AI Data Center Market Size, Share & Forecast to 2035
Market at a glance
The AI Data Center market was worth USD 390.19 billion in 2025 and is forecast to reach USD 1,612.86 billion by 2035, growing at a 15.25% CAGR. North America accounts for 54.8% of the market.
Source: DC Market Insights analysis
Key findings
Hyperscale Capital Spending Reaches USD 291.97 Billion at Three Companies
Rack-Scale Systems Push Densities Far Past the Industry Average
Power Demand More Than Doubles to 945 TWh by 2030
GPU Clouds and Neoclouds Grow 18.45% a Year
What is inside the report
- 01Introduction
- 02Research Methodology
- 03Executive Summary
- 04Market Dynamics
- 05Pricing Analysis
- 06Standards and Regulation
- 07Supply Chain Analysis
- 08AI Data Center Market Size and Forecast
- 09Market by Data Center Type
- 10Market by Component
- 23chapters
- 2020-2024Historical period
- 2026-2035Forecast period
- 17Companies profiled
Regional insights
- North America54.8%
- Asia Pacific25.6%
- Europe11.9%
- Middle East and Africa4.6%
- Latin America3.1%
Segmentation
Companies profiled
- NVIDIA
- AMD
- Broadcom
- Dell Technologies
- Supermicro
- Hewlett Packard Enterprise (HPE)
- Cisco
- Vertiv
- Schneider Electric
- Microsoft
- Amazon Web Services (AWS)
- Meta
- Oracle
- CoreWeave
- Equinix
- Digital Realty
Recent developments
OpenAI, SoftBank, Oracle and MGX announced Stargate, which intends to invest USD 500 billion over four years in new AI infrastructure, starting with USD 100 billion (Source: OpenAI announcement).
OpenAI and NVIDIA announced a letter of intent to deploy at least 10 GW of NVIDIA systems, with NVIDIA intending to invest up to USD 100 billion in OpenAI (Source: NVIDIA press release).
AMD and OpenAI announced an agreement for OpenAI to deploy 6 GW of AMD GPUs over multiple generations (Source: AMD filing with the US Securities and Exchange Commission).
Aligned Data Centers, with 5 GW of active and planned capacity, agreed to be sold to a consortium including BlackRock's GIP and MGX in a deal worth USD 40 billion (Source: Data Center Dynamics news report).
CoreWeave signed an expanded long-term agreement with Meta worth about USD 21 billion through December 2032 (Source: CoreWeave filing with the US Securities and Exchange Commission).
NVIDIA reported Data Center revenue of USD 89.0 billion for the quarter that ended in July 2026, up 117% from a year earlier (Source: NVIDIA financial results release).
Full analysis
The AI Data Center Market covers the spending needed to build and equip data center capacity for artificial intelligence (AI) workloads: accelerated servers, networking and storage, the power, cooling and building shell of AI halls, and the software and services that run them. DC Market Insights values the global AI Data Center Market at USD 390.19 billion in 2025 and forecasts USD 1,612.86 billion by 2035, a CAGR of 15.25%. The 2025 value is built from deployed capacity: about 9.04 gigawatts (GW) of AI IT load put into service at roughly USD 31.62 million of IT hardware per megawatt (MW) gives USD 285.84 billion, 4.20 GW of newly completed AI halls at USD 13.08 million of power, cooling and shell per MW adds USD 54.94 billion, and software and services add USD 49.41 billion. Inference is the fastest-growing workload segment at 20.56% a year because every trained model is then served to users around the clock, and the Middle East and Africa is the fastest-growing region at 22.17% a year as sovereign AI programs in Saudi Arabia and the United Arab Emirates build gigawatt-scale campuses.
Executive Summary
The Global AI Data Center Market size was valued at USD 19.75 billion in 2020, reached USD 390.19 billion in 2025, and is anticipated to reach USD 1,612.86 billion by 2035, at a CAGR of 15.25% during the forecast period.
| Report attribute | Details |
|---|---|
| Historical Period | 2020-2024 |
| Base Year | 2025 |
| Forecast Period | 2026-2035 |
| AI Data Center Market Size 2025 | USD 390.19 Billion |
| AI Data Center Market, CAGR | 15.25% |
| AI Data Center Market Size 2035 | USD 1,612.86 Billion |
The AI Data Center Market grew at 81.60% a year between 2020 and 2025, and 2025 alone added 84.2% over 2024 as generative AI moved into large-scale production. DC Market Insights expects another 54.3% jump in 2026, to USD 602.05 billion, as hyperscale capital budgets rise again. Growth then averages 15.25% a year to 2035 because IT hardware spend per MW falls from USD 31.62 million to USD 22.84 million while deployed AI capacity grows more than fivefold.
The AI compute and data landscape is concentrated in a small group of buyers. Alphabet, Amazon and Meta alone reported USD 291.97 billion of capital expenditure in 2025, and NVIDIA’s Data Center revenue reached USD 193.7 billion in its fiscal year that ended in January 2026, up 68% (Source: NVIDIA results filed with the US Securities and Exchange Commission).

