AI data center spend surge fuels liquid cooling

The global data center market is undergoing a dramatic expansion fueled by the surging demand for AI workloads and hyperscale infrastructure. Projections indicate the OCP rack market alone will more than double from $2.02 billion in 2026 to $4.32 billion by 2030, driven by AI training and inference applications that command the largest market share. This growth is mirrored in the broader data center investment landscape, with Arizton forecasting a near doubling of global investments from $514.26 billion in 2025 to $959.19 billion by 2031, propelled by hyperscale giants like AWS, Google, Meta, and Microsoft aggressively scaling AI-ready capacity worldwide.

Regionally, North America remains the dominant hub for data center investment and capacity growth, supported by its mature digital infrastructure and a strong emphasis on renewable energy integration. The North American data center construction market is expected to nearly double its investment from $85.23 billion in 2025 to $158.20 billion by 2031, with hyperscale operators such as Microsoft unveiling AI-focused campuses like the Fairwater facility housing hundreds of thousands of NVIDIA GPUs. Meanwhile, Asia Pacific is emerging as the fastest-growing region, led by rapid AI adoption in China and capacity expansions across Japan, South Korea, Singapore, and India, highlighting a global shift toward diversified regional growth hotspots.

The U.S. market exemplifies the AI-driven transformation with data center capacity projected to double over the next three years, supported by over 700 planned facilities adding 45 gigawatts of power. Hyperscale operators like AWS, Google, and Meta are spearheading this expansion, investing heavily in sustainable energy projects—including Meta's 850 MW solar PPAs—and advanced cooling technologies to meet the intense power demands of AI workloads. Simultaneously, the neocloud segment is scaling at an unprecedented pace, reaching $9 billion in Q4 2025 with a 223% year-over-year increase, underscoring its rising influence as a cost-efficient and power-optimized alternative that complements traditional hyperscale growth.

Industry players like Coresight are capitalizing on the AI boom by targeting high-profile customers and investing over $700 million in 2026 to replenish sold data center capacity, driven by insatiable demand for AI inferencing and machine learning services that currently outpace their capacity. This selective investment strategy reflects a broader market trend where quality and specialization in AI infrastructure are becoming critical differentiators amid a rapidly expanding and competitive global data center ecosystem.

Japan's data center market is undergoing rapid expansion, with projections soaring from USD 12.76 billion in 2025 to nearly USD 39 billion by 2031, driven primarily by Tokyo's dominance alongside emerging hubs like Osaka, Hokkaido, and Kyushu that benefit from government incentives and lower costs. This growth is coupled with technological advancements such as pilot liquid cooling projects by Preferred Networks and Internet Initiative Japan, and strategic collaborations exemplified by SoftBank and Oracle's sovereign cloud joint venture launching in 2026. Sustainability is also gaining traction, highlighted by Microsoft's 100 MW solar power agreement in Kyushu and Chugoku, signaling a strong regional commitment to green energy integration.

Indonesia's data center landscape is entering a phase of 'hyper growth with specialization,' shifting focus from mere capacity expansion to tailored infrastructure that meets the high power density demands of AI workloads, with racks now supporting upwards of 100 kilowatts. Despite abundant land and power resources, the market grapples with challenges in securing reliable, large-scale green power due to transmission and distribution bottlenecks, compounded by a regional shortage of skilled talent for both construction and operations. This talent gap is prompting cross-border knowledge transfers within the Asia-Pacific, underscoring the complexity of scaling data center capabilities in emerging markets.

Europe's data center market is accelerating at a CAGR of nearly 23% through 2031, propelled by surging AI and HPC workloads that demand advanced liquid cooling technologies such as direct-to-chip and immersion cooling. Western Europe, particularly the UK and Nordic countries, leads in power capacity and investment, supported by proactive government policies targeting digitalization and carbon footprint reduction, while emerging markets like Spain and Poland attract growth due to strategic location and cost advantages. Industry giants including AWS, Google, and Spode Works Regeneration Ltd are expanding sustainable infrastructure portfolios, emphasizing renewable energy integration and efficient power management.

