Data Center Cooling: Why 120 kW AI Racks Are Rewriting a USD 72.02 Billion Market
Updated: 4 days ago
Executive summary. For decades, data centre cooling was a facilities problem solved with more air: raised floors, chillers and rows of air handlers pushing cold air past servers. AI has broken that model. A single rack of modern AI accelerators can draw many times the power of a traditional server rack, and air simply cannot carry that much heat away. Liquid has moved from niche supercomputers to the mainstream of new data centre design. Strategic Revenue Insights estimates the global data center cooling market at USD 18.85 billion in 2025, rising to USD 72.02 billion by 2035 at a 14.30 percent compound annual growth rate. Demand for energy-efficient cooling, the expansion of high-density computing, cloud growth and sustainability rules are driving it. This article sets out five lessons from the cooling suppliers, data centre operators and server makers moving first, with takeaways for strategy, product, operations, channels and revenue.
The structure of the market shows where value is moving. Strategic Revenue Insights' Data Center Cooling Market research finds that air conditioners remain the largest product segment while liquid cooling is the fastest growing, that air is still the most widely adopted technology while direct liquid and immersion cooling are gaining traction, and that row and rack cooling techniques are gaining ground on room-level cooling. Asia Pacific is the fastest growing region. The research names the high initial cost of advanced cooling, rapid technology change, pricing pressure and supply chain dependence as challenges, and lists Vertiv, Schneider Electric, STULZ, Rittal, Johnson Controls, Daikin, Trane Technologies, Carrier, Airedale and Munters among the key companies, with CoolIT Systems and Motivair also notable.
AI hardware has set a new baseline for heat. NVIDIA's GB200 NVL72 rack is designed for around 120 kilowatts and is cooled by direct-to-chip liquid, according to NVIDIA and partner specifications, many times the power of a conventional enterprise rack. Cooling suppliers are seeing the demand: Vertiv reported organic orders up about 60 percent year on year in the third quarter of 2025 and a backlog of USD 9.5 billion, tying the momentum to AI infrastructure.
Why the Data Center Cooling Market Is Entering a New Phase
Three structural forces are converging.
Rack density has jumped. AI training and inference clusters concentrate huge power in a small space. SRI's research names the expansion of high-density computing as a key driver and identifies liquid cooling as the fastest growing product.
Energy and water are now constraints. Cooling can account for a large share of a data centre's non-IT energy use, and many regions face limits on power and water. SRI's research names energy efficiency and sustainability regulation as drivers.
Speed to capacity matters. Operators racing to bring AI capacity online need cooling designs that can be built quickly, standardised and repeated across campuses.
These forces explain SRI's forecast of strong growth to 2035, with value concentrating in liquid cooling, rack-level systems and Asia Pacific.
Five Lessons From the Companies Moving First
The suppliers and operators turning these forces into advantage share a recognisable pattern. We have distilled it into five lessons, each followed by a takeaway for a specific leadership function.
Lesson 1: Design for Hybrid Air and Liquid
Liquid cooling is the fastest growing segment, but air conditioners remain the largest in SRI's research. Most data centres will run both for years: liquid for AI racks and air for networking, storage and conventional servers. Even liquid-cooled AI racks still reject part of their heat to air.
The operators moving first design halls that can switch between densities, with liquid distribution ready to install as AI demand arrives, rather than building separate facilities for each.
"The future data centre is not air or liquid. It is a building that can be both, rack by rack, as the workload changes."
Takeaway on strategy. Plan facilities for hybrid air and liquid cooling, with liquid-ready infrastructure that can be activated as AI racks arrive.
Lesson 2: Direct-to-Chip First, Immersion Where It Fits
SRI's research identifies direct liquid and immersion cooling as the technologies gaining traction. Direct-to-chip cooling, where cold plates sit on processors and coolant flows through the rack, is the route chosen by the leading AI hardware platforms. Immersion, where servers sit in tanks of dielectric fluid, suits specific high-density and edge uses.
The companies moving first build complete liquid portfolios: coolant distribution units, cold plates, manifolds, heat rejection and monitoring. The bandwidth and power of AI clusters, described in why AI clusters drove a USD 23.8 billion optics year, are what make these systems necessary.
Takeaway on product. Offer complete direct-to-chip systems, from cold plates to coolant distribution and heat rejection, and add immersion for the uses where it fits best.
Lesson 3: Prefabricate and Standardise
AI capacity is being built at a pace that traditional site-built mechanical systems cannot match. The suppliers moving first deliver prefabricated cooling modules, skids and power and cooling blocks that arrive tested, connect quickly and repeat across sites. SRI's research lists rapid technology change and supply chain dependence as challenges, which standardised designs help manage.
Close work with chip and server makers on reference designs means cooling is ready when new hardware ships.
"In the AI build-out, the cooling that wins is the cooling that arrives on a truck, tested and ready to connect."
Takeaway on operations. Standardise cooling into prefabricated, tested modules, co-develop reference designs with chip and server makers, and secure component supply early.
Lesson 4: Sell Lifecycle Service, Not Just Equipment
Liquid cooling adds pumps, fluids, filters and leak detection to the data centre, and operators need confidence that systems will run continuously. The companies moving first sell installation, commissioning, monitoring, fluid management and maintenance as long-term service contracts, and offer performance guarantees.
Hyperscale cloud providers, colocation companies and AI cloud specialists are the core customers. The rise of AI agents and inference workloads, covered in why the AI agents scaling gap decides who wins, will keep demand for dense, reliable capacity growing.