Market Scope: What Does the AI Data Center Market Cover?
The AI Data Center Market covers the first-sale value, in current US dollars, of everything bought to put AI compute capacity into service inside data centers, sized by DC Market Insights at USD 390.19 billion in 2025.
Included. Accelerated servers built on graphics processing units (GPUs) and custom AI accelerators (application-specific integrated circuits, or ASICs); AI cluster networking and storage; the power, cooling and shell of new or converted AI halls; cluster software and orchestration; and design, deployment and managed services. Accelerators placed in existing halls, including refresh cycles, count as deployed AI IT load.
Excluded. Land, utility-side grid assets, general-purpose servers without accelerators, devices outside data center buildings, electricity bills, and cloud or colocation rental revenue, which would count the same equipment twice. Smart or autonomous data center software counts only when bought for an AI hall.
How the segments fit together. The component split (hardware, software and orchestration, services) and the hardware-layer split re-cut the same USD 390.19 billion; hardware is USD 340.78 billion, of which AI servers and accelerators are USD 237.82 billion.
Market Drivers: What Drives the AI Data Center Market?
Three drivers explain most of the growth in the AI Data Center Market: USD 291.97 billion of 2025 capital spending at three hyperscalers, gigawatt-scale compute contracts, and the shift from training to inference.
Hyperscale Capital Spending Reaches USD 291.97 Billion at Three Companies
Alphabet’s purchases of property and equipment reached USD 91.45 billion in 2025, Amazon’s cash capital expenditure reached USD 128.3 billion and Meta’s capital expenditure, including finance leases, reached USD 72.22 billion. The 2026 plans are larger again: Meta narrowed its 2026 outlook to USD 130 to 145 billion in July 2026, Amazon guided about USD 200 billion in February 2026, and Microsoft expects about USD 175 billion for calendar 2026 after moving some leases out of capital expenditure. DC Market Insights estimates that hyperscale operators and other cloud service providers bought USD 216.17 billion of AI data center capacity in 2025, 55.4% of the market, and that their spend grows 13.77% a year, slower than the market, as neoclouds, colocation and sovereign buyers take a larger share.
Gigawatt Contracts Lock In AI Data Center Demand Through 2029
AI buyers now contract compute in gigawatts, not racks. In September 2025, OpenAI and NVIDIA signed a letter of intent to deploy at least 10 GW of NVIDIA systems, with NVIDIA intending to invest up to USD 100 billion in OpenAI as the systems are deployed. OpenAI then agreed to deploy 6 GW of AMD GPUs and announced a collaboration with Broadcom for 10 GW of custom accelerators. Anthropic expects its expanded use of Google tensor processing units to bring well over a gigawatt of capacity online in 2026. Together these commitments add up to at least 27 GW of AI capacity. At USD 31.62 million of IT hardware per MW, each gigawatt represents about USD 31.62 billion of equipment, counted when deployed.
Inference Overtakes Training as the Larger AI Workload
DC Market Insights estimates that training clusters took 58.3% of 2025 spend, USD 227.48 billion, while inference took USD 162.71 billion. Inference grows at 20.56% a year to USD 1,055.87 billion by 2035, when it holds 65.5% of the market, because every model in production must answer users continuously. Digital Realty described a 200 MW lease in Charlotte in April 2026 as AI inference-oriented.
Market Trends: How Is the AI Compute and Data Landscape Changing?
The AI compute and data landscape is moving to rack-scale systems, liquid cooling as the default and campuses measured in gigawatts, and DC Market Insights expects direct-to-chip liquid cooling to serve 75.4% of AI data center spend by 2035.
Rack-Scale Systems Push Densities Far Past the Industry Average
The NVIDIA GB200 NVL72 connects 36 Grace CPUs and 72 Blackwell GPUs in one liquid-cooled rack, and NVIDIA’s chief executive said its planned Kyber rack for Rubin Ultra will consume about 600 kW. The Uptime Institute’s 2026 survey, as reported by Data Center Knowledge, put the average rack at 7.8 kW, excluding a small number of facilities designed above 30 kW, up from 7.5 kW in 2025.
Liquid Cooling Becomes the Default for AI Halls
Direct-to-chip liquid cooling takes 44.1% of AI Data Center Market spend in 2025 and grows at 21.61% a year, while air-cooled AI spend declines at 1.85% a year. Immersion cooling is the fastest-growing cooling technology at 28.48% a year from a small base of USD 10.15 billion. Schneider Electric completed the purchase of a controlling interest in Motivair in February 2025, and the Direct-to-Chip Liquid Cooling Market report sizes the cold plates and manifolds inside these halls.
Campuses Measured in Gigawatts
The largest AI data centers are now planned at 1 GW to 5 GW. Meta said its Prometheus cluster comes online in 2026 and its Hyperion campus will be able to scale up to 5 GW over several years. The five Stargate sites announced in September 2025 brought Stargate to nearly 7 GW of planned capacity and over USD 400 billion of investment. DC Market Insights models new AI hall completions rising from 4.20 GW in 2025 to 13.20 GW in 2030 and 17.82 GW in 2035.