North America remains the largest and most mature data center market globally, with investments projected to nearly double from USD 85 billion in 2025 to over USD 158 billion by 2031, fueled by hyperscale cloud infrastructure growth and AI-driven demand. The U.S. leads this surge, with hotspots like Northern Virginia, Texas, California, and the Pacific Northwest attracting major expansions from Meta, Google, and Microsoft, the latter unveiling its Fairwater AI data center equipped for hundreds of thousands of NVIDIA GPUs. Regional incentives and renewable energy priorities, such as Meta's 850 MW solar power purchase agreement across Oklahoma, Texas, and Mississippi, are shaping development, while Texas is poised to become a sustainability beacon with AWS's planned nuclear-powered campus by 2026. Additionally, North America's connectivity is being enhanced through 27 upcoming subsea cable systems linking it to Asia-Pacific, Europe, and Latin America, reinforcing its global data center leadership.

By early 2026, the data center liquid cooling market was experiencing explosive growth, projected to surge from $870 million in 2024 to $10.7 billion by 2030 at a staggering CAGR of nearly 52%, driven primarily by escalating rack power densities and the insatiable demands of AI and HPC workloads. Hyperscale data centers, led by giants like Meta and Microsoft, spearheaded this adoption, leveraging liquid cooling to manage dense GPU clusters and pushing the boundaries of traditional air cooling limitations. This rapid expansion underscores liquid cooling's critical role in sustaining the next generation of AI infrastructure.

Direct-to-chip liquid cooling emerged as the dominant technology within this transformation, favored for its precision in delivering coolant directly to heat-generating components, thereby enhancing thermal management and extending hardware lifespan. Single-phase systems using water-glycol mixtures gained popularity due to their efficiency and seamless integration with existing data center infrastructure, effectively handling heat loads between 60 kW and 120 kW per rack. Market forecasts anticipate the direct-to-chip cooling segment to balloon to $17.31 billion by 2032, with hyperscale operators continuing to lead adoption amid challenges such as leakage risks and integration complexity.

Geographically, North America commands the largest share of the liquid cooling market, accounting for 45%, fueled by substantial investments in purpose-built AI campuses like Microsoft's Fairwater facility, which supports hundreds of thousands of NVIDIA GPUs with advanced liquid cooling systems. Meanwhile, regions such as China, Europe, the Middle East, and Latin America are rapidly advancing their liquid cooling capabilities, driven by greenfield projects, hot climates, and soaring electricity costs. This global diffusion is further accelerated by sustainability imperatives, with liquid cooling technologies significantly improving power usage effectiveness (PUE) to as low as 1.02–1.03 and reducing greenhouse gas emissions.

Despite its clear advantages, the liquid cooling revolution faces hurdles including high upfront capital expenditures, supply chain disruptions, and a lack of standardized industry protocols, which complicate integration and slow broader adoption. However, ongoing market consolidation and strategic partnerships among key players such as Vertiv, Johnson Controls, Schneider Electric, and innovative startups like CoolCentric and Iceotope are fostering technological advancements and expanding global reach. This dynamic landscape suggests that while challenges persist, the momentum behind liquid cooling as the backbone of AI-ready data centers is both robust and accelerating.

By mid-2026, the power infrastructure supporting AI-driven data centers is undergoing a transformative evolution marked by the integration of advanced battery technologies and intelligent power management. Google’s pioneering use of lithium-ion backup systems sets a high operational standard, while nickel-zinc batteries gain traction for their superior energy density, safety, and environmental benefits. This innovation wave is mirrored in the global data center power market, which Polaris Market Research projects will soar to $70.94 billion by 2034, fueled by AI, cloud, and hyperscale demands, with solutions like UPS systems and power distribution units dominating 70.6% of the market and AI-enabled power management optimizing energy use and predictive maintenance.

The race toward 1-megawatt racks is catalyzing a fundamental rethinking of data center architecture, as hyperscalers including Google, Meta, and Microsoft collaborate through the Open Compute Project’s Mount Diablo initiative to standardize cooling, power delivery, and rack design. Central to this shift is the adoption of ±400V DC power distribution, inspired by electric vehicle technologies, which promises to streamline megawatt-scale power delivery while reducing losses and cabling complexity. This evolution transforms racks from passive enclosures into integrated compute ecosystems, exemplified by Nvidia’s DGX platform, where power, cooling, interconnects, and management coalesce at the rack level to meet relentless AI workload demands.

Open Rack Wide (ORW) standards extend the Open Compute Project’s architecture to accommodate the heftier, more power-dense AI systems emerging in hyperscale environments, with AMD’s Helios AI rack architecture adopting Meta’s ORW specification submitted in 2025. This innovation aligns with a booming OCP rack market projected to grow from $1.68 billion in 2025 to $4.32 billion by 2030 at a 21% CAGR, driven by hyperscalers and wholesale colocation providers leveraging uniform, modular designs that integrate advanced liquid cooling and power delivery. North America leads this expansion, capitalizing on its concentration of hyperscale data centers and early embrace of 21-inch Open Rack designs that enhance thermal management and power density.

