AI Data Centers Drive 76% Surge in PJM Electricity Prices – Watchdog Calls for Direct Power Procurement

Announcement

The independent oversight body Monitoring Analytics released a report this week linking a 75.5 % increase in wholesale electricity prices across the PJM Interconnection – the largest U.S. grid operator – to the rapid expansion of AI‑focused data centers. Wholesale rates rose from $77.78 /MWh in Q1 2025 to $136.53 /MWh in the same quarter of 2026, a jump the watchdog describes as “irreversible” without immediate policy action. The report also flags the upcoming Base Residual Auction (BRA) scheduled for June 2026 as a potential catalyst for even higher prices if data‑center demand remains baked into the market’s capacity forecasts.

Technical specs and supply‑chain context

Metric	Q1 2025	Q1 2026	Change
Wholesale price (USD/MWh)	77.78	136.53	+75.5 %
AI data‑center load (GW)	5.2	9.1	+75 %
Total PJM peak demand (GW)	173	180	+4 %
New generation capacity (GW) added 2024‑2025	3.4	3.6	+6 %

The data‑center load grew ~75 % year‑over‑year, outpacing the modest ~4 % increase in overall system demand. Because the PJM capacity market clears three years ahead of actual need, the surge in compute power has forced the market to schedule additional megawatts that are not yet built, compressing the supply‑demand gap and driving up the clearing price.

Why the price spike matters for the grid

• Heat‑dissipation: AI workloads push server utilization to 90‑95 % CPU/GPU occupancy, raising rack‑level power draw to 30 kW per rack, compared with the 10‑12 kW typical of web‑hosting farms.
• Cooling load: Higher density translates to 2‑3 × the cooling water flow, increasing the need for chilled‑water plants and, in some regions, supplemental natural‑gas‑fired chillers.
• Transmission stress: The PJM network’s 345‑kV corridors are approaching 85 % of thermal limits in the Ohio‑Pennsylvania corridor, leaving little headroom for additional large‑scale loads.

Market implications

Direct power procurement as a mitigation path

The watchdog proposes that AI data centers negotiate bilateral power purchase agreements (PPAs) with generators rather than participating in the BRA. A PPA‑only model would:

1. Isolate demand – only the data‑center’s load would be factored into the capacity auction, keeping residential and small‑business demand out of the high‑price calculation.
2. Lock in price – long‑term contracts (10‑15 years) can hedge against spot‑market volatility, potentially capping the effective cost at $90 /MWh for the compute tenant.
3. Accelerate renewable build‑out – many hyperscalers already sign renewable PPAs; scaling this to AI‑intensive sites could add 2‑3 GW of wind/solar capacity annually.

Potential push‑back from PJM

PJM argues that integrating large, predictable loads into the capacity market improves overall system reliability. However, the report shows that the capacity price for the June 2026 BRA is projected to exceed $200 /MWh, a level that would ripple through distribution tariffs and raise residential bills by an estimated $15‑$20 per month on average.

Legislative backdrop

In March 2024, the White House convened the major AI hyperscalers and secured a “ratepayer protection pledge” – a non‑binding commitment that the firms would fund the infrastructure needed for their power consumption. The pledge lacks statutory force; without a congressional amendment to the Federal Energy Regulatory Commission (FERC) rules, PJM can still allocate the cost of new generation to all ratepayers.

Outlook

If PJM proceeds with the current BRA design, analysts project a 10‑12 % year‑over‑year increase in average residential electricity rates across the PJM footprint through 2027. Conversely, a shift to direct PPAs could contain the incremental cost to 3‑4 %, preserving competitiveness for both data‑center developers and end‑users.

---

Key takeaways

• AI data centers added roughly 3.9 GW of load in one year, outstripping overall demand growth.
• Wholesale power prices jumped $58.75 /MWh YoY, a 75.5 % rise.
• Direct PPAs are presented as the most viable tool to prevent further price inflation for non‑compute customers.
• Legislative action is required to enforce cost‑allocation rules; otherwise, the burden will continue to flow to households and small businesses.

For a deeper dive into PJM’s capacity market mechanics, see the official PJM capacity auction guide. The full Monitoring Analytics report is available as a PDF download.