GPU向け液冷市場シェア分析、業界動向と統計、成長予測 2026-2031年

GPU向け液冷市場の規模は、2025年の69億米ドルから2026年には83億5,000万米ドルへと拡大し、2026年から2031年にかけて年平均成長率（CAGR）26.37%で推移して、2031年には269億1,000万米ドルに達するとMordor Intelligenceでは予測しています。1チップあたり1,000ワットを超える熱を発生させるAIアクセラレータの急速な普及に伴い、データセンターは従来の空冷方式からの転換を余儀なくされており、現在では液冷ソリューションがハイパースケール・キャンパスにおける容量拡張を支える基盤となっています。機器メーカーは、液冷システムを工場出荷時に組み込んだサーバーの提供を開始しており、これにより、高コストな後付け改修が不要となり、導入期間の短縮も実現しています。北米や欧州におけるPUE（電力使用効率）規制の強化は、規制への対応という圧力を、液冷プロジェクトへの即時的な設備投資予算へと転換させています。導入量ではアジア太平洋地域が先行していますが、北米では二相式蒸発冷却プラットフォームの早期導入を通じて、技術的なベンチマークが確立されつつあります。

レポートの主なポイント

• 冷却方式別では、2025年のGPU液冷市場において単相式ダイレクト・ツー・チップ（Direct-to-Chip）方式が売上高シェア73%を占めて首位となりました。一方、二相式システムは2031年まで年平均成長率（CAGR）27.80%で拡大すると予測されています。
• 冷却レベル別では、2025年のGPU液冷市場においてコンポーネントレベルのコールドプレートがシェア56%を占めました。対照的に、ラックレベルの液浸冷却は2026年から2031年にかけて28.10%の成長が見込まれています。
• 導入形態別では、2025年のGPU液冷市場の売上高においてハイパースケーラーおよびクラウドプロバイダーが64%を占めましたが、エンタープライズ分野も2031年までCAGR 26.80%で推移する見通しです。
• GPUの電力密度別では、2025年のGPU液冷市場において300W～700Wの区分が52%を占めました。一方、700W超の区分は2031年までCAGR 28.60%で拡大しています。
• 地域別では、2025年のGPU液冷市場においてアジア太平洋地域がシェア68%を占めて市場を主導し、2026年から2031年にかけてはCAGR 29.10%で拡大すると予測されています。

冷却方式別：極めて高い熱負荷下で高まる二相冷却への機運

2025年の売上高において単相冷却システムが73%を占めました。これは、同技術が10年にわたり現場で成熟し、既存のチラー（冷却水供給装置）設備とも円滑に統合できるためです。これらのシステムは、30℃～45℃の水・グリコール混合液を循環させてチップから発生する熱の最大85%を回収し、既存の建物の冷水循環系に直接組み込むことが可能です。単相冷却ソリューション向けのGPU液冷市場は、ハイパースケーラーがOpen Compute Project（OCP）の下でマニホールド設計を標準化した恩恵を受け、システム統合にかかる工数を5分の1近く削減しました。Lenovo Neptune、Dell PowerEdge XE9680L、HPE Cray EXといったサーバーの継続的な出荷により、特に700W級のアクセラレーターが主流となっているワークロードにおいて、導入済みシステムの基盤が維持されています。

二相冷却プラットフォームは年平均成長率（CAGR）27.80%で拡大すると予測されており、GPU液冷市場の中で最も急成長するセグメントとなる見込みです。蒸発式コールドプレートは、1,000Wを超える熱負荷のほぼ全量をシリコン上で直接回収し、高温の状態でチラーへ戻すため、一年の大半において「ニア・フリー・クーリング（外気利用等による低エネルギー冷却）」が可能になります。Vertiv Liebert PCWやZutaCore HyperCoolを用いた実証実験では、チップレベルでの熱回収率98%を達成し、冷却エネルギー消費を半減させるとともに、電力変換機器を設置するためのラックスペースを確保できることが実証されました。誘電性冷媒の管理に伴い材料適合性に関するより厳しい要件が課されるものの、ハイパースケーラーは、将来のBlackwell Ultraクラスターを収容するために全く新しいデータセンター棟を建設する場合のコストと、こうした要件に伴うコストを天秤にかけて検討を進めています。