The industry’s pivot toward higher-tier (Tier III and IV) data centers underscores a commitment to fault tolerance and redundancy essential for critical AI workloads, while sustainability remains a core pillar with increasing integration of battery storage, renewables, and high-efficiency UPS equipment. This holistic approach balances the competing demands of reliability, efficiency, flexibility, and environmental stewardship, as AI-enabled power management systems further refine energy consumption and maintenance. Collectively, these innovations reflect a maturing ecosystem where power infrastructure and high-density rack designs evolve in tandem to meet the relentless scale and complexity of AI-driven data centers worldwide.

By early 2026, neocloud providers emerged as a dynamic force in the compute supply landscape, rapidly growing within the Open Compute Project (OCP) rack market with a projected CAGR of 34.3%. These agile players address critical compute shortages caused by the surging AI workload demand not by competing with hyperscalers’ massive 100 MW+ projects, but by aggregating smaller, scattered pockets of power capacity—such as 2 MW waste-to-energy plants—that larger operators typically overlook. This strategic focus on cost-efficient, power-optimized, and greenfield deployments enables neoclouds to complement traditional hyperscale data centers effectively, filling gaps in compute supply with nimble infrastructure solutions.

The neocloud ecosystem is notably diverse, comprising ex-crypto miners, real estate operators, and entrepreneurs pivoting from other sectors like Allbirds, all leveraging existing infrastructure or unique access to scarce inputs to supply flexible, GPU-centric compute capacity. Despite hyperscalers’ entrenched scale advantages, the unprecedented speed of AI demand growth and its specialized compute requirements create a fertile environment for these newcomers to thrive, potentially rewriting the traditional cloud market playbook.

By mid-2026, neocloud providers such as Nebius, Lambda, and CoreWeave had scaled impressively, with the neocloud market hitting $9 billion in Q4 2025 alone—an astounding 223% year-over-year increase according to Synergy Research Group. This surge not only complements the ongoing expansion of hyperscale data centers, which are set to double U.S. capacity to over 45 gigawatts with 700 new facilities in the pipeline, but also creates fresh channel opportunities. Partners like World Wide Technology are capitalizing on this trend by collaborating with neoclouds to deliver enterprise-grade GPU-as-a-Service offerings, meeting the specialized infrastructure demands of AI workloads amid local concerns over power availability and sustainability.

Sustainability imperatives have become a central driver in the evolution of data center infrastructure, with major hyperscale operators like Amazon Web Services and Meta committing to substantial renewable energy investments, including AWS's ambitious nuclear-powered campus in Texas and Meta's 850 MW solar PPAs across multiple states. This green push is not only reshaping investment decisions but also influencing site selection, as seen in the attractiveness of the Midwestern U.S. for its renewable resources and tax incentives, despite regulatory hurdles such as Ohio's bans. These dynamics underscore how environmental goals and regional policy frameworks are tightly interwoven in shaping the future data center landscape.

To meet the dual challenges of soaring AI-driven power demands and stringent sustainability targets, data center operators are rapidly adopting advanced cooling technologies and energy-efficient designs. Industry leaders like Meta and Microsoft are pioneering liquid and hybrid cooling systems, with Microsoft exploring microfluidics to enhance efficiency, while water conservation and low-carbon building practices are becoming standard to align with corporate net-zero commitments. This evolution reflects a strategic pivot where infrastructure must balance high-density computing needs with aggressive energy reduction goals.

Emerging power technologies and energy storage solutions are increasingly integral to sustainable data center operations, as exemplified by Google's integration of lithium-ion backup systems and the rising adoption of nickel-zinc batteries for their superior safety and lower environmental impact. The transition from optional upgrades to core infrastructure components is further evidenced by landmark deals like the 1 GW solar capacity agreements between TotalEnergies and Google, signaling a maturation of renewable energy and battery storage as foundational elements in data center power strategies.

Supply chain constraints and the complexity of modern data center environments are driving a trend toward larger, more scalable campuses equipped with advanced cooling and power systems to support next-generation workloads. This shift is fueling robust growth in specialized services—including design, integration, maintenance, and energy management—with the sector projected to expand at a CAGR of 14.9%. Operators are thus navigating a landscape where infrastructure scale, technological sophistication, and supply chain resilience are critical to sustaining rapid AI and cloud expansion while meeting sustainability mandates.