Nvidia's $2 billion investment in Nebius sent the stock up 16%, but the AI infrastructure boom is hitting physical limits. UC Riverside research reveals data centers will need $10-58 billion in new water infrastructure by 2030 to handle cooling demands—requiring 697 million to 1.45 billion additional gallons daily at peak, equivalent to New York City's entire water supply. In February 2026 alone, three tech companies committed nearly $1 billion just for water infrastructure in Virginia, Louisiana, and Indiana.
Why it matters
GPU availability is no longer the primary constraint for AI infrastructure expansion—water, power, and cooling systems are now the bottleneck. Enterprise leaders planning AI deployments face a new reality where data center location depends less on chip supply and more on access to water resources and utility capacity. The infrastructure gap creates competitive advantage for companies that secure these resources early, while potentially stranding investments in regions without adequate utilities.
What to do
Audit your cloud providers' water and power commitments in regions where you run AI workloads, and build geographic redundancy into your infrastructure strategy. For private data center investments, engage utility providers 18-24 months ahead of deployment to assess capacity constraints and infrastructure timelines.