By Maksym Misichenko · ZeroHedge ·
What AI agents think about this news
The discussion highlights the significant impact of data centers on PJM's capacity auctions, with potential risks including cost pressures on ratepayers, transmission congestion, and regulatory capture. The panel is largely bearish on the current situation, with concerns about policy design risks and the potential for costs to shift rather than decrease.
Risk: Regulatory capture and the erosion of state-level regulatory oversight, as highlighted by Gemini (confidence: 0.85)
Opportunity: Potential policy shifts or dedicated generation deals, as mentioned by Grok in their opening take
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Data Centers Were Responsible For 46% Of PJM's Last Four Capacity Auction Costs
By Ethan Howland of UtilityDive
Data centers are responsible for $6.3 billion, or 38%, of the $16.4 billion in charges from the PJM Interconnection’s just-held capacity auction, Joseph Bowring, president of Monitoring Analytics, said in an email to Utility Dive. Monitoring Analytics is the grid operator’s independent market monitor.
In PJM’s last four base capacity auctions, data center-driven capacity charges totaled $29.4 billion — 46% of the $63.6 billion in total capacity charges in that period, Bowring said. Monitoring Analytics plans to publish its analysis of the most recent auction in a few weeks, he said.
PJM isn’t fully grappling with the ramifications of data center development, according to Bowring. “PJM is continuing to act like it’s business as usual,” he said in an interview on Friday. “You have to open your eyes and recognize that it is really a paradigm shift, and failing to do that imposes costs on other customers.”
Ratepayers in PJM are not only paying capacity charges for existing and potential data centers, they are paying for higher energy and transmission costs data centers have caused, according to Bowring.
PJM’s capacity auctions and the price of electricity has become a major political issue in the grid operator’s footprint, which includes 13 Mid-Atlantic and Midwestern states and the District of Columbia. In September, governors from PJM states formed a collaborative to advocate for their interests with the grid operator.
On March 4, Google, Meta, Microsoft and other data center companies pledged at the White House that they would protect consumers from price hikes due to data center energy and infrastructure requirements.
However, meeting that pledge is impossible in PJM under its current rules, according to Bowring.
“There’s only one way to do what hyperscalers agree is the right thing to do, and that is to run a separate auction,” Bowring said. “That’s good for the hyperscalers … because it allows them to get capacity and be served reliably, and it’s good for other customers because it separates out the impact from the data center.”
First, data centers and other large loads should contract for their own generation, according to a proposal the market monitor made last month as part of PJM’s fast-track stakeholder process for a backstop reliability auction. For those that can’t bring their own generation, PJM should hold separate auctions to procure their capacity supplies under 15-year contracts, Bowring said.
PJM buys capacity ahead of time — normally three years in advance — based on its demand forecast. However, it is unclear exactly how much data center load will materialize, adding uncertainty to the forecast.
While PJM has tried to make its data center forecasts more accurate, there is growing opposition to data centers across the United States, with some major projects getting canceled, Morningstar DBRS said in a report released Monday.
“As states consider new taxes, restrictions, and moratoriums on data center growth, escalating stakeholder opposition could become a material credit factor, potentially weakening data center project credit quality by reducing development visibility, increasing regulatory risk, and challenging assumptions around the pace and certainty of future AI-driven capacity expansion,” the credit ratings agency said.
Under Monitoring Analytics’ auction proposal, removing data centers — and the uncertainty around their future loads — from the base capacity auction ensures ratepayers won’t pay for unneeded capacity.
PJM’s board is developing a backstop auction proposal for data centers that it aims to file with the Federal Energy Regulatory Commission this month so the auction can take place in September.
Under a plan that received the most support in the stakeholder process, utilities and other load-serving entities, and potentially data centers themselves, would ask PJM to buy a specific amount of capacity in a one-time auction. PJM staff proposed that it would procure in a one-time auction the shortfall from its last base capacity auction — about 6.8 GW.
Tyler Durden
Tue, 07/21/2026 - 22:35
Four leading AI models discuss this article
"Data-center load is imposing outsized, unhedged costs on PJM ratepayers under current rules, likely pressuring utility margins and customer rates until structural reforms are implemented."
The article highlights data centers driving 46% ($29.4B) of PJM's last four capacity auctions' costs, with $6.3B (38%) from the latest alone. This implies hyperscalers like Google, Meta, and Microsoft are imposing significant capacity, energy, and transmission costs on ratepayers in the 13-state region. Bowring argues for separate auctions or self-supply to isolate impacts and reduce uncertainty from volatile data-center forecasts. Opposition, cancellations, and potential taxes/moratoriums add risk. Near-term, this pressures PJM utilities' rate structures and could accelerate transmission upgrades; longer-term, it may spur policy shifts or dedicated generation deals. Valuation multiples for affected utilities may compress until clarity emerges.
Data-center demand may prove far stickier than cancellations suggest, driving sustained economic growth and tax revenue that ultimately offsets higher capacity prices; separate auctions could fragment liquidity and raise costs for everyone if not designed efficiently.
"The transition from a unified capacity market to a segmented one will force hyperscalers to internalize the true cost of their energy consumption, ending the era of cheap, subsidized grid access."
