Liquid cooling goes mainstream as AI power limits bite

The liquid cooling market for data centers is on a steep growth trajectory, with the direct-to-chip cooling segment expected to surge from $3.33 billion in 2026 to $17.31 billion by 2032, reflecting a robust CAGR of 26.5%. Complementing this, the broader data center heat exchangers and chillers markets are also expanding rapidly, projected to reach $14.40 billion and $2.81 billion respectively by 2032, underscoring the escalating demand for advanced thermal management solutions as data centers evolve.

Regionally, North America dominates the liquid cooling landscape with a commanding 45% market share, driven by hyperscale data center operators investing heavily to support AI, machine learning, and other data-intensive workloads. Asia Pacific and Europe follow with 25% and 20% shares respectively, reflecting significant regional adoption fueled by expanding cloud infrastructure and government initiatives. This geographic distribution highlights a global embrace of liquid cooling technologies, with hyperscale players at the forefront of this transformation.

Within the technology mix, single-phase direct-to-chip cooling systems paired with water-glycol-based coolants are leading the market due to their proven reliability and seamless integration with existing data center infrastructures. This preference for mature, compatible solutions facilitates faster adoption among major cloud and tech companies, enabling them to efficiently manage the thermal demands of high-density AI and HPC workloads without disruptive overhauls.

The explosive growth of AI and high-performance computing (HPC) workloads is fundamentally reshaping data center infrastructure, driving a shift from traditional air cooling to advanced liquid cooling solutions such as direct-to-chip and immersion cooling. Rack densities have surged from around 10 kW per rack in the CPU era to over 100 kW with current AI hardware like GB200 and GB300, with next-generation systems like Fire Rubin expected to push densities beyond 200 kW, making air cooling insufficient and liquid cooling indispensable. This transition is exemplified by BDC’s pioneering efforts in South Asia, where since 2022 they have operated 90 megawatts of direct-to-chip cooled capacity alongside 10 megawatts of immersion cooling, demonstrating liquid cooling’s critical role in managing the thermal demands of modern AI workloads.

The rapid escalation in AI workloads, particularly the shift from training to inference which is projected to constitute 80% of AI IT load by 2030, is intensifying power and density requirements in data centers worldwide. This surge is driving data center operators to scale from tens of megawatts to gigawatt-level campuses, with global AI infrastructure demand expected to reach nearly 90 GW by 2030—outstripping supply by over 500%. Moreover, inference workloads’ need for low-latency service delivery is pushing data center deployments closer to urban centers, where higher power densities and advanced liquid cooling technologies like direct-to-chip and immersion cooling are essential to efficiently handle the thermal and spatial constraints.

The complexity and power intensity of AI-driven data centers are also accelerating the adoption of sophisticated power management and thermal control technologies. The power stage segment, crucial for efficient power delivery at high currents and low voltages, is forecasted to dominate the data center semiconductor market by 2029, while multi-channel ADC/DAC solutions are gaining traction for their ability to provide precise real-time monitoring and control of power and cooling systems. As Linglan Wang from Gartner highlights, 'AI capacity is now constrained by power availability,' underscoring the critical need for high-efficiency liquid cooling systems to ensure sustainable and scalable growth amid soaring electricity consumption, with AI-optimized servers alone expected to consume 175 TWh in 2026, an 84.2% increase from the previous year.

Europe is channeling a massive €6.8 billion into water infrastructure for data centers by 2036, with Germany, the U.K., and France leading the charge by accounting for over 40% of this spend. This investment surge is propelled not only by the AI boom driving soaring thermal loads but also by stringent EU regulations that now treat water availability as critical as power access. Countries like Spain, Italy, Poland, and the Nordic nations are tailoring their approaches—ranging from water-resilient development to grid modernization—reflecting a nuanced regional strategy to embed liquid cooling as a core solution amid rising water stress and energy constraints.

In the Asia-Pacific region, government-led initiatives and strategic investments are turbocharging the expansion of AI-ready data centers, with liquid cooling technologies at the forefront. South Korea’s collaboration with OpenAI and its $30 million ASEAN digital transformation project, alongside Taiwan’s $3.2 million AI infrastructure fund aiming for global leadership by 2040, exemplify this momentum. Meanwhile, colossal investments like Digital Edge’s $4.5 billion AI-centric facility in Jakarta and Equinix’s $201 million projects in Malaysia underscore how public-private synergy is reshaping the region’s data center landscape toward sustainable, liquid-cooled architectures.

Sustainability is a linchpin in APAC’s data center evolution, with governments promoting green infrastructure through incentives and regulatory bodies such as Malaysia’s Data Center Task Force. Operators are integrating low-carbon power solutions including lithium-ion batteries and biofuel generators, highlighted by Microsoft’s renewable generator deployment in South Korea. Complementing this, infrastructure projects like Japan’s Society 5.0 smart city initiative and Google’s TalayLink submarine cable between Australia and Thailand are enhancing connectivity and supporting the scalable adoption of liquid cooling technologies across key hubs like Beijing, Sydney, Tokyo, Singapore, and Indonesia.