Data center cooling demand is forecast to roughly quadruple in the 2026-2033 window, with the global market rising from $18.20 billion in 2026 to $43.13 billion by 2033 at a 13.1% CAGR, while the narrower hyperscale cooling segment grows from $9.8 billion in 2025 to a projected $60.7 billion by 2034 at a 22.5% CAGR [S1][S4]. The widening gap between air and liquid cooling adoption rates is the single most actionable signal for specifiers through 2030.
Four independent 2026 forecasts converge on the same direction but disagree on absolute size: Persistence Market Research sizes the market at $13.6 billion in 2026 rising to $46.3 billion by 2033 (19.2% CAGR) [S3], Coherent Market Insights at $18.20 billion to $43.13 billion (13.1% CAGR) [S4], Market Intelo at $9.8 billion to $60.7 billion for the hyperscale slice alone (22.5% CAGR) [S1], and MarketsandMarkets' dedicated chiller subset at $1.67 billion to $2.81 billion (9.0% CAGR) [S2].
Air vs Liquid vs Hybrid: Where Each Technology Stands in 2026
Air cooling still leads the installed base with over 53% technology share in 2026, held in place by cost efficiency and brownfield compatibility with legacy raised-floor rooms [S3]. Room-based cooling as a deployment type holds 62.2% of the 2026 market, reflecting the maturity of CRAC and CRAH architectures in enterprise and colocation sites [S4].
Liquid cooling is the fastest-growing technology at 26.4% CAGR, with the segment projected to climb from $5.7 billion in 2026 to $29.2 billion by 2033, and direct liquid cooling (DLC) alone held 38.4% of hyperscale technology share in 2025 [S1][S3]. Within DLC the 700-3,000 RT (approximately 2.46-10.55 MW) capacity band is forecast to dominate chiller procurement through 2032, and water-cooled chillers are projected to retain the largest product share while liquid-coolant chillers post the highest product-level CAGR at 20.0% [S2].
Rack Density Is the Real Driver Behind the Forecast
AI GPU clusters are pushing rack power densities to 50-100 kW, with NVIDIA H100 and H200 accelerators each generating up to 700 W of thermal load and next-generation GPU-based AI servers exceeding 1,200 W per accelerator [S1][S3]. Conventional CRAC architectures are now inadequate for these densities, which is why an estimated 68% of newly commissioned hyperscale facilities globally are specifying liquid cooling in their 2025 designs, up from fewer than 20% five years earlier [S1].
Hyperscale operators are responding with a 10 MW-class centrifugal chiller test unit shipped by Mitsubishi Heavy Industries to the Port of Brunswick, Georgia, in July 2026, rated at approximately 34.1 million BTU/h and intended to support high-density AI workloads [S2]. For instrument specifiers, this density shift directly increases the load on secondary-loop flow meters and plant-side pressure transmitters at every chilled-water branch, since flow accuracy now determines whether a 10 MW chiller meets its part-load efficiency target.
Regional Demand Map: North America Holds, Asia Pacific Grows Fastest
North America led 2025-2026 with a 40.2% share of hyperscale cooling revenue ($3.94 billion) and a 36-38.3% share of the broader cooling market, anchored by hyperscale campuses in Virginia, Oregon, Iowa, and Texas housing AWS, Microsoft Azure, Google Cloud, and Meta [S1][S3][S4]. Microsoft, AWS, Google, and Meta collectively committed over $200 billion in data center capex for 2025 alone, which is the upstream signal that pulls cooling orders through 2027 [S1].
Asia Pacific is the fastest-growing region across every report, with a projected 25.8% CAGR for hyperscale cooling through 2034 and a 24.6% CAGR for the broader cooling market through 2033, plus an 11.5% CAGR for chillers specifically [S1][S2][S3]. Growth is concentrated in China, India, and Southeast Asia, where humid tropical climates are forcing region-specific liquid and hybrid designs rather than direct lifts of North American templates [S4].
Standards, Sustainability, and the PUE Below 1.2 Push
The European Union Energy Efficiency Directive and corporate net-zero pledges targeting Power Usage Effectiveness (PUE) below 1.2 are forcing the retirement of legacy air-cooling infrastructure in new builds [S1]. North America's pivot toward renewable energy-powered data centers is driving integration of cooling systems with onsite generation, which in turn pushes the data logger and building-automation layer into the cooling plant's control loop for verifiable efficiency reporting [S4].
Stringent environmental regulations and corporate ESG commitments are accelerating demand for low-carbon cooling technologies, including liquid cooling and renewable energy-integrated HVAC systems, and growth of environmentally certified products [S3][S4]. For procurement, the practical signal is that cooling systems without measurable PUE telemetry, often delivered through a PLC with a high-speed acquisition chain, are starting to be excluded from hyperscale RFPs.
Competitive Landscape and Sourcing Signals
Vertiv led the hyperscale cooling competitive landscape in 2025 on the strength of its broadest integrated cooling portfolio, while the chiller-specific market is led by Johnson Controls, Trane Technologies, Carrier, Daikin, and Vertiv [S1][S2]. Strategic collaborations between cooling vendors and semiconductor companies are accelerating innovation in liquid and AI-driven thermal management systems, which means the bill of materials on a 2027-2028 AI cooling skid will increasingly look like a hybrid of mechanical and compute-thermal suppliers [S3].
The underlying AI compute roadmap, including accelerators such as those mapped in this AI accelerator spec gate for 2026, sets the thermal load envelope that every downstream cooling order must serve.