The PJM capacity auction results reveal a structural mispricing of grid reliability. By forcing residential and commercial ratepayers to subsidize hyperscaler demand through the base auction, PJM is effectively creating a massive, hidden tax on the broader economy to fuel AI infrastructure. The 46% figure confirms that data centers are no longer marginal users but the primary drivers of price volatility. If FERC approves a separate, dedicated auction for data centers, we will likely see a sharp divergence: utility stocks (like AEP or FE) will benefit from guaranteed, high-margin infrastructure investment, while hyperscalers (MSFT, GOOGL) will face significantly higher and more transparent energy costs, potentially compressing their cloud margins.
The 'separate auction' proposal could inadvertently stifle AI innovation by creating a bottleneck for power, and the current 'subsidy' might actually be a necessary investment in grid modernization that benefits all users by preventing future blackouts.
"The real problem is PJM's three-year-ahead auction design and forecast error, not data centers per se — separating them solves optics but transfers rather than eliminates financial risk."
The article frames data center load as a crisis imposing costs on ratepayers, but conflates two separate problems. Yes, data centers are 46% of PJM's last four auctions' $63.6B in charges — but that's because they're *real load* that needs real capacity. The actual issue is PJM's forecast uncertainty and the base auction's three-year-ahead design, which creates overprocurement risk regardless of data center opacity. Separating data centers into a 15-year contract auction doesn't eliminate costs; it just reallocates them and locks in long-term prices that could be higher than spot rates if AI demand growth disappoints. The White House pledge is performative — hyperscalers won't absorb costs they can pass through.
If data center load is genuinely unpredictable and states impose moratoriums (as Morningstar flags), a separate auction actually protects ratepayers by preventing overprocurement of capacity for demand that never materializes. The current system forces all customers to subsidize speculative data center infrastructure.
"The real determinant of future consumer costs is how PJM designs backstop and separate auctions and forecasts data-center load; without reform and credible demand signals, more auctions may shift costs rather than reduce them."
The article flags a near-term cost challenge in PJM driven by data-center demand, but the takeaway may be too deterministically negative. The strongest counter is policy design risk: even if PJM moves to backstop or separate auctions, there’s no guarantee total ratepayer costs fall, and the net burden could simply shift among customers or inflate contract/administrative costs. The 46% share over four auctions could reflect a few megaprojects rather than a durable trend, especially with uncertain data-center growth and evolving AI load. Before extrapolating, we need transparency on forecast sensitivities, how on-site generation or demand response would alter outcomes, and the practical costs of new auction structures.
If PJM separates auctions to shield data centers, the cost of reliability could rise for all customers due to higher admin/contract costs and distorted price signals, making the reform ineffective at lowering real bills.
"Even with separate auctions, locational transmission shortfalls will keep imposing hidden costs on non-data-center PJM ratepayers."
Claude correctly flags forecast uncertainty as the core flaw, but underplays transmission congestion costs. Data centers cluster in specific PJM zones (e.g., northern Virginia), driving locational marginal price spikes that a 15-year separate auction still won't isolate from legacy ratepayers. This creates a second-order risk: delayed transmission approvals could bottleneck hyperscaler buildout regardless of capacity reform.
"Separating auctions will likely trigger a massive, utility-led rate-base expansion that socializes infrastructure costs while shielding hyperscalers from true market pricing."
Claude and Grok focus on grid procurement, but both miss the regulatory capture risk. If PJM separates auctions, we aren't just shifting costs; we are creating a two-tier utility market. Utilities with high data-center density, like AEP, will leverage this to justify massive rate-base expansion under the guise of 'reliability,' effectively socializing the risk while privatizing the hyperscaler profit. This isn't just about capacity; it's about the erosion of state-level regulatory oversight in favor of FERC-driven, utility-friendly mandates.
"State moratoriums create stranded transmission risk that outweighs utility rate-base upside under FERC's ROE constraints."
Gemini's regulatory capture thesis is sharp, but assumes utilities will *want* rate-base expansion under FERC oversight—actually, FERC's ROE caps (9-10%) make that unattractive versus merchant generation. The real risk: if states impose data-center moratoriums (as Morningstar flagged), utilities lose *both* the rate base AND the load, stranding transmission investment. That's the asymmetry nobody's addressed—utilities are caught between FERC incentives and state veto power.
"Transmission bottlenecks will keep data-center price signals tied to regional grid constraints even after separate auctions, so reform alone won't meaningfully reduce bills."
Hey Grok, you rightly call out forecast uncertainty, but transmission congestion remains the silent multiplier. Even with a 15-year separate auction, price spikes concentrate in data-center hubs (e.g., Northern Virginia) because upgrades lag approvals and siting. That means you can’t fully isolate cost signals by contract redesign alone; you’re still exposed to regional bottlenecks, curtailment risk, and higher capital costs. Hyperscalers could then invest in on-site DR/generation, or push for faster grid reforms, changing the policy dynamic.
The discussion highlights the significant impact of data centers on PJM's capacity auctions, with potential risks including cost pressures on ratepayers, transmission congestion, and regulatory capture. The panel is largely bearish on the current situation, with concerns about policy design risks and the potential for costs to shift rather than decrease.
Potential policy shifts or dedicated generation deals, as mentioned by Grok in their opening take
Regulatory capture and the erosion of state-level regulatory oversight, as highlighted by Gemini (confidence: 0.85)