GPU Liquid Cooling Market Analysis by Mordor Intelligence

The GPU liquid cooling market size is expected to increase from USD 6.90 billion in 2025 to USD 8.35 billion in 2026 and reach USD 26.91 billion by 2031, growing at a CAGR of 26.37% over 2026-2031. Rapid adoption of AI accelerators that dissipate more than 1,000 watts per chip is forcing data centers to abandon conventional air cooling, and liquid solutions now underpin capacity expansions at hyperscale campuses. Equipment makers have begun shipping factory-integrated liquid-cooled servers, eliminating costly retrofits and compressing deployment timelines. Stricter power-usage-effectiveness (PUE) mandates across North America and Europe are converting regulatory pressure into immediate capital budgets for liquid projects. Asia-Pacific leads deployment volumes, yet North America is setting technology benchmarks through early adoption of two-phase evaporative platforms.

Key Report Takeaways

• By cooling type, single-phase direct-to-chip solutions led the GPU liquid cooling market with a 73% revenue share in 2025, while two-phase systems are projected to expand at a 27.80% CAGR through 2031.
• By cooling level, component-level cold plates held 56% of the GPU liquid cooling market share in 2025, whereas rack-level immersion is forecast to grow at 28.10% between 2026 and 2031.
• By deployment, hyperscale and cloud providers captured 64% revenue of the GPU liquid cooling market in 2025, yet the enterprise segment is on track for a 26.80% CAGR to 2031.
• By GPU power density, the 300-watt to 700-watt band accounted for 52% of the GPU liquid-cooling market in 2025, while the above-700-watt category is advancing at a 28.60% CAGR through 2031.
• By geography, Asia-Pacific dominated the graphics processing unit (GPU) liquid cooling market with a 68% share in 2025 and is anticipated to expand at a 29.10% CAGR over 2026-2031.

Note: Market size and forecast figures in this report are generated using Mordor Intelligence’s proprietary estimation framework, updated with the latest available data and insights as of January 2026.

Global GPU Liquid Cooling Market Trends and Insights

Drivers Impact Analysis*

Hyperscale Rack Density Pushes Liquid Cooling Into the Mainstream

Rack-level power density ballooned from 27 kW in 2024 to well above 100 kW in 2026, and leading campuses already test 250 kW racks. At these thermal loads, airflow velocity becomes impractical, and acoustic limits prohibit further fan speed increases. Operators therefore retrofit existing halls with coolant distribution units that tolerate 45 °C supply temperatures, allowing chillers to run in economizer mode all year. Early adopters documented 35% smaller building footprints, saving more than USD 180 million in avoided concrete and steel. The outcome is a visible migration budgeted directly into hyperscale capital-expenditure plans rather than deferred operational spending.[1]

OEM Platforms Remove Integration Barriers

Server manufacturers moved liquid cooling from a specialized option to a default for high-performance GPU nodes. Factory-installed cold plates, quick disconnects, and leak-detection loops now ship on NVIDIA GB200, AMD MI325X, and Supermicro X14 systems. The turnkey approach cuts rack-level installation time from 16 hours to less than three, reducing commissioning labor by 80%. End users gain immediate performance headroom, as liquid-cooled variants sustain up to 18% higher clock frequencies under the same power envelope. The commercial message is clear that liquid cooling is no longer exotic; it is the shipping configuration for flagship AI hardware.[2]

Two-Phase Technology Answers 1 kW-Plus Chips

With chip thermal design power doubling in three years, single-phase water-glycol loops approach their material limits. Two-phase evaporative modules exploit the latent heat of vaporization, absorbing roughly tenfold more energy per kilogram of coolant. Field pilots kept 1,100-watt accelerators below 65 °C junction temperature while slashing chiller loads by 30%. Equipment makers integrate micro-evaporators directly on silicon packages, eliminating external pumps and shrinking coolant inventory. Standardization work inside the Open Compute Project now aligns refrigerant chemistry and manifold geometry, removing interoperability barriers for hyperscalers.[3]

