Nvidia Claims Data Center Water Crisis Solved with New Warm Liquid Cooling System
The AI compute build-out continues with major new datacenter projects and significant power deals, while efficiency solutions like advanced cooling are becoming key to scaling infrastructure.
The story
A notable development in AI infrastructure emerged this week as Nvidia announced a significant advancement in data center cooling. At London Climate Week on June 26, Nvidia revealed that its latest AI systems can be fully cooled using liquid warm enough to negate the need for additional chilling equipment.
Josh Parker, Nvidia's chief sustainability officer, stated that this innovation largely resolves the water consumption challenges facing data centers. This claim, if widely adopted, could reshape the environmental footprint and geographic viability of future AI factories.
Data centers currently consume a substantial amount of electricity and water, with global AI data center electricity demand projected to more than quadruple by 2030. Nvidia's new approach, which redesigns how these facilities operate beyond just faster chips, aims to remove a critical physical bottleneck that has concerned investors and environmental advocates. This shift towards more efficient, closed-loop liquid cooling systems could convert wasted power directly into revenue-generating AI computation, fueling the next wave of data center expansion.
The build-out
| Project | Who | Scale | Where |
|---|---|---|---|
| Pecos AI Data Center Campus | Microsoft, Chevron (Energy Forge One LLC) | ~2 GW IT capacity, scaling to 2.67 GW with 20-year power agreement | Pecos, Reeves County, Texas, US |
| Mount Pleasant AI Campus | Microsoft | USD 3.3 billion investment | Mount Pleasant, Wisconsin, US |
| US Federal AI Infrastructure Expansion | AWS | Up to USD 50 billion commitment | US Top Secret, Secret, and GovCloud (US) regions |
| Helix Digital Infrastructure | KKR, Kuwait Investment Authority, NVIDIA, Vistra | Over USD 10 billion in committed capital | Global (focused on integrated power supply AI campuses) |
| GTA AI Infrastructure Facility | Buzz HPC (Hive Digital Technologies subsidiary) | Approximately 320 MW capacity | Greater Toronto Area, Ontario, Canada |
Supply & policy signals
Nvidia's new AI system liquid cooling reduces need for chilling equipment.
Implication: Addresses water consumption as a major data center bottleneck, potentially easing site selection and environmental concerns for AI factory expansion.
Gartner projects a 49% increase in spending on AI-optimized servers in 2026, contributing $401 billion to AI infrastructure spending this year.
Implication: Indicates sustained, rapid growth in demand for specialized AI hardware and supporting infrastructure, driving significant capital expenditure across the sector.
AI infrastructure financing is consolidating around energized land and secured, often off-grid, power under long-term contracts, exemplified by Microsoft's 2.67GW Project Kilby with Chevron.
Implication: Highlights that reliable, dedicated power, rather than grid dependency, is becoming the scarce and decisive asset for scaling AI compute capacity.
What we'll be watching
- Groundbreaking for AWS's $50 billion federal AI infrastructure expansion, scheduled for 2026.
- First power for Microsoft's 2 GW Pecos, Texas data center campus, targeted by 2028.
- Further development planning and investment announcements for Buzz HPC's 320 MW AI infrastructure facility in the Greater Toronto Area.
Reporting + analyst voices: grounded via Google Search at publish time.