Carlos, I’ve spent my career straddling the control room and the commission docket, so I see both the physics and the policy. Right now, PJM is in a moment where data center demand and aging thermal fleets are colliding, and record-high capacity prices landed even as the region missed its reliability target. Against that backdrop, a proposed 1,267‑MW peaker acquisition—split between a 677‑MW simple‑cycle gas unit in Dixon, Illinois, and a 590‑MW gas/diesel unit near Dayton, Ohio—has kicked up urgent questions about the flow of capacity out of public markets and into private data center load. My goal here is to unpack the market mechanics, the regulatory levers, and the practical safeguards that can keep reliability whole while still enabling growth.
PJM faces tightening supply-demand due to rapid data center growth. How is this changing capacity procurement dynamics, and what concrete metrics do you track weekly to flag emerging scarcity? Share an example of a recent inflection point and the operational steps you took in response.
Capacity procurement has shifted from a comfortable surplus mindset to a hunt for assured megawatts, with bidders pricing not only fuel and maintenance but the option value of redirecting capacity to dedicated data center load. Weekly, I watch offer stacks and the depth of marginal supply in the prompt and next auctions, outage patterns versus historic baselines, interconnection withdrawals, and the pace of new load adds tied to data centers. The inflection point came when PJM hit record-high capacity prices at the same time it fell short of its reliability target; that was the smoke alarm going off in a quiet house. My immediate steps were to tighten derating assumptions in our portfolios, bring forward demand response enrollments, and pre-negotiate seasonal bilateral option strips to hedge against further capacity scarcity while pressing for near-term uprates at existing sites.
A 1,267‑MW acquisition of two peaking plants is proposed during record-high capacity prices. How could such a deal influence clearing prices and reliability, and what scenarios would you model first? Walk us through your methodology and the assumptions you’d stress-test.
With prices already elevated, moving 1,267 MW of fast-start peaking capability under a buyer linked to data center development alters both supply elasticity and the threat of capacity withdrawal. I’d model three bookends: status quo participation in PJM’s market, partial redirection of capacity to private data center load, and full withdrawal framed as on-site or dedicated bilateral service. Methodologically, I’d run nodal to zonal translations to see how local deliverability constraints amplify the price effect, then apply sensitivity runs for outage rates, gas constraints, and offer behavior consistent with peaker scarcity rents. I’d stress-test assumptions around the timing of capacity commitments, maintenance windows, and the impact of any retirements like the buyer’s planned retirements of a 65‑MW and a 600‑MW plant by June 1, because that combination can tighten the stack even further.
One facility is a 677‑MW simple-cycle gas plant in Dixon, Illinois; the other is a 590‑MW gas/diesel plant near Dayton, Ohio. How do their operating characteristics affect their leverage in tight peak hours, and what unit-specific data would you analyze to gauge market impacts?
The 677‑MW simple-cycle in Dixon has classic peaker leverage: fast starts, high flexibility, and the ability to set price in scarcity intervals. The 590‑MW gas/diesel unit near Dayton carries fuel-switch optionality that can be pivotal when gas delivery is tight or prices spike intra-day. I’d analyze start times, minimum run durations, ramp rates, forced outage patterns, and historic performance in peak hours alongside any local transmission constraints that elevate their local capacity value. Fuel inventory logistics and dual-fuel switchover protocols at the gas/diesel unit matter a lot—down to how frequently it successfully tested transitions—because in shoulder seasons and heat waves, that reliability premium can translate directly into capacity market leverage.
Some developers are repurposing power plant sites into energy-and-data center campuses. What incentives drive shifting generation from capacity markets to dedicated data center load, and how do you quantify the cost shifts to other customers? Offer a case study with numbers.
The incentives are straightforward: stable offtake from a creditworthy data center, avoidance of market price volatility, and potential capture of both energy and resiliency premiums. When capacity is redirected to private load, the broader market loses that MW count, which can lift clearing prices and shift risk. As a case study, assume a 590‑MW plant redirects a meaningful tranche to a campus; when data centers already accounted for 40% of PJM capacity costs in the last auction, even a modest shift tightens the stack and pushes more cost onto the remaining customer base. You quantify it by comparing the cleared capacity volumes before and after redirection and attributing the incremental price increase to the removal; that delta—times the remaining load—represents the cost shift borne by other customers.
When capacity is redirected to private load, you’ve called it a form of withholding. What evidence patterns would convince you this is happening, and what practical screens or thresholds should market monitors apply? Describe a step-by-step detection process.
I look for a consistent pattern: a generator with a history of offering at competitive levels suddenly withdraws or derates capacity concurrent with public data center development milestones. Step one is timeline alignment—map application filings and site announcements to capacity offer changes. Step two is behavioral analysis—evaluate whether offers became noncompetitive or if deratings clustered around peak seasons without commensurate maintenance needs. Step three is counterfactual modeling—estimate what the unit would have cleared at with prior behavior. Step four is disclosure review—cross-check any bilateral arrangements or on-site supply claims. If the evidence shows removal tied to data center load with market price impacts, that’s functional withholding.