Market Challenges: What Holds the AI Data Center Market Back?
Electricity supply and short hardware lives are the two constraints that slow the AI Data Center Market from 54.3% growth in 2026 to about 6.5% a year by 2035, in DC Market Insights estimates.
Power Demand More Than Doubles to 945 TWh by 2030
The International Energy Agency (IEA) estimates that data centers used about 415 terawatt-hours (TWh) of electricity in 2024, around 1.5% of world consumption, and projects about 945 TWh by 2030, with demand from AI-optimized data centers more than quadrupling. BloombergNEF said in December 2025 that US data center power demand could reach 106 GW in 2035. DC Market Insights expects grid delays to push developers toward on-site generation and models facility cost per MW rising from USD 13.08 million in 2025 to USD 13.52 million in 2028 before design standardization brings it down.
Rapid Hardware Cycles Compress Asset Lives
AI hardware is replaced far sooner than the buildings around it. DC Market Insights models a large share of future deployments as refresh of existing AI halls rather than new capacity: of 49.83 GW of AI IT load deployed in 2035, only 17.82 GW goes into newly completed halls. DC Market Insights sees residual-value risk for buyers that finance hardware over five or six years, because each new generation lowers the cost per unit of compute. Component costs add to the strain: Microsoft said its quarterly capital expenditure of USD 41 billion in the June 2026 quarter included the impact of higher component pricing.
Market Opportunities: Where Are the New Revenue Pools in AI Data Centers?
The two largest new revenue pools in the AI Data Center Market are GPU cloud and neocloud operators, growing at 18.45% a year, and sovereign and research AI programs, growing at 19.15% a year.
GPU Clouds and Neoclouds Grow 18.45% a Year
Specialist GPU cloud operators bought 12.6% of 2025 AI data center capacity, and DC Market Insights forecasts their spend to grow 18.45% a year, the fastest of the large buyer groups. CoreWeave ended June 2026 with about USD 104 billion of revenue backlog, 1.5 GW of active power and about 3.7 GW of contracted power, up from more than 850 MW active at the end of 2025. In April 2026 it signed an expanded agreement with Meta worth about USD 21 billion through December 2032.
Sovereign AI Spending Grows Faster Than the Market
Governments now fund AI capacity as national infrastructure. The European Commission’s InvestAI initiative aims to mobilize EUR 200 billion for AI, including a EUR 20 billion fund for AI gigafactories. The UAE-US AI campus in Abu Dhabi is planned at 5 GW, and Saudi Arabia’s HUMAIN, a subsidiary of the Public Investment Fund (PIF), targets 1.9 GW of data center capacity by 2030. DC Market Insights estimates that sovereign and research buyers hold 4.1% of the market and grow 19.15% a year.
Pricing: How Much Does an AI Data Center Cost per MW?
An AI data center costs about USD 44.70 million per MW of IT load in 2025 when built new and fully equipped: USD 31.62 million of IT hardware and USD 13.08 million of power, cooling and building shell, before software and services. These are DC Market Insights estimates.
| AI data center cost per MW (USD million) | 2020 | 2025 | 2030 | 2035 |
|---|---|---|---|---|
| IT hardware (servers, networking, storage) | 16.10 | 31.62 | 28.12 | 22.84 |
| Power, cooling and shell (new AI hall) | 9.46 | 13.08 | 13.24 | 12.19 |
| AI IT load deployed (GW) | 0.92 | 9.04 | 27.86 | 49.83 |
| New AI hall completions (GW) | 0.25 | 4.20 | 13.20 | 17.82 |
IT hardware cost per MW doubled from USD 16.10 million in 2020 to USD 31.62 million in 2025 as each GPU generation packed costlier silicon and memory into the same power; it peaks at USD 33.08 million in 2026, then falls about 4% a year as accelerator competition widens. Facility cost per MW rises with liquid cooling, then eases after 2028. For buyers planning scalable AI data center infrastructure, a 1 GW AI campus at USD 44.70 million per MW represents about USD 44.70 billion of equipment and construction.
Standards and Regulation: Which Rules Shape AI Data Center Buildouts?
Three sets of rules shape where AI data centers are built: US federal permitting policy for projects above 100 MW, European Union (EU) energy reporting for sites of at least 500 kW, and EU funding for AI gigafactories.
US permitting (July 2025). The Executive Order on Accelerating Federal Permitting of Data Center Infrastructure targets projects in which the sponsor has committed at least USD 500 million or that add more than 100 MW of electric load. America’s AI Action Plan, published the same day, lists over 90 federal policy actions, including rapid buildout of data centers.
EU energy reporting. Commission Delegated Regulation (EU) 2024/1364 requires operators of data centers with installed IT power of at least 500 kW to report key performance indicators to a European database by 15 May each year. The Commission proposed a common rating scheme for data centers on 21 September 2026, a step toward comparable efficiency ratings for AI halls across member states.
EU AI gigafactories. The Commission plans up to seven AI gigafactories across Europe, supported by up to EUR 10 billion of EU and national funding, and InvestAI adds a EUR 20 billion fund for these gigafactories.
