The mechanism

The surge in artificial intelligence adoption, from large language models to complex scientific simulations, has dramatically increased power consumption and heat generation within data center racks. Modern GPUs and AI accelerators generate heat loads far exceeding the approximate 41.3 kW per rack limit of traditional air cooling, making liquid cooling a fundamental requirement for the entire AI data center stack. Next-generation AI racks may reach 370kW densities in 2026, a challenge air-based systems cannot handle. This demand is driving a structural shift in how data centers are designed, built, and operated. Direct-to-chip (DTC) cooling, where liquid coolant flows through cold plates mounted directly on high-heat-generating components, is the prevalent solution. Single-phase direct-to-chip cooling, often using water-glycol mixtures, currently holds an estimated 2026 market share of 55% for direct-to-chip solutions, favored for its ease of deployment and lower cost. However, as chip thermal design power (TDP) continues to climb beyond 1500W and approaches 2000W, two-phase direct-to-chip cooling is expected to become essential, with large-scale adoption projected around 2026-2027. Two-phase systems, while more complex and costly upfront, offer higher heat transfer rates by leveraging phase change at the chip level. Immersion cooling, which submerges entire servers in a dielectric fluid, is also gaining traction, projected to grow at a 26.62% CAGR through 2031. The industry is grappling with supply chain bottlenecks, with lead times for Coolant Distribution Units (CDUs) extending to 20-26 weeks as of early 2026, posing risks to deployment timelines. Furthermore, the environmental impact, particularly water usage, is prompting a shift towards more sophisticated closed-loop cooling technologies and higher liquid-cooling system temperatures to reduce reliance on water-intensive evaporative methods.

Who is exposed

Companies providing the foundational infrastructure for AI data centers are directly exposed to these evolving thermal demands. Vertiv (VRT), a global leader in critical digital infrastructure, is deeply embedded in this trend. Its recent acquisitions of King Environmental Services and PurgeRite highlight a strategic push into end-to-end fluid management for high-density, liquid-cooled data centers. Vertiv’s collaboration with NVIDIA on high-density reference architectures, including GB200 and GB300 systems, confirms its pivotal role in the direct-to-chip liquid cooling shift. Super Micro Computer (SMCI) is another key beneficiary, positioning itself as a leader in delivering AI-optimized servers with integrated liquid cooling. Its recent shipment of NVIDIA Vera Rubin NVL72 racks with direct liquid cooling underscores its capacity to meet the demands of advanced AI platforms. With AI GPU-related platforms contributing over 80% of its Q3 FY 2026 revenue, Supermicro's performance is tightly linked to the continued buildout of AI infrastructure. Eaton (ETN), an intelligent power management company, is leveraging its expertise from grid to chip. Its $9.5 billion acquisition of Boyd Thermal, a leader in thermal components and liquid cooling solutions, significantly enhances its integrated offerings for AI data centers. Eaton's collaboration with NVIDIA on the Beam Rubin DSX platform further solidifies its position in providing comprehensive power and cooling blueprints for AI factories. These companies are not just selling components; they are providing integrated solutions that are becoming non-negotiable for high-performance AI deployments, tying their growth prospects directly to the continued expansion and densification of AI computing.

Quantitative Outlook

The market data reflects the ongoing transition and the significant investments being made in AI data center infrastructure. Vertiv (VRT) shares, despite a 16.65% dip over the last 90 days, have seen robust 1-year growth of 79.03%, indicating strong underlying demand for its solutions. Its raised full-year 2026 guidance, including