Section 203 reviews center on competition, rates, and regulation. How should regulators adapt these tests to account for capacity withdrawal risks tied to data centers, and what concrete commitments would you require at approval? Provide sample clauses and enforcement mechanics.
Regulators should add an ability-and-incentive lens specifically for capacity withdrawals to serve data centers, because merger policy historically didn’t contemplate this. I’d require an explicit commitment: “Purchaser shall not remove, derate, or dedicate capacity from the Lee County (677‑MW) and Tait (590‑MW) units to serve data center load outside PJM’s capacity market absent Commission-approved tariffs.” Include mandatory disclosure of any data center affiliations and bilateral contracts. Enforcement needs automatic remedies: immediate re-offer requirements in the next auction window and financial make-whole obligations to ratepayers for any demonstrated price impacts during the breach period, with escalation for repeated violations.
A developer involved in the deal is also advancing a data center campus on a retired plant site. How do you assess the “ability and incentive” to self-supply in this circumstance, and what structural safeguards can separate roles? Share governance and ring‑fencing examples.
Ability is established if the owner controls dispatchable capacity and interconnection rights; incentive exists where the developer can monetize reliability premiums at a campus, like the Sammis site redevelopment. I’d implement corporate separations with independent boards, no shared executives, and arm’s‑length contracts subject to audit. Information firewalls should prohibit the data center arm from accessing nonpublic bidding, outage, or maintenance plans of the generation arm. Ring‑fencing can include dividend restrictions triggered by market power findings, cash management separation, and covenants barring cross‑collateralization between the plants and the data center development entity.
If owners pledge not to remove specific units from the capacity market, what monitoring and penalties make such commitments credible? Detail measurement points, reporting cadence, default remedies, and how you’d address partial deratings versus full withdrawal.
Credibility starts with clear measurement points: unit-level offered and cleared capacity, seasonal UCAP/ICAP values, and forced/maintenance outage logs. Reporting should be quarterly to FERC and the market monitor, supplemented by event-driven notices for any material derate. Default remedies include an automatic requirement to re-offer at cost-based parameters in the next window and funding of an escrow to cover estimated uplift impacts until compliance is verified. For partial deratings, require root‑cause documentation and independent engineering review; if the derate aligns with private load milestones without technical justification, treat it as a pro‑rata breach subject to the same remedies as full withdrawal.
Some older plants are slated for near-term retirement while others might be repurposed. What criteria determine whether units retire, retool, or shift to bilateral service for data centers, and how do interconnection rights factor in? Provide a recent, concrete decision tree.
Start with economics under record-high capacity prices: if a unit still can’t cover going‑forward costs, it leans toward retirement; if it can, evaluate retool options and bilateral pathways. Next, check physical constraints—fuel flexibility, emissions headroom, and the cost/benefit of reliability upgrades. Interconnection rights are pivotal; retaining them can make a repowered or bilateral solution attractive. A decision tree I’ve used: if the owner lacks a captive offtaker and has acceptable maintenance risks, keep in the PJM market; if aligned with a campus and can technically island without jeopardizing compliance, seek FERC‑approved exceptions; if costs overwhelm even with high prices, retire—much like the planned retirements of a 65‑MW and a 600‑MW unit by June 1 cited in filings.
Data centers accounted for a large share of recent capacity costs in PJM. How should cost allocation evolve so hyperscalers bear their fair share without stifling investment, and what transition timeline is realistic? Offer tariff concepts and numerical examples.
With data centers at 40% of capacity costs in the last auction, the tariff should more granularly align charges with coincident responsibility. Concepts include a peak-coincident allocator tailored to data center load shapes and a dedicated reserve requirement for large campuses. A pragmatic transition could phase in specialized allocators over multiple auctions while offering credit for verifiable demand flexibility or on-site resources that remain inside PJM’s capacity framework. For example, allocate capacity charges such that high-load campuses pick up a proportionate share of the 40% burden, then offer a decrement if they commit capacity back to PJM rather than withdrawing it.
PJM recently missed its reliability target while prices surged. What near-term measures best stabilize reserve margins—demand response, fast-track uprates, transmission fixes, or interim capacity procurements—and how would you prioritize them? Share implementation sequences and KPIs.
In the near term, the fastest relief is demand response paired with targeted uprates at existing thermal sites; transmission helps, but lead times can be stubborn. I’d sequence actions as follows: expand demand response enrollment, green‑light low‑risk uprates with accelerated testing, run a focused interim capacity procurement for near‑term delivery years, and prioritize congestion‑busting transmission fixes. KPIs include enrolled MW of demand response, tested uprate MW placed in service, interim procurement cleared volumes, and reductions in localized scarcity events. The goal is to bridge the gap exposed when PJM fell short of its reliability target amid record-high prices, buying time for structural fixes.
