[[Hyperscalers]] are deploying $300B+ in 2025 capex (median 53% increase year-over-year), forcing a fundamental shift from air to liquid cooling as AI chips exceed 1000W thermal design points. The physics are unforgiving. Air cooling tops out at 1000W [[Thermal Design Power - TDP]]. [[Direct-to-Chip Cooling]] handles 1000-1600W. Beyond 1600W, only [[Two-Phase Cooling]] or [[Immersion Cooling]] works. Nvidia's latest chips are pushing these boundaries, making liquid cooling mandatory rather than optional. Two dominant approaches emerged. Immersion submerges entire servers in dielectric fluid, capturing all waste heat uniformly but requiring specialized equipment. DTC targets individual processors with cold plates, integrating with existing rack infrastructure for faster deployment and easier maintenance. The market validated both paths. Over $1B in venture capital flowed into cooling startups since 2023. Companies like Submer (immersion, $131M raised) and Iceotope (DTC, $78M raised) are scaling production. Strategic acquirers paid 15-25x revenue multiples. Schneider bought Motivair for $1.1B at 25x revenue, KKR took CoolIT at 23x. Liquid cooling penetration in AI servers jumps from 36% in 2024 to 59% by 2029. HPC follows similar trajectory, 40% to 62%. The transition mirrors historical infrastructure shifts: once physical limits force change, adoption accelerates regardless of operational inertia. DTC wins on serviceability and deployment speed. Immersion wins on total energy efficiency and heat capture. The answer depends on power density requirements and existing infrastructure constraints. Most hyperscalers are deploying both, optimizing by workload and facility. Capital flows to atoms, not bits. Data centers become thermal management companies that happen to run compute. The constraint shifted from silicon performance to heat removal. This connects to broader trends in [[Cooling and Energy Efficiency Technologies]] and [[Power Usage Effectiveness - PUE]]. The infrastructure build-out also touches [[Geothermal Energy and the Future of Data Centers]] for waste heat recovery and [[SMRs x Data Center Opportunities]] for reliable baseload power to support these massive deployments. --- #deeptech #datacenters #AI #infrastructure