The vote and what it settled
On June 30, 2026 PJM's Members Committee endorsed one package for a reliability backstop procurement and rejected all of the connect-and-manage options that would have let the grid operator curtail large loads during capacity-short years. That is the pivotal choice. Faced with a projected shortfall driven largely by data centers, PJM's stakeholders chose to procure new supply rather than to switch demand off. The board will now turn the endorsement into filings with FERC under Section 205 in July.
The mechanism is a one-time, two-phase auction. In Phase I, from September 2026 through March 2027, data centers and load-serving entities can negotiate bilateral capacity contracts directly with new supply projects. After March 2027, PJM holds a central procurement to buy whatever capacity is still needed. The original target was 14.9 gigawatts of new resources online by mid-2031; the revised process points to roughly 9 gigawatts, with the procurement's average cost capped at $555 per MW-day.
Why owners of large loads should read this closely
The subtext of the vote matters more than the megawatt figure. By rejecting connect-and-manage, PJM's stakeholders declined to make large loads the shock absorber for grid stress. In the alternative that lost, a data center could have been ordered to power down when the system ran tight. The path that won instead commits the region to building or buying firm supply so that those loads can stay on. For a business whose economics depend on uptime, that is a meaningful signal about how the largest US grid intends to treat you.
It also front-loads a negotiation. Phase I lets large loads strike direct contracts with new projects before the central auction, and one analysis notes that data centers are well positioned to lock up available gas and battery supply early, at favorable pricing, before the rest goes to competitive auction. The advantage runs to whoever is ready to contract in the fall of 2026. Being slow to the table is a cost.
Who ends up paying
Someone funds this new supply, and the design points at load. Under the plan, load-serving entities bill the large loads for the procurement, which keeps the intent that the demand driving the need also carries the cost. But the reach is longer than one auction. Because backstop contracts can run up to 15 years, an analysis of the program warns that the prices struck now could remain embedded in customer bills into the mid-2040s, and that limited available supply hands developers pricing power that can pad bids.
The honest read is that this program is unlikely to lower power or capacity prices; it is a reliability instrument, not a discount. Anyone modeling the total cost of a large load in PJM should assume the backstop adds a durable line to the bill, not a temporary one.
The pattern this sets
PJM is the largest grid operator in the United States and the front line of data-center-driven demand, with new large load growth forecast at 55 gigawatts by 2030 and 100 gigawatts by 2037. How it resolves the collision between load growth and supply becomes a template other regions study. The choice made on June 30 was to guarantee supply and price it, rather than to ration demand.
Two things now decide how this lands for you. First, whether the FERC filings this month preserve the load-pays cost structure or shift some burden onto ratepayers, which will draw scrutiny. Second, how much firm supply actually shows up at a $555 per MW-day cap when developers hold pricing power. The region committed to keeping large loads powered. What that commitment costs, and who carries it, is the number to track through the summer.
Read next: A single storage contract in Europe now runs to 25 gigawatt-hours | Three privately built microreactors reached criticality inside one month
