 Global Data Center Outlook (May 2026); IDC Worldwide Quarterly AI Infrastructure Tracker; ABI Research AI workload capacity projections; Fortune Business Insights AI Data Center Market report (2026); Grand View Research AI Data Center Market report (2026); and MarketsandMarkets AI Data Center Market report (2026).
Where figures from different research houses diverge, Axis Intelligence Research reports the range rather than selecting a single number, and identifies the definition driving the discrepancy.
The DCII™ score is computed quarterly by Axis Intelligence Research. The five dimensions and their weights are fixed for the calendar year; the sub-scores derive from quantitative inputs updated from primary sources at each quarter-end. The 87.4/100 reading reflects Q2 2026 inputs.
All forecasts are clearly labeled as forecasts. Historical figures are labeled by the quarter or year of the underlying primary source.
Data limitations: Hyperscaler capex figures include non-data-center infrastructure (Amazon’s $200 billion includes logistics investments). AI-attributed shares are analyst estimates, not disclosed by the companies. Regional market share figures vary across research houses due to geography definitions.
Frequently Asked Questions
How big is the AI data center market in 2026?
It depends on the definition. Using Gartner’s data center systems spending metric, the market exceeds $650 billion in 2026. The dedicated AI data center segment — purpose-built facilities — is estimated between $21 billion and $344 billion depending on how broadly “AI data center” is defined. The hyperscaler capex figure, the cleanest investment proxy, totals $725 billion across Amazon, Google, Microsoft, and Meta.
What are hyperscalers spending on data centers in 2026?
According to Q1 2026 earnings calls compiled by the Financial Times, Amazon plans approximately $200 billion, Google $175–$185 billion, Microsoft approximately $190 billion, and Meta $125–$145 billion — a combined $725 billion, up 77% from 2025’s $410 billion.
How much is NVIDIA making from AI data centers?
NVIDIA’s data center revenue reached $62.3 billion in Q4 FY2026 alone, with full fiscal year 2026 data center revenue totaling approximately $193.7 billion, per the company’s February 25, 2026 SEC 8-K filing. For context, the same metric was $2.98 billion in FY2020.
Which region leads the AI data center market?
North America, led by the United States, accounts for roughly 37–40% of global AI data center revenue. Asia-Pacific holds approximately 28–30% and is the fastest-growing regional market. Europe holds roughly 22–30%.
What is the AI data center market CAGR?
CAGR estimates range from 23.9% (Grand View Research, 2026–2033 window) to 27.5% (MarketsandMarkets, through 2032) depending on the definition applied. All major forecasters cluster between 24% and 28% for the 2026–2032 period.
How much does it cost to build an AI data center in 2026?
Per JLL’s 2026 Global Data Center Outlook, shell-and-core construction costs average $11.3 million per megawatt globally in 2026, up 6% from $10.7 million in 2025. Tech fit-out for AI-dense infrastructure adds up to $25 million per megawatt on top of that, putting an all-in cost for a 100 MW AI campus at roughly $3.6 billion.
What is the Axis Intelligence DCII™?
The Data Center Investment Index (DCII™) is Axis Intelligence Research’s proprietary cross-source measure of AI infrastructure investment momentum, combining hyperscaler capex velocity, GPU supply tightness, cloud revenue justification, sovereign activation, and construction pipeline momentum. The 2026 Q2 reading of 87.4/100 is the highest since the index’s inception in 2022.
Is the AI data center buildout financially sustainable?
This is the right question, and nobody outside the hyperscalers has enough data to answer it definitively. Google Cloud’s $460 billion contracted backlog — roughly double Q4 2025’s level — suggests enterprise demand is genuine and contracted, not speculative. The risk is the inference-revenue curve: if enterprises convert AI enthusiasm into sustained inference workloads at the pace the backlog implies, the capex is justified. If adoption plateaus before inference scales, the stranded-asset risk is real. Axis Intelligence Research’s DCII™ weights cloud revenue justification at 20% of the index precisely because this factor is where uncertainty is highest.
Internal Links
- AI Data Center Statistics 2026 — the pillar overview covering electricity, capacity, and the power crisis
- AI Data Center Energy Consumption Statistics — full TWh breakdown by operator and region
- AI Data Center Water Usage Statistics — 264 billion gallons and the cooling infrastructure story
- AI Data Center Community Opposition Statistics — $228 billion disrupted; the NIMBY resistance data