Regulation Turns Efficiency From Nice-to-Have Into Legal Requirement

Germany, France, California, and the European Union have codified PUE thresholds between 1.15 and 1.20 for facilities commissioned after 2027. For AI workloads that push rack densities beyond 30 kW, meeting those limits without liquid cooling is mathematically impossible. Fiscal penalties include higher grid-connection fees and, in some regions, mandatory waste-heat recovery plans. Consequently, sustainability officers, facilities engineers, and finance teams converge on liquid cooling as the lowest-risk compliance path, accelerating deal cycles and pulling enterprise adoption forward by at least two years.[4]

Restraints Impact Analysis*

Capital Cost Premium Slows Immersion Uptake

Immersion tanks cost 30%-50% more than air-based cooling solutions and continue to carry a 15% cost uplift over single-phase cold plates, making upfront capital expenditure a key barrier for many buyers. Dielectric fluids further increase deployment costs, adding USD 15,000-30,000 per rack, and operators must budget for yearly top-offs that can approach 10% of total fluid volume. For enterprises that typically refresh or renew hardware every three years, these added costs extend the payback period beyond acceptable investment horizons, slowing adoption and delaying broader rollouts. Vendors are now pursuing cost reductions by using commodity fluids, simplifying system designs, and standardizing tank dimensions to improve manufacturing scale and installation efficiency. However, most observers expect immersion cooling to reach price parity with competing solutions only after 2028.

Environmental Rules Complicate Coolant Choice

European F-gas revisions ban high-GWP fluorocarbons by 2030, and the Netherlands already outlawed PFAS fluids in 2026, accelerating regulatory pressure on immersion cooling deployments that rely on these chemistries. The exit of 3M Novec from the market further forces operators and suppliers to transition rapidly to hydrocarbon or bio-based alternatives, which offer lower dielectric strength and higher viscosity compared with incumbent fluids. As a result, operators face unexpected retrofit bills for seal replacements and pump upgrades, creating additional technical and financial burdens. These requirements introduce uncertainty into capital planning, complicate procurement decisions, and temporarily curb the momentum of the immersion market.

*Our forecasts treat driver/restraint impacts as directional, not additive. The impact forecasts reflect baseline growth, mix effects, and variable interactions.

Segment Analysis

By Cooling Type: Two-Phase Momentum Builds at Extreme Thermal Loads

Single-phase systems accounted for 73% of revenue in 2025, reflecting a decade-long field maturity and seamless integration with legacy chillers. These installations circulate water-glycol mixtures between 30 °C and 45 °C, capture up to 85% of chip heat, and slot directly into existing building chilled-water loops. The GPU liquid-cooling market for single-phase solutions benefited from hyperscalers standardizing on manifold designs under the Open Compute Project, trimming integration labor by nearly one-fifth. Continued shipments of Lenovo Neptune, Dell PowerEdge XE9680L, and HPE Cray EX servers sustain the installed base, especially for workloads where 700-watt accelerators remain dominant.

Two-phase platforms are projected to grow at a 27.80% CAGR, making them the fastest-growing segment of the GPU liquid cooling market. Evaporative cold plates capture almost the entire 1,000-watt-plus heat load directly on silicon, returning liquid to chillers at elevated temperatures that support near-free cooling for much of the year. Vertiv Liebert PCW and ZutaCore HyperCool field trials demonstrate 98% chip-level heat capture, halving cooling energy use and freeing rack space for power-conversion gear. Though dielectric management imposes stricter material compatibility rules, hyperscalers weigh those costs against the alternative of building entirely new halls to house future Blackwell Ultra clusters.

By Cooling Level: Immersion Tanks Accelerate Deployment Timelines

Component-level cold plates retained 56% share in 2025, underpinned by ship-through volumes from server OEMs. Direct-to-chip plates require no new server enclosures and leverage standardized quick disconnects, making them attractive for enterprises performing gradual retrofits. Cold-plate loops also minimize fluid inventory, reducing environmental risk and meeting insurance-carrier guidelines with minimal paperwork. However, they still demand per-node plumbing labor and cannot fully exploit heat-re-use schemes because exhaust temperatures sit below 50 °C.

Rack-level immersion is forecast to expand at 28.10% per year. Submer, GRC, and LiquidStack tanks enable 200 kW racks within existing floor grids, shrink installation windows to a single afternoon, and simplify maintenance with live-service capabilities. The GPU liquid cooling market share for immersion rose sharply after ByteDance documented PUE 1.08 and 85% waste-heat recovery, influencing procurement strategies across mainland China. Financial models show that once electricity exceeds USD 0.10 per kilowatt-hour, immersion’s operational savings eclipse its capital premium in under three years, tilting total-cost-of-ownership calculations decisively.

