SemiAnalysis • 10809 implied HN points • 03 Mar 26
- PJM’s simulation-driven capacity market and optimistic datacenter load forecasts caused capacity auction prices to soar, shifting roughly $16 billion in costs onto customers and adding about $25–$30 a month to household bills.
- ERCOT’s energy-only model with real-time scarcity pricing and skeptical planning absorbed similar datacenter growth without a 9x price spike, and its operational reforms helped the grid hold up during Winter Storm Fern.
- The crisis highlights that market design and regulatory speed—not AI datacenters alone—drive price shocks; fixing forecasting methods, capacity incentives, and treating datacenters as flexible grid resources is needed to avoid political fallout and misallocated costs.