We have more details on how PJM plans to manage the 70 gigawatts of large load growth barrelling its way between now and 2038.
Following months of back and forth over how to run an emergency auction for data centers and whether the RTO is “too big to function,” PJM last week rolled out a multi-part plan for data centers that want to secure their own power supply, as well as a demand response requirement for loads that don’t.
The proposed framework is largely in line with previous proposals and stakeholder expectations. First, as the RTO detailed in plans filed to FERC on Friday, PJM will look to spend up to $20 billion to secure new capacity for data center load growth. That includes a bilateral matchmaking phase to pair up new large loads with new generation; any capacity shortfall that’s left over will be addressed by a one-time auction in phase two.
In total, PJM seeks to procure only the 6.8 gigawatts of capacity that the latest capacity auction failed to secure (as opposed to the full 24 GW shortfall the region anticipates by 2030).
The second part of PJM’s plan includes requiring data centers that don’t bring their own supply to shut off during times of grid stress, starting in 2027. Both parts of the plan rely on the creation of a large load registry, which will track sites greater than 50 megawatts, with information like location, size, and any onsite generation.
Taken together, the procurement plan, the curtailment requirements, and the registry offer a workable stopgap for the intersecting problems facing the region, said Julia Hoos, head of USA East at Aurora Energy Research — problems it needs to fix on an extremely short timeline. However, Hoos added, “this is a system that is designed to break within a year or two,” and may not bring down costs for ratepayers in the meantime.
Eventually, PJM will still need to procure more capacity for its non-data center load, presumably through its central capacity auction, she explained. But that complicates incentives for generators. If they can sign long-term deals directly with a data center via this new matchmaking process, they’re unlikely to show up in the main auction.
“If I’m a new generator trying to build in 2032, and this current system still exists, why in the world would I participate in the regular [capacity auction]?” Hoos said.
Even without data centers, load growth from other causes would still eventually cause the same capacity market cost increases, explained Brent Nelson, managing director of markets and strategy at Ascend Analytics.
“It doesn’t really matter what causes your system to go from long to short, whether it’s a large data center or a bunch of new housing developments and a new manufacturing facility and a new EV charging network,” Nelson explained. “If you push demand past where supply sits, you still end up with a shortage problem.”
The only way to get around that reality is to funnel revenue specifically to new sources, rather than to the entire existing fleet. That means relying less on the general capacity auction and more on targeted mechanisms like state-sponsored solicitations or direct bilateral contracts between retailers and developers, Nelson explained to Latitude last year. And so far, that’s only an option for large loads.
Data center demand response is back on the table
For large loads that can’t or won’t procure their own capacity, PJM is proposing a backup option in the form of mandated demand response. Under the “Interim Resource Adequacy Service,” or IRAS, a data center can still get served if it doesn’t bring 100% of its own capacity — but it would be forced to curtail its power use during tight conditions.
PJM itself doesn’t have authority to force load to curtail, which is one of the many issues PJM’s independent market monitor raised last summer amid a debate over whether data center flexibility in the region was a “regulatory fiction.”
To get around that problem, PJM has essentially assigned the responsibility of enforcement — both cost allocation and physical curtailment — to states and utilities, said Mary Yang, