Power Constraints Threaten 50% of AI Data Centers, Impacting NVIDIA Chip Demand
NVDA•Over 70% of US grid interconnection requests are withdrawn and 50% of planned AI data centers may never be built due to power constraints, potentially limiting NVIDIA's chip demand. Hyperscalers are signing long-term nuclear and Nordic hydroelectric deals to secure low-cost power, highlighting infrastructure risks that underpin NVIDIA's AI ecosystem.
1. US Grid Constraints Limit AI Data Centers
US grid interconnection requests are failing at over 70%, with utilities unable to deliver on three-year timelines and projections indicating half of planned AI data centers may never be built under existing infrastructure.
2. Hyperscalers Lock in Long-Term Power
Major cloud providers have committed to long-term power agreements, including restarting dormant nuclear facilities and co-locating data centers with nuclear stations, to secure the hundreds of megawatts needed for AI workloads.
3. Nordic Hydroelectric Advantage
Nordic countries offer abundant hydroelectric and nuclear capacity, cold climates that reduce cooling expenses, and strong data sovereignty protections, but new power allocations have been capped, closing the window for latecomers.
4. Implications for NVIDIA
NVIDIA's AI chip sales depend on data center build-outs, making the availability of secured, low-cost power a critical factor in the company's growth trajectory and market valuation.
