What Happened (The Mechanism, Not the Marketing)

PJM’s proposal is a two-phase, one-time procurement intended to address projected capacity shortfalls driven by large load growth, including data centers. Phase one runs September through March, with PJM and a consultant (Charles River Associates, per PJM’s proposal) acting as confidential intermediaries to facilitate bilateral contracting between buyers (large loads) and sellers (power suppliers). PJM would not be a counterparty, but it would play the institutional matchmaker.

Phase two is the “fine, I’ll do it myself” clause: in March, PJM would procure any remaining capacity needed through a PJM-run central procurement. Bidders can offer terms from two to 15 years, and winning resources would still participate in PJM’s base capacity auctions (offering at $0, receiving the clearing price), with the backstop contract structured as a contract-for-differences to top up (or claw back) to the agreed revenue.

The Non-Obvious Angle: This Is the Grid Turning AI Into a Contract, Not a Vibe

Everyone loves talking about AI like it’s software. PJM is treating it like what it is in practice: a new class of industrial load that shows up faster than the physical supply chain can respond. The interesting shift here is not “PJM needs more capacity.” PJM always needs more capacity. The shift is that the grid is trying to force coupling: if you want to add giant demand, you should bring (or fund) giant supply.

That’s compute policy by spreadsheet. Not a national industrial strategy. Not a keynote about innovation. A procurement mechanism with a timeline, a target number, and a bill.

Why This Matters (Beyond PJM’s Footprint)

• Large loads become price-makers: when your incremental demand can destabilize adequacy forecasts, the market stops being “competitive” in the idealized sense. It becomes a negotiation about who pays for risk.
• Permitting and gas infrastructure sneak in through the back door: PJM explicitly points at transmission build-out and “other infrastructure needs” (including natural gas infrastructure) as part of the reality. You can’t procure your way around pipes and interconnection queues.
• Defining “who pays” is the whole war: PJM is aligning itself with a White House and PJM-governor statement of principles that emphasizes cost allocation to the load buying the capacity. In plain terms: the AI boom doesn’t get to socialize the grid bill indefinitely without a fight.

A Quick Scenario Map (Because This Is Horizon, Not a Press Release)

Scenario A: The “Bring Your Own Megawatts” Future

Bilateral deals dominate phase one. Data center developers pair with generation developers and lock in bespoke terms. This accelerates private-market coupling (and quietly increases the power of whoever can finance and permit generation fastest). The grid becomes the referee, not the builder.

Scenario B: The “Central Procurement” Future

Bilateral interest underperforms. PJM ends up buying more of the shortfall itself through central procurement, pushing costs through utilities and then into the political arena. Expect state regulators and consumer advocates to ask why residential customers are exposed to over-procurement risk. PJM itself flags a review period where the quantity could be scaled back, because “overbuying” is how you lose friends and get hearings.

Scenario C: The “Curtailment-as-a-Service” Future

Parallel PJM work on “connect-and-manage” for large loads develops into a more formal curtailment regime for early interconnections. Large loads connect earlier, but accept curtailment under defined conditions or provide their own backup (storage or generation). In other words, you can plug in early, but you agree to be the first thing turned off when the grid gets nervous.