How could behind-the-meter generation at data centers change market dynamics, and what rules are needed to prevent double counting of capacity or uplift cost shifting? Propose verification steps and metering standards with practical audit trails.
Behind-the-meter (BTM) resources at campuses can either relieve or exacerbate scarcity depending on whether they stay inside or participate in PJM’s markets. To prevent double counting, require exclusive election: either register as a PJM capacity resource with full telemetry and testing or remain purely BTM with no capacity credit. Verification requires revenue-grade interval metering at the generator and facility boundary, witness-tested performance during system stress, and immutable logs for audits. Tie settlement eligibility to those audits; if a resource claims PJM credits, it must deliver during scarcity or face clawbacks that discourage uplift shifting.
For competition concerns, when does aggregation of peakers plus data center load cross a market power threshold? Describe the market share and pivotal supplier tests you’d apply, and give a worked example using realistic summer peak assumptions.
I’d apply both market share and pivotal supplier screens at the relevant capacity deliverability area. Market share asks whether the portfolio holds enough MW to materially influence price; pivotal supplier asks if the market can meet demand without the supplier’s capacity. In a summer peak scenario, if the system needs nearly all fast-start units to clear reliability, even a modest share aggregated with a large data center offtake can render the owner pivotal—especially if they can redirect supply to private load. I’d flag any instance where the combined 677‑MW and 590‑MW units appear necessary to meet zonal requirements under stress, particularly when aligned with a developer advancing a data center campus on a retired site.
What role should states play—siting, air permits, and incentives—when capacity might migrate from public markets to private load? Outline a cooperative state-federal playbook, including transparency requirements and timelines that keep reliability whole.
States control the ground game—siting, air, and incentives—so they should condition approvals on preserving regional reliability. The playbook: states require disclosure of any intention to withdraw capacity, coordinate with the RTO and FERC under a shared timeline, and enforce transparency on fuel plans and backup operations. In parallel, FERC embeds conditions in Section 203 approvals that sync with state permits, ensuring that any deviation triggers joint enforcement. This cooperative approach keeps eyes on both the substation fence and the docket, so capacity doesn’t quietly slip from public service to private benefit.
If FERC conditioned approval on keeping the Dixon and Dayton units in PJM’s capacity market, how would you structure exceptions for maintenance, retrofits, or co-location projects? Provide a clear matrix of allowed actions, notice requirements, and penalties.
I’d allow three narrow exceptions. For maintenance: permitted seasonal derates with 30‑day advance notice and post-work test results. For retrofits: approved outages tied to documented upgrades with a commissioning test before re-entry. For co-location: data center interconnection allowed only if capacity remains registered in PJM with no net reduction; any temporary testing-related curtailments must be scheduled off-peak with day‑ahead notification. Penalties escalate from re-offer mandates to financial restitution for demonstrated market impacts and, for repeated violations, suspension of eligibility to participate in future Section 203‑related approvals.
PJM market monitor urged FERC to reject a 1.3‑GW gas plant deal between Hull Street and Rockland Capital, citing risks that capacity could be redirected to data centers. How should regulators document and enforce a “no removal” commitment in this specific context to protect customers from cost shifts the monitor warned about?
The commitment should be unit‑specific and time‑bound, naming the 677‑MW Lee County and 590‑MW Tait units and binding successors and assigns. It must include a prohibition on offering derates or withholding tied to serving data center load and require disclosure of any affiliations with entities developing the Sammis energy-and-data center campus. Enforcement should incorporate automatic reporting to the market monitor, with failure triggering immediate corrective re-offers and restitution to address the cost shifts the monitor highlighted—where removing capacity to serve data centers would otherwise push costs and risks onto other PJM customers. Finally, require a re-filing if ownership or development plans change, ensuring FERC re-tests competition, rates, and regulation under Section 203.
What is your forecast for PJM’s capacity market and data center integration?
Over the next few auctions, I expect tight but stabilizing conditions as PJM and FERC harden rules around capacity withdrawal and as owners adapt to clearer guardrails. Data centers aren’t going away; the more realistic path is structured integration that keeps the bulk of large peakers—like the 677‑MW and 590‑MW units—inside PJM’s capacity stack while allowing thoughtfully designed co-location. If regulators follow through on explicit “no removal” commitments and targeted cost allocation, the pressure that drove record-high capacity prices when PJM missed its reliability target should ease, though not evaporate. In short, we’ll move from a scramble to a managed coexistence—still taut, but with better transparency and fewer surprises for customers footing the bill.