Market Segmentation: Which Segment Leads the AI Data Center Market?
Hardware leads the AI Data Center Market with 87.3% of 2025 value, hyperscale data centers lead buyers with 55.4%, and generative AI and large language models lead applications with 61.5%. Inference is the fastest-growing segment at 20.56% a year, against 9.37% for training, because each trained model is then served to users at scale; among data center types, edge and micro data centers grow fastest at 21.94% a year from a small base.
By Component: Hardware Leads with 87.3%
| Component | 2025 (USD billion) | Share 2025 | 2035 (USD billion) | CAGR 2025-2035 |
|---|---|---|---|---|
| Hardware | 340.78 | 87.3% | 1,355.34 | 14.80% |
| Software and Orchestration | 17.04 | 4.4% | 94.87 | 18.73% |
| Services | 32.37 | 8.3% | 162.64 | 17.52% |
Software and orchestration grows fastest at 18.73% a year as cluster schedulers, inference-serving platforms and data center infrastructure management (DCIM) tools take a larger share of each deployment.
By Hardware Layer: AI Servers and Accelerators Lead with 60.9%
| Hardware layer | 2025 (USD billion) | Share 2025 | 2035 (USD billion) | CAGR 2025-2035 |
|---|---|---|---|---|
| AI Servers and Accelerators | 237.82 | 60.9% | 904.80 | 14.30% |
| Networking and Interconnect | 32.30 | 8.3% | 149.09 | 16.53% |
| Storage | 15.72 | 4.0% | 84.22 | 18.27% |
| Power and Cooling Infrastructure | 34.06 | 8.7% | 139.68 | 15.16% |
| Facility Shell and Construction | 20.88 | 5.4% | 77.55 | 14.02% |
Storage, from direct-attached AI storage inside GPU servers to networked flash arrays, grows fastest at 18.27% a year as inference and retrieval keep large datasets close to the accelerators. The accelerator volumes behind the largest layer are sized in the Data Center GPUs Market report.
By Data Center Type: Hyperscale Leads with 55.4%
| Data center type | 2025 (USD billion) | Share 2025 | 2035 (USD billion) | CAGR 2025-2035 |
|---|---|---|---|---|
| Hyperscale | 216.17 | 55.4% | 785.37 | 13.77% |
| Colocation | 67.11 | 17.2% | 305.39 | 16.36% |
| GPU Cloud and Neocloud | 49.16 | 12.6% | 267.36 | 18.45% |
| Enterprise | 35.51 | 9.1% | 117.10 | 12.67% |
| Sovereign and Research | 16.00 | 4.1% | 92.26 | 19.15% |
| Edge and Micro | 6.24 | 1.6% | 45.38 | 21.94% |
Colocation grows faster than hyperscale at 16.36% a year because hyperscalers lease more capacity from developers such as Aligned, which had 5 GW of active and planned capacity when it agreed to be sold in a USD 40 billion deal in October 2025. Enterprise AI data centers grow slowest at 12.67%: enterprises run AI and machine learning in the cloud first and bring only steady inference and sensitive data on premises, which forces denser racks and liquid cooling into existing rooms.
By Workload: Training Leads with 58.3% in 2025
Training grows at 9.37% a year to USD 556.99 billion by 2035, while inference reaches USD 1,055.87 billion.
By Cooling Technology: Air Cooling Leads with 47.8%
Air-cooled AI capacity is still the largest cooling group at USD 186.51 billion in 2025 but shrinks to USD 154.80 billion by 2035. Rear-door and hybrid cooling holds USD 21.46 billion (5.5%) and immersion cooling USD 10.15 billion (2.6%).
By Application: Generative AI and Large Language Models Lead with 61.5%
Generative AI and large language models (LLMs) account for USD 239.97 billion in 2025 and grow at 16.81% a year to USD 1,134.63 billion. Recommendation and predictive machine learning accounts for USD 69.45 billion (17.8%) and grows at 11.32%. Scientific computing and others hold USD 30.04 billion (7.7%), computer vision USD 28.87 billion (7.4%) and speech and natural language processing USD 21.85 billion (5.6%).
By Vertical: IT and Telecom Lead with 63.4%
IT and telecom, which includes cloud and internet platforms, accounts for USD 247.38 billion of 2025 demand and grows at 14.47% a year. Banking, financial services and insurance (BFSI) follows with USD 30.44 billion (7.8%), then automotive and manufacturing with 6.6%, healthcare and life sciences with 5.9%, retail and e-commerce with 5.2%, government and defense with 5.1%, media and entertainment with 3.4% and others with 2.6%. Healthcare and life sciences grows fastest at 18.46% a year.
