50% of U.S. AI Data Centers Estimated Canceled, Hitting NVIDIA’s GPU Demand
NVDA•Analysts estimate that 50% of planned U.S. AI data centers will be abandoned due to grid interconnection failures, with 70% of requests withdrawn. This constraint may dampen NVIDIA’s GPU sales growth by limiting new AI infrastructure deployments.
1. Grid Capacity Shortfall Threatens AI Data Center Pipeline
The U.S. electrical grid, built for 1–2% annual demand growth, is straining under AI workloads that draw power equivalent to small cities. Berkeley Lab found over 70% of interconnection requests are withdrawn, and analysts project 50% of planned AI data centers will never be built due to grid limitations.
2. Major Tech Firms Secure Nuclear Power Deals
Microsoft signed a 20-year deal to restart Three Mile Island, Amazon paid $650 million for a data center campus beside Susquehanna nuclear station, Google agreed to small modular reactors, and Meta requested proposals for 4 GW of nuclear capacity. These moves highlight hyperscalers bypassing the grid to secure stable power.
3. Implications for NVIDIA’s GPU Demand Growth
With half of planned AI data centers at risk of cancellation, demand for high-performance GPUs could soften as fewer facilities come online. NVIDIA’s future revenue growth may be challenged if AI infrastructure expansion slows due to persistent power delivery bottlenecks.