By Deployment: Enterprise Momentum Builds Under Sustainability Scrutiny

Hyperscalers accounted for 64% of 2025 revenue, leveraging million-square-foot campuses and custom power-distribution schemes. Their early leadership established volume orders that anchored supplier cost curves and catalyzed broad ecosystem standardization. Even so, enterprise demand is catching up quickly, posting a forecast 26.80% CAGR as mid-tier colocation providers face Scope 2 disclosure obligations. For example, an OVHcloud retrofit cut PUE from 1.42 to 1.18, sparing EUR 3.20 million (USD 3.62 million) in annual grid surcharges and making liquid cooling a CFO-endorsed sustainability lever.

Edge AI and government research deployments remain niche but strategically important within the graphics processing unit (GPU) cooling solutions market. Edge rollouts favor sealed, pre-filled cooling modules that arrive ready for immediate power-on, as this approach eliminates the need for on-site coolant handling and simplifies deployment in distributed locations. In the public sector, procurement cycles often extend for 5 years or more due to funding approvals, technical evaluations, and compliance requirements. However, once these projects receive funding, large-scale systems such as Lawrence Livermore’s El Capitan require the highest thermal envelopes, making liquid cooling effectively mandatory for another decade.

By GPU Power Density: Above-700-Watt Nodes Drive New Revenue Streams

The 300-watt to 700-watt class still leads installed base metrics, but the above-700-watt tier is where suppliers extract margins and differentiation in the graphics processing unit (GPU) liquid-cooling market. NVIDIA GB200 and AMD MI325X accelerators, each cresting the 1,000-watt mark, are available only with liquid cooling, expanding the GPU liquid-cooling market with every shipment. Rack designs now host eight of these chips per 4U chassis, hitting 200 kW densities that are simply infeasible with any form of air cooling. Microsoft reported a 35% reduction in floor space after retrofitting H100 racks with liquid loops, highlighting that even mid-power silicon benefits from liquid cooling in high-density layouts.

Longer term, chip roadmaps indicate that devices could reach 1,500 watts by 2028, increasing the need for cooling infrastructure capable of supporting higher thermal loads. Suppliers are therefore future-proofing today’s infrastructure by oversizing coolant manifolds and adopting modular CDU frames that can accommodate future capacity requirements. Operators installing two-phase evaporators now can avoid another round of capital retrofits in three years, strengthening the business case for early investment. This argument resonates strongly with treasury teams scrutinizing cash yields, as it supports more predictable capital planning and reduces the risk of near-term infrastructure replacement.

Complete Report Scope:

• By Cooling Type
  • Single-Phase Liquid Cooling
  • Two-Phase Liquid Cooling
• By Cooling Level
  • Component-Level Cooling
  • Server Rack-Level Cooling
• By Deployment
  • Hyperscale Cloud
  • Enterprise
  • Government and Research HPC
  • Edge AI
• By GPU Power Density
  • Below 300 W
  • 300 W – 700 W
  • Above 700 W
• By Geography
  • North America
    • United States
    • Canada
    • Mexico
  • Europe
    • Germany
    • United Kingdom
    • France
    • Italy
    • Rest of Europe
  • Asia-Pacific
    • China
    • Japan
    • South Korea
    • India
    • Southeast Asia
    • Rest of Asia-Pacific
  • South America
    • Brazil
    • Rest of South America
  • Middle East and Africa

Geography Analysis

Asia-Pacific accounted for 68% revenue in 2025 and is forecast to grow 29.10% annually through 2031. Sovereign AI programs in China inject billions of dollars into national GPU clusters, while Southeast Asia leverages tax incentives to localize production of CDUs and quick disconnects. ByteDance documented a PUE of 1.08 at its Tianjin campus and shared blueprints with domestic cloud peers, creating a demonstration effect that ripples through the region’s colocation market. In parallel, Japan and South Korea channel state grants toward liquid-ready fabs, ensuring regional supply security against export-control uncertainty.