DC Exclusive: The AI Capacity Cost Stack, Data Center Type × Spend Layer
At USD 138.04 billion in 2025, accelerated servers bought for hyperscale data centers are the single largest cell of the AI Capacity Cost Stack and 35.4% of all AI data center spending.
| 2025, USD billion | Servers and Accelerators | Networking and Storage | Power, Cooling and Shell | Software and Services | Total |
|---|---|---|---|---|---|
| Hyperscale | 138.04 | 26.03 | 30.44 | 21.65 | 216.17 |
| Colocation | 34.52 | 8.14 | 15.43 | 9.03 | 67.11 |
| GPU Cloud and Neocloud | 32.51 | 6.96 | 2.90 | 6.80 | 49.16 |
| Enterprise | 19.73 | 4.44 | 3.37 | 7.98 | 35.51 |
| Sovereign and Research | 9.35 | 1.76 | 1.95 | 2.93 | 16.00 |
| Edge and Micro | 3.68 | 0.69 | 0.85 | 1.02 | 6.24 |
| Total | 237.82 | 48.02 | 54.94 | 49.41 | 390.19 |
Neoclouds spend 66.1% of their budget on servers and accelerators and only 5.9% on power, cooling and shell, because they lease powered buildings. Colocation spend is the most facility-heavy at 23.0%, so colocation developers are the main buyers of power and cooling equipment for AI. Enterprises spend 22.5% on software and services, the highest share, because they buy expertise along with hardware. The AI Capacity Cost Stack is a DC Market Insights model, not a shipment count: it is built from 112 inputs: 16 annual cohorts each of AI IT deployments, IT hardware cost per MW, new AI hall completions, facility cost per MW, software ratio and services ratio (96 inputs), 6 data center type shares, 7 spend layer totals and 3 supplier disclosures used as a cross-check. The model and its assumptions are explained on our research methodology page.
Regional Insights: Which Region Leads the AI Data Center Market?
North America leads the AI Data Center Market with 54.8% of 2025 value, and the Middle East and Africa is the fastest-growing region at 22.17% a year because sovereign AI programs in the Gulf Cooperation Council (GCC) countries are building gigawatt campuses from a small base.
North America Leads with 54.8% Share
North American AI data center spend reached USD 213.83 billion in 2025 and is forecast to reach USD 772.02 billion by 2035, a CAGR of 13.70%. The region hosts the largest hyperscale and neocloud campuses, including Stargate’s Abilene site, which OpenAI described as already running on Oracle Cloud Infrastructure in September 2025.
Asia Pacific Holds 25.6% Share
Asia Pacific is worth USD 99.89 billion in 2025 and USD 459.60 billion by 2035, a CAGR of 16.49%. DC Market Insights estimates that China accounts for more than half of the region, while Japan, South Korea, India and Australia add hyperscale and sovereign capacity; India’s national program alone counted 34,333 GPUs in May 2025.
Europe Holds 11.9% Share
Europe is worth USD 46.43 billion in 2025 and USD 186.14 billion by 2035, a CAGR of 14.89%. The EU plans up to seven AI gigafactories, and InvestAI adds a EUR 20 billion fund for them.
Middle East and Africa Holds 4.6% Share
The Middle East and Africa is worth USD 17.95 billion in 2025 and USD 132.94 billion by 2035, growing fastest at 22.17% a year. The GCC drives the region: Stargate UAE is a 1 GW cluster in Abu Dhabi with 200 MW expected to go live in 2026, the UAE-US AI campus in Abu Dhabi is planned at 5 GW, and HUMAIN plans to grow from 1.9 GW in 2030 to 6.6 GW four years later. We assess power, fiber and permitting for these markets in our data center site selection and feasibility work.
Latin America Holds 3.1% Share
Latin America is worth USD 12.10 billion in 2025 and USD 62.16 billion by 2035, a CAGR of 17.78%. DC Market Insights allocates USD 5.59 billion of 2025 spend to Brazil, growing 18.63% a year to USD 30.85 billion, and USD 3.28 billion to Mexico, growing 19.08% a year to USD 18.79 billion.
Country Analysis: Which Countries Spend the Most on AI Data Centers?
The United States is the largest country market for AI data centers at USD 195.01 billion in 2025, 50.0% of global value, followed by China at USD 54.74 billion (14.0%) and Canada at USD 15.82 billion (4.1%). The 12 largest countries account for 86.6% of 2025 spending.
| Rank | Country | Region | 2025 (USD billion) | Share 2025 | 2035 (USD billion) | CAGR 2025-2035 |
|---|---|---|---|---|---|---|
| 1 | United States | North America | 195.01 | 50.0% | 699.20 | 13.62% |
| 2 | China | Asia Pacific | 54.74 | 14.0% | 232.68 | 15.57% |
| 3 | Canada | North America | 15.82 | 4.1% | 62.15 | 14.66% |
| 4 | Japan | Asia Pacific | 11.79 | 3.0% | 48.02 | 15.08% |
| 5 | United Kingdom | Europe | 10.96 | 2.8% | 42.18 | 14.43% |
| 6 | Germany | Europe | 9.47 | 2.4% | 38.19 | 14.96% |
| 7 | South Korea | Asia Pacific | 8.09 | 2.1% | 37.52 | 16.58% |
| 8 | India | Asia Pacific | 7.19 | 1.8% | 63.37 | 24.31% |
| 9 | France | Europe | 6.83 | 1.7% | 31.09 | 16.37% |
| 10 | Saudi Arabia | Middle East and Africa | 6.32 | 1.6% | 62.92 | 25.84% |
| 11 | Australia | Asia Pacific | 5.79 | 1.5% | 28.55 | 17.29% |
| 12 | United Arab Emirates | Middle East and Africa | 5.71 | 1.5% | 49.33 | 24.07% |
Saudi Arabia is the fastest-growing country market at 25.84% a year, ahead of India at 24.31% and the United Arab Emirates at 24.07%. The country figures are a DC Market Insights allocation of the regional totals, weighted by each country’s estimated installed and planned AI capacity.