Takeaway on channels. Bundle equipment with commissioning, monitoring and maintenance contracts, and build deep relationships with hyperscale, colocation and AI cloud operators.
Lesson 5: Asia Pacific Grows Fastest, Efficiency Rules Tighten Everywhere
SRI's research identifies Asia Pacific as the fastest growing region, driven by data centre expansion in India, Southeast Asia, Japan and Australia, often in hot and humid climates where cooling efficiency matters even more. Across regions, rules on energy efficiency, water use and heat reuse are tightening, and SRI's research names sustainability regulation as a driver.
The capabilities that decide who captures this demand are specific:
Hybrid air and liquid facility designs.
Complete direct-to-chip portfolios with immersion options.
Prefabricated, standardised cooling modules.
Lifecycle service and performance guarantees.
Efficiency, water and heat reuse capability for each region.
Companies evaluating these competitive dynamics in greater detail can use SRI's Data Center Cooling Market analysis to assess product, technique and technology segments and the positioning of the leading suppliers.
Takeaway on market development. Build local manufacturing and service capacity in Asia Pacific, and design for regional efficiency, water and heat reuse rules.
Where the Opportunity Is Concentrating
Value is moving toward liquid cooling, rack-level systems and service.
By product, air conditioners remain the largest segment while liquid cooling is the fastest growing.
By technique, row and rack cooling are gaining ground on room-level cooling.
By technology, air remains the most widely adopted, while direct liquid and immersion cooling are gaining traction.
By geography, Asia Pacific is the fastest growing region.
At the Horizon: What to Watch Over the Next Three Years
Rack power. Each new generation of AI hardware will raise rack power further and push more of the market to liquid.
Heat reuse and water. Rules and incentives for reusing waste heat and cutting water use will shape cooling designs.
Supply capacity. Coolant distribution units, cold plates and chillers in short supply will decide who can deliver on time.
"The next phase of data centre cooling will be decided by the suppliers who can deliver liquid-cooled capacity as fast as chips are shipped."
The Strategic Revenue Opportunity
For cooling suppliers, the task is to scale liquid portfolios and service. For data centre operators, the opportunity is faster, denser, more efficient capacity. For chip and server makers, the priority is cooling that is ready for each new platform. For investors, the signal is orders, backlog and service revenue.
Market intelligence turns these questions into decisions. Segment-level sizing shows where product and technology demand is concentrating, competitive analysis shows how suppliers are positioned, and regional analysis shows where data centre build-outs will pull growth first. This is the path from market signals to strategic insight to revenue impact.
Takeaway on revenue. Treat the data center cooling market as four revenue pools: air systems sold for conventional capacity, liquid systems sold for AI racks, prefabricated modules sold on speed, and lifecycle service sold by contract. Each needs its own pricing model.
The Next Leg: Preparing for the Journey Ahead
The defining shift in data centre cooling is that it is moving from air-based facilities engineering to liquid-based infrastructure tied to every generation of AI hardware. That changes what is bought, how fast it must arrive and who supplies it. The companies moving first are not the ones with the largest catalogues; they are the ones designing for hybrid cooling, leading on direct-to-chip, prefabricating, selling service and meeting efficiency rules.
The next three years will decide which suppliers become the default thermal partners for AI data centres. Decision-makers should invest in liquid capability, standardisation and service and track rack power, heat reuse rules and supply capacity. Companies evaluating growth opportunities in this market can use Strategic Revenue Insights' Data Center Cooling Market report to assess market size, competitive dynamics, regional opportunities and strategic growth priorities.
Sources: Strategic Revenue Insights, Data Center Cooling Market report (2026); NVIDIA and partner specifications; Vertiv Q3 2025 results.
Frequently Asked Questions
How big is the data center cooling market?
Strategic Revenue Insights estimates the global data center cooling market at USD 18.85 billion in 2025, rising to USD 72.02 billion by 2035 at a 14.30 percent compound annual growth rate. Air conditioners are the largest product segment, and Asia Pacific is the fastest growing region.
What is driving growth in data center cooling?
Demand for energy-efficient and sustainable cooling, the expansion of data centres and high-density AI computing, cloud growth and digital transformation, and regulation that promotes green technologies are the main drivers. AI racks drawing around 120 kilowatts are pushing operators toward liquid cooling.
Why is liquid cooling growing so fast in data centers?
Liquid cooling is the fastest growing segment, according to Strategic Revenue Insights, because AI accelerators produce more heat than air can remove efficiently. Direct-to-chip cooling carries heat away from processors with coolant, while immersion cooling places servers in dielectric fluid for very dense uses.
What are the main challenges in data center cooling?
The main challenges are the high initial cost of advanced cooling, especially for smaller operators, rapid technology change that requires continuous innovation, rising competition and pricing pressure, and reliance on global supply chains that expose suppliers to disruption and geopolitical risk.
About the authors: This article was prepared by the Strategic Revenue Insights research team, which combines market intelligence, competitive analysis and strategic advisory for growth-stage, mid-market and enterprise organisations across the ICT, semiconductor and electronics sector. Sector analyst: Lukas Schmidt, Technology and Infrastructure Analyst. Our market estimates combine primary research (interviews, surveys and expert panels) with secondary data (industry reports, databases and public disclosures); read more about how Strategic Revenue Insights works. To discuss this market with our analysts, contact our team.
Comments