# Capacity Market A market that pays generators to **exist and be available**, separately from paying them for the electricity they actually produce. The product is a promise: be ready to run on the delivery day, or face a penalty. Capacity markets exist because energy-only markets have a theoretical problem. A generator that runs a handful of hours a year — the peaker that covers the hottest afternoon — cannot recover its fixed costs from those hours unless prices spike enormously, and regulators are reluctant to allow price spikes of the necessary magnitude. Rather than tolerate the spikes, most eastern US markets pay a separate capacity charge to keep the fleet large enough. [[ERCOT]] declined to, which is the cleanest available comparison. ## How the auction works 1. The operator forecasts peak demand for a delivery year, typically three years ahead, and adds a **reserve margin** — a required cushion above forecast peak. 2. That total is the **reliability requirement**, the quantity the auction must procure. 3. Resources bid the price at which they will commit. The operator stacks bids from cheapest and takes them until the requirement is met. 4. The last bid accepted sets the **clearing price**, which is paid to **every** cleared resource, not just the marginal one. 5. The total cost is allocated to load-serving entities in proportion to their share of peak demand, and flows through to customer bills. Step 4 is why the number moves so violently: a modest increase in the requirement can pull in a much more expensive marginal resource, and that new price is then paid to the entire fleet. > [!important] This is the mechanism that converts a data center into your bill > A data center does not appear on a residential bill as a line item. It appears as a larger forecast peak, which raises the reliability requirement, which lifts the clearing price paid to all cleared capacity, which raises the total charge allocated across load. The path is real, arithmetically traceable, and completely invisible to the person paying. ## The PJM case [[PJM Interconnection]] clearing prices ran $28.92/MW-day for delivery year 2024/2025, $269.92 for 2025/2026, and $329.17 for 2026/2027 — roughly eleven-fold in two years. PJM's Independent Market Monitor attributed **$6.3 billion of the $16.4 billion** cost of the 2028/2029 auction to data center load, and **$29.4 billion of $63.6 billion (46%)** across the last four base auctions. An earlier Monitoring Analytics analysis calculated that current and projected data-center demand raised capacity costs for the 2025/2026 delivery year by **$9.3 billion, or 174%**, against a no-data-center counterfactual, attributing 63% of the year-over-year increase to data centers. ## Price caps and what they actually do FERC approved a price **collar** — a cap and a floor — for PJM's auctions, and extended it in April 2026 through the 2029/2030 delivery-year auction. The 2027/28 and 2028/29 auctions both cleared **at** the cap. The instructive part is the counterfactual PJM itself published: without the collar, the 2028/2029 auction would have cleared at **$554.72/MW-day**, roughly $30 billion, against $16.4 billion actually charged. The cap saved load about half the bill in that auction — and the auction came in **6,831 MW short** of its reliability requirement, at a 14.4% reserve margin against a 20% target. That is the trade in one sentence: **a price cap converts a cost problem into a quantity problem.** PJM then proposed a separate backstop auction with a **$555/MW-day** cap — essentially the suppressed price — to procure the missing capacity. The money did not disappear. It moved to a different auction with a different name and a fifteen-year commitment. ## The critique from both directions **From the left and from consumer advocates:** capacity markets have paid tens of billions to existing plants that were never going to retire, the clearing-price mechanism overpays inframarginal resources by design, and the cost is socialized across captive ratepayers who had no say in the forecast that drove it. **From market economists and from Texas:** capacity payments are a subsidy that suppresses the scarcity pricing which would otherwise call forth investment and, more importantly, demand response. Paying for availability lets load ignore the real-time cost of its own consumption. Both critiques point at the same feature — the market pays for a forecast rather than for behavior — and disagree about the remedy. Note that the 2026 reforms in PJM move toward the Texas answer: [[Demand Response]] obligations, curtailable service, and "bring your own capacity" all make large load bear consequences it previously avoided. ## See Also - [[PJM Interconnection]] — the largest and most contested capacity market - [[ERCOT]] — the energy-only counterexample - [[Cost Allocation]] — how the cleared cost gets divided - [[Federal Energy Regulatory Commission]] — the approver of every auction parameter