Forecast Scenarios: How Could the 2035 Forecast Change?
The AI Data Center Market reaches USD 1,612.86 billion by 2035 in the base case, with a range of USD 1,205.21 billion to USD 1,934.58 billion across the low and high scenarios.
| Scenario | Assumption from 2026 | 2035 (USD billion) | CAGR 2025-2035 |
|---|---|---|---|
| Low | AI IT deployments 22% below base; hardware cost per MW 5% lower | 1,205.21 | 11.94% |
| Base | Deployments rise to 49.83 GW a year; new AI halls reach 17.82 GW a year | 1,612.86 | 15.25% |
| High | Deployments 17% above base; hardware cost per MW 3% higher | 1,934.58 | 17.36% |
The high case assumes grid connections and memory supply keep pace with gigawatt plans such as Stargate and Hyperion. Power availability is the main swing factor: a gigawatt of AI IT load that cannot be energized on time moves about USD 44.70 billion of spend into a later year.
Competitive Insights: Who Leads the AI Data Center Market?
- NVIDIA
- AMD
- Broadcom
- Dell Technologies
- Supermicro
- Hewlett Packard Enterprise (HPE)
- Cisco
- Vertiv
- Schneider Electric
- Microsoft
- Amazon Web Services (AWS)
- Meta
- Oracle
- CoreWeave
- Equinix
- Digital Realty
NVIDIA is the largest supplier in the AI Data Center Market with about half of 2025 value (DC Market Insights estimate: NVIDIA’s Data Center revenue of USD 193.7 billion in the fiscal year that ended in January 2026, set against the USD 390.19 billion market). NVIDIA’s Data Center revenue reached USD 62.3 billion in the January 2026 quarter and USD 89.0 billion in the quarter that ended in July 2026, up 117% from a year earlier. AMD’s Data Center segment revenue reached USD 16.6 billion in 2025, and Broadcom’s AI hardware revenue reached about USD 20 billion in its fiscal 2025. Dell shipped more than USD 25 billion of AI-optimized servers in its fiscal 2026 and entered fiscal 2027 with a USD 43 billion backlog, and Supermicro’s net sales reached USD 39.1 billion in its fiscal year to June 2026. In the facility layer, Vertiv ended 2025 with a USD 15.0 billion backlog, up 109%, and data centers made up 30% of Schneider Electric’s end-market exposure, measured by 2025 orders.
Recent Developments
- In January 2025, OpenAI, SoftBank, Oracle and MGX announced Stargate, which intends to invest USD 500 billion over four years in new AI infrastructure, starting with USD 100 billion (Source: OpenAI announcement).
- In September 2025, OpenAI and NVIDIA announced a letter of intent to deploy at least 10 GW of NVIDIA systems, with NVIDIA intending to invest up to USD 100 billion in OpenAI (Source: NVIDIA press release).
- In October 2025, AMD and OpenAI announced an agreement for OpenAI to deploy 6 GW of AMD GPUs over multiple generations (Source: AMD filing with the US Securities and Exchange Commission).
- In October 2025, Aligned Data Centers, with 5 GW of active and planned capacity, agreed to be sold to a consortium including BlackRock’s GIP and MGX in a deal worth USD 40 billion (Source: Data Center Dynamics news report).
- In April 2026, CoreWeave signed an expanded long-term agreement with Meta worth about USD 21 billion through December 2032 (Source: CoreWeave filing with the US Securities and Exchange Commission).
- In August 2026, NVIDIA reported Data Center revenue of USD 89.0 billion for the quarter that ended in July 2026, up 117% from a year earlier (Source: NVIDIA financial results release).
We follow these deals, capacity announcements and policy moves as they happen in our data center industry updates.
Methodology: How We Built the AI Data Center Market Model
DC Market Insights built the AI Data Center Market bottom-up from 16 annual cohorts (2020 to 2035) of AI IT load deployed, from 0.92 GW in 2020 and 9.04 GW in 2025 to 49.83 GW in 2035, each priced at an IT hardware cost per MW (USD 31.62 million in 2025). New AI hall completions, taken as an AI share of total data center completions (4.20 GW of 12.0 GW in 2025), were priced at a facility cost per MW (USD 13.08 million in 2025). Software and orchestration (5.0% of hardware in 2025, rising to 7.0%) and services (9.5%, rising to 12.0%) were added on top. The total was split by 3 components, 5 hardware layers, 6 data center types, 2 workloads, 4 cooling technologies, 5 applications, 8 verticals and 5 regions, and every split sums to the market total. The result was checked top-down against NVIDIA’s USD 193.7 billion of fiscal 2026 Data Center revenue, USD 291.97 billion of 2025 capital expenditure at Alphabet, Amazon and Meta, and disclosures by Dell, Supermicro and Vertiv. The analyst named on this page built the model, and the reviewer checked the arithmetic, the sources and the dates before publication.
