Data Center Energy Analyst

Water and Cooling as Emerging Constraints

Water and Cooling as Emerging Constraints

Key Questions

Why is liquid cooling becoming mandatory for AI data centers?

Direct-to-chip liquid cooling is required to handle high-density AI workloads, with CoolIT's 15kW cold plates and NVIDIA's fanless systems leading adoption. Liquid cooling CAGR is projected at 18.2% through 2035 amid water constraints.

What companies are benefiting from the cooling technology boom?

ZutaCore raised $100M at $600M valuation; Modine saw 158% data center revenue growth with $4B reservations; and nVent reported $2.3B backlog. Vertiv, AAON, and EnerSys are highlighted as key cooling plays.

How are water constraints affecting data center projects globally?

Severe water shortages have impacted India's Visakhapatnam project and Malaysia's Johor hub, prompting Meta to use desalinated seawater cooling in India. UN reports now emphasize water and land impacts beyond carbon emissions.

What innovations address water use in data center cooling?

Microsoft claims near-zero water cooling while Ferveret’s immersion technology offers zero-water, 15% better efficiency. Texas now mandates closed-loop systems, and Amazon reports a 42% drop in water use to 0.12 L/kWh.

How are material and supply chain issues impacting cooling adoption?

Liquid cooling manifolds face steel-vs-plastic choices with a $3B market doubling; ASHRAE is evaluating polypropylene. NVIDIA’s reference design stops at facility walls, leaving external water and infrastructure challenges unresolved.

ZutaCore $100M round at $600M valuation with strategic investors confirmed. Modine 158% data center revenue jump, $4B capacity reservation. CoolIT 15kW coldplate extends single-phase DLC roadmap. Direct-to-chip liquid cooling mandatory. nVent Electric $1B sales, $2.3B backlog, Siemens/NVIDIA partnership, liquid cooling CAGR 18.2% through 2035. Xing immersion-cooled BBU. Cadence chip-to-chiller simulation. Zacks highlights AAON, Comfort Systems, EnerSys, Vertiv as cooling plays. India's Visakhapatnam project and Malaysia's Johor highlight severe water constraints. UN report reframes AI's environmental impact beyond carbon to water and land—inference dominance (80-90% of energy) shifts focus to runtime efficiency and cooling choices. Liquid cooling materials become a supply chain concern—steel vs plastic manifolds, $3B market doubling, ASHRAE stance on polypropylene. Microsoft claims near-zero water cooling. Georgia Fayette County water usage (29M gallons during construction). Meta's India AI data center uses desalinated seawater cooling. Ecolab introduces Cooling-as-a-Service. New: Ferveret introduces nuclear-inspired subcooled boiling immersion cooling claiming 15% better efficiency and 35% more tokens per watt, zero-water, low-PFAS, partnered with CleanSpark, FuriosaAI, Switch. Texas mandates closed-loop cooling for data centers. Amazon touts water efficiency gains (42% drop in NoVA, 0.12 L/kWh vs 0.84 industry average). KAIST demonstrates 10x liquid cooling efficiency (COP 106,000) using plain water. Intelligent liquid cooling controls improve adaptive performance. NVIDIA introduces fully liquid-cooled system eliminating fans and nearly zeroing water use.

Sources (4)
Updated Jun 26, 2026
Why is liquid cooling becoming mandatory for AI data centers? - Data Center Energy Analyst | NBot | nbot.ai