North America trails Asia-Pacific in volume but leads in technology adoption and regulatory impetus. California’s Title 24 and the United States federal push for district-energy coupling obligate new builds to meet PUE below 1.18 within the decade. Hyperscale operators pre-empt compliance risk through aggressive two-phase pilots in Virginia and Arizona. Canada’s colder climate further sweetens the server heat-export narrative, with Quebec utilities offering tariff rebates for waste-heat recovery connections, compressing payback to 24 months in some metro sites.

Europe presents a mosaic of drivers that Germany enforces statutory PUE caps, France ties grid fees to efficiency scores, and the Nordics monetize heat reuse within well-established district networks. Combined, these policies accelerate adoption schedules and generate attractive secondary revenue from heat sales. However, European operators must simultaneously pivot away from high-GWP coolants, compelling suppliers to certify hydrocarbon or ester formulations before facility-commissioning deadlines. This dual pressure fosters a vibrant vendor landscape comprising transformer oil specialists, chemical majors, and newly formed fluid recyclers.

Competitive Landscape

The top five suppliers accounted for roughly 45% of revenue in 2025, indicating a moderately concentrated graphics processing unit (GPU) liquid cooling market. Vertiv, Schneider Electric, and CoolIT anchor the CDU and manifold niche, while Asetek and Boyd lead silicon-level cold plates through OEM design wins with Dell, HPE, and Lenovo. Immersion specialists Submer, GRC, and LiquidStack differentiate on rapid deployment and high heat-reuse factors, yet face margin compression from fluid reformulation costs. Consolidation accelerated when Eaton bought Boyd for USD 9.50 billion, and Ecolab purchased CoolIT for USD 4.75 billion, forming vertically integrated stacks that mix chemistry, mechanics, and long-term service contracts.

Hyperscalers are meanwhile drafting open interface specifications that erode historical vendor lock-in. The Open Compute Project’s 2025 liquid spec unified manifold dimensions and sensor protocols, enabling second-tier suppliers to compete on price rather than proprietary fittings. Patent filings for micro-channel evaporators jumped 60% in 2025, led by ZutaCore and Accelsius, signaling the next battlefront for chip-level performance. Edge AI remains a white-space segment where Iceotope’s sealed KUL AI modules promise plug-and-play installation in remote telecom huts, potentially expanding total addressable units by hundreds of thousands over five years.

Finally, chemical giants Shell and Chemours are accelerating efforts to supply PFAS-free fluids that can meet performance requirements while maintaining competitive viscosity levels. At the same time, logistics firms in Singapore, Malaysia, and Thailand are positioning themselves to support OEMs by delivering assembled CDUs just in time, helping reduce lead times and improve deployment efficiency. Together, these trends indicate that while the market continues to consolidate around mechanical infrastructure, opportunities in fluid chemistry and edge deployment formats will remain open to a diverse group of entrants, supporting continued innovation across the ecosystem.

Recent Industry Developments

• April 2026: Iceotope Technologies surpassed 200 granted patents in precision immersion, opening a 15,000-square-foot research laboratory in Sheffield to advance two-phase micro-channel cooling for GPUs above 1,200 watts.
• April 2026: Carrier Global increased its stake in ZutaCore to 25% and committed USD 30 million for HyperCool production aimed at 200 kW-plus racks.
• March 2026: Ecolab closed its USD 4.75 billion acquisition of CoolIT Systems, pairing coolant chemistry with cold-plate mechanics for end-to-end solutions.
• January 2026: Vertiv launched the Liebert PCW phase-change platform, achieving 98% chip-level heat capture on NVIDIA GB200 racks and enabling 250 kW densities.

List of Companies Covered in this Report:

• Vertiv Group Corporation
• Schneider Electric SE
• CoolIT Systems Inc.
• LiquidStack Holdings Inc.
• Submer Technologies SL
• Green Revolution Cooling Inc.
• Asetek AS
• Iceotope Technologies Ltd.
• Fujitsu Ltd.
• Dell Technologies Inc.
• Hewlett Packard Enterprise Company
• Lenovo Group Ltd.
• Super Micro Computer Inc.
• BOYD Corporation
• Delta Electronics Inc.
• Parker Hannifin Corporation
• CPC Colder Products Company
• Danfoss AS
• Staubli International AG
• Zutacore Ltd.
• LiquidCool Solutions Inc.
• Asperitas BV

Additional Benefits:

• The market estimate (ME) sheet in Excel format
• 3 months of analyst support

Table of Contents

1 INTRODUCTION
1.1 Study Assumptions and Market Definition
1.2 Scope of the Study

2 RESEARCH METHODOLOGY

3 EXECUTIVE SUMMARY

4 MARKET LANDSCAPE
4.1 Market Overview
4.2 Market Drivers
4.2.1 Hyperscale GPU Rack Densities Exceeding 120 kW Accelerate Liquid Cooling Adoption
4.2.2 OEM Launch of Liquid-Cooled GPU Platforms NVIDIA AMD Supermicro
4.2.3 Stricter Data-Centre PUE Mandates in North America and Europe
4.2.4 Heat Re-use Incentives in District Energy Networks
4.2.5 Next-Generation AI ASICs Above 1 kW TDP Require Two-Phase Solutions
4.2.6 Supply-Chain Localisation of CDUs and Quick Disconnects in Southeast Asia
4.3 Market Restraints
4.3.1 High CAPEX Premium for Immersion Systems
4.3.2 Fluorinated Coolant Environmental Regulations
4.3.3 Limited Field Expertise for Edge AI Retrofits
4.3.4 Long Qualification Cycles for Government HPC Facilities
4.4 Industry Value Chain Analysis
4.5 Regulatory Landscape
4.6 Technological Outlook
4.7 Impact of Macroeconomic Factors on the Market
4.8 Porter’s Five Forces Analysis
4.8.1 Threat of New Entrants
4.8.2 Bargaining Power of Buyers
4.8.3 Bargaining Power of Suppliers
4.8.4 Threat of Substitutes
4.8.5 Competitive Rivalry

5 MARKET SIZE AND GROWTH FORECASTS (VALUE)
5.1 By Cooling Type
5.1.1 Single-Phase Liquid Cooling
5.1.2 Two-Phase Liquid Cooling
5.2 By Cooling Level
5.2.1 Component-Level Cooling
5.2.2 Server Rack-Level Cooling
5.3 By Deployment
5.3.1 Hyperscale Cloud
5.3.2 Enterprise
5.3.3 Government and Research HPC
5.3.4 Edge AI
5.4 By GPU Power Density
5.4.1 Below 300 W
5.4.2 300 W – 700 W
5.4.3 Above 700 W
5.5 By Geography
5.5.1 North America
5.5.1.1 United States
5.5.1.2 Canada
5.5.1.3 Mexico
5.5.2 Europe
5.5.2.1 Germany
5.5.2.2 United Kingdom
5.5.2.3 France
5.5.2.4 Italy
5.5.2.5 Rest of Europe
5.5.3 Asia-Pacific
5.5.3.1 China
5.5.3.2 Japan
5.5.3.3 South Korea
5.5.3.4 India
5.5.3.5 Southeast Asia
5.5.3.6 Rest of Asia-Pacific
5.5.4 South America
5.5.4.1 Brazil
5.5.4.2 Rest of South America
5.5.5 Middle East and Africa

6 COMPETITIVE LANDSCAPE
6.1 Market Concentration
6.2 Strategic Moves
6.3 Market Share Analysis
6.4 Company Profiles
6.4.1 Vertiv Group Corporation
6.4.2 Schneider Electric SE
6.4.3 CoolIT Systems Inc.
6.4.4 LiquidStack Holdings Inc.
6.4.5 Submer Technologies SL
6.4.6 Green Revolution Cooling Inc.
6.4.7 Asetek AS
6.4.8 Iceotope Technologies Ltd.
6.4.9 Fujitsu Ltd.
6.4.10 Dell Technologies Inc.
6.4.11 Hewlett Packard Enterprise Company
6.4.12 Lenovo Group Ltd.
6.4.13 Super Micro Computer Inc.
6.4.14 BOYD Corporation
6.4.15 Delta Electronics Inc.
6.4.16 Parker Hannifin Corporation
6.4.17 CPC Colder Products Company
6.4.18 Danfoss AS
6.4.19 Staubli International AG
6.4.20 Zutacore Ltd.
6.4.21 LiquidCool Solutions Inc.
6.4.22 Asperitas BV

7 MARKET OPPORTUNITIES AND FUTURE OUTLOOK
7.1 White-Space and Unmet-Need Assessment

---

---