Table of contents
- 01Introduction
- 1.1Market Definition and Scope
- 1.2Report Coverage and Exclusions
- 1.3Currency, Units and Base Year
- 1.4Key Stakeholders
- 02Research Methodology
- 2.1Bottom-Up Capacity Cohort Model
- 2.2Top-Down Validation Against Supplier and Buyer Disclosures
- 2.3Primary and Secondary Sources
- 2.4Forecasting Assumptions
- 03Executive Summary
- 3.1Market Snapshot, 2025 and 2035
- 3.2Key Findings
- 3.3Analyst Takeaways
- 04Market Dynamics
- 4.1Drivers
- 4.2Challenges
- 4.3Opportunities
- 4.4Trends in the AI Compute and Data Landscape
- 05Pricing Analysis
- 5.1IT Hardware Cost per MW, 2020-2035
- 5.2Facility Cost per MW, 2020-2035
- 5.3All-In Cost of a 1 GW AI Campus
- 06Standards and Regulation
- 6.1US Federal Permitting of Data Center Infrastructure
- 6.2EU Data Center Energy Reporting and Rating Scheme
- 6.3EU AI Gigafactories and InvestAI
- 07Supply Chain Analysis
- 7.1Accelerator and Server Suppliers
- 7.2Networking and Storage Suppliers
- 7.3Power and Cooling Equipment Suppliers
- 7.4Developers, Colocation and Neocloud Operators
- 08AI Data Center Market Size and Forecast
- 8.1Market Size, 2020-2035
- 8.2AI IT Load Deployed and New AI Hall Completions, 2020-2035
- 8.3Year-on-Year Growth
- 09Market by Data Center Type
- 9.1Hyperscale
- 9.2Colocation
- 9.3GPU Cloud and Neocloud
- 9.4Enterprise
- 9.5Sovereign and Research
- 9.6Edge and Micro
- 10Market by Component
- 10.1Hardware
- 10.2Software and Orchestration
- 10.3Services
- 11Market by Hardware Layer
- 11.1AI Servers and Accelerators
- 11.2Networking and Interconnect
- 11.3Storage
- 11.4Power and Cooling Infrastructure
- 11.5Facility Shell and Construction
- 12Market by Workload
- 12.1Training
- 12.2Inference
- 13Market by Cooling Technology
- 13.1Air Cooling
- 13.2Direct-to-Chip Liquid Cooling
- 13.3Rear-Door and Hybrid Cooling
- 13.4Immersion Cooling
- 14Market by Application
- 14.1Generative AI and Large Language Models
- 14.2Recommendation and Predictive Machine Learning
- 14.3Computer Vision
- 14.4Speech and Natural Language Processing
- 14.5Scientific Computing and Others
- 15Market by Vertical
- 15.1IT and Telecom
- 15.2BFSI
- 15.3Automotive and Manufacturing
- 15.4Healthcare and Life Sciences
- 15.5Retail and E-commerce
- 15.6Government and Defense
- 15.7Media and Entertainment
- 15.8Others
- 16DC Exclusive: AI Capacity Cost Stack
- 16.1Data Center Type × Spend Layer Matrix, 2025
- 16.2Model Inputs and Cross-Checks
- 17Market by Region
- 17.1North America (United States, Canada)
- 17.2Asia Pacific (China, Japan, South Korea, India, Australia)
- 17.3Europe (United Kingdom, Germany, France)
- 17.4Middle East and Africa (Saudi Arabia, United Arab Emirates)
- 17.5Latin America
- 18Country Analysis
- 18.1Top 12 Countries, 2025 and 2035
- 18.2Fastest-Growing Countries
- 19Regional Cross-Sections
- 19.1Region Share and Growth Comparison, 2025 and 2035
- 19.2Country Share Within Each Region, 2025
- 20Forecast Scenarios
- 20.1Low, Base and High Scenarios to 2035
- 20.2Key Swing Factors
- 21Competitive Landscape
- 21.1Market Share Analysis, 2025
- 21.2Supplier Positioning by Layer
- 21.3Recent Developments
- 22Company Profiles
- 22.1NVIDIA
- 22.2AMD
- 22.3Broadcom
- 22.4Dell Technologies
- 22.5Supermicro
- 22.6Hewlett Packard Enterprise (HPE)
- 22.7Cisco
- 22.8Vertiv
- 22.9Schneider Electric
- 22.10Microsoft
- 22.11Amazon Web Services (AWS)
- 22.12Google
- 22.13Meta
- 22.14Oracle
- 22.15CoreWeave
- 22.16Equinix
- 22.17Digital Realty
- 23Analyst Recommendations
- 23.1For Equipment Suppliers
- 23.2For Operators and Developers
- 23.3For Investors
- Fig. 1AI Data Center Market Size, 2020-2035 (USD billion)
- Fig. 2AI Data Center Market by Region, 2025 and 2035 (USD billion)
- Fig. 3AI Data Center Market Share by Data Center Type, 2025
- Fig. 4AI Data Center Market by Hardware Layer, 2025 and 2035
- Fig. 5Training vs Inference Spend, 2025 and 2035
- Fig. 6Cooling Technology Mix, 2025 and 2035
- Fig. 7AI Capacity Cost Stack, 2025
- Fig. 8Forecast Scenarios, 2035
- Table 1AI Data Center Market Size, 2020-2035 (USD billion)
- Table 2AI Data Center Cost per MW, 2020, 2025, 2030 and 2035
- Table 3Market by Component, 2025 and 2035
- Table 4Market by Hardware Layer, 2025 and 2035
- Table 5Market by Data Center Type, 2025 and 2035
- Table 6Market by Workload, 2025 and 2035
- Table 7Market by Cooling Technology, 2025 and 2035
- Table 8Market by Application, 2025 and 2035
- Table 9Market by Vertical, 2025 and 2035
- Table 10AI Capacity Cost Stack: Data Center Type × Spend Layer, 2025
- Table 11Market by Region, 2025 and 2035
- Table 12Top 12 Countries, 2025 and 2035
- Table 13Forecast Scenarios, 2035
About this report
Written by Amit Jain and reviewed by Deepti Agrawal, Senior Editor, Research. Figures are built top-down and bottom-up and reconciled before publication, with 2025 as the base year.
Read our methodologyHow we built this
- Historical period2020-2024
- Base year2025
- Forecast period2026-2035
- Sizing approachTop-down + bottom-up
Need more than the report?
Advisory on this market
Due diligence, site selection and market entry work, by the analysts who wrote this report.
See consulting servicesFrequently asked questions
USD 390.19 billion in 2025 and USD 1,612.86 billion by 2035, a CAGR of 15.25%, according to DC Market Insights estimates. The 2025 value covers 9.04 GW of AI IT load deployed and 4.20 GW of newly completed AI halls.
Up to 5 GW for the largest campuses: Meta says its Hyperion campus will be able to scale up to 5 GW, and the UAE-US AI campus in Abu Dhabi is planned at 5 GW. The Uptime Institute’s 2026 survey puts the average rack at 7.8 kW, while NVIDIA’s planned Kyber rack will consume about 600 kW.
44.1% of 2025 AI data center spend already uses direct-to-chip liquid cooling, and DC Market Insights expects 75.4% by 2035. Rack-scale systems such as the liquid-cooled NVIDIA GB200 NVL72 and campuses measured in gigawatts are the other two trends.
USD 35.51 billion of enterprise AI data center spend in 2025, growing 12.67% a year, is the DC Market Insights estimate. Enterprises run AI in the cloud first and bring steady inference and sensitive data on premises, which forces denser racks and liquid cooling into existing rooms.
54.8% of 2025 value, or USD 213.83 billion, sits in North America. The Middle East and Africa grows fastest at 22.17% a year as sovereign AI programs in the GCC, such as Stargate UAE and Saudi Arabia’s HUMAIN, build gigawatt campuses.
About half of 2025 value goes to NVIDIA, in DC Market Insights estimates based on its USD 193.7 billion of Data Center revenue in the fiscal year that ended in January 2026. AMD, Broadcom, Dell, Supermicro, Vertiv, Schneider Electric and the hyperscalers follow.
About USD 44.70 million per MW in 2025: USD 31.62 million of IT hardware and USD 13.08 million of power, cooling and shell, in DC Market Insights estimates. A 1 GW AI campus therefore represents about USD 44.70 billion of equipment and construction.
65.5% of AI data center spend will go to inference by 2035, up from USD 162.71 billion in 2025, as inference grows 20.56% a year in DC Market Insights estimates. New AI hall completions rise from 4.20 GW in 2025 to 17.82 GW in 2035.
USD 390.19 billion in 2025 is the value of everything bought to put AI compute capacity into service. It covers accelerated servers, AI networking and storage, power, cooling and shell of AI halls, cluster software and services. Smart or autonomous data center software counts only when bought for an AI hall.
Land, utility-side grid assets, general-purpose servers without accelerators, devices outside data center buildings, electricity bills, and cloud or colocation rental revenue are excluded. Rental revenue is left out because it would count the same equipment twice.
Hardware is USD 340.78 billion of the USD 390.19 billion market in 2025, and AI servers and accelerators account for USD 237.82 billion of that. Software and services add USD 49.41 billion.
Get this report
How do I get the full report?
Request a quote; we reply by email within one business day.
How and when will I receive the report?
By email, within 24-48 business hours of payment.
How can I pay?
Once you accept the quote, we email you a secure payment link.
Can I see a sample first?
Yes. Request a free sample and we will email it to you.
Can the report be tailored to my needs?
Yes. We can add countries, segments or data cuts. Talk to an analyst about the scope you need.
Can I get an invoice or pay against a purchase order?
Write to sales@dcmarketinsights.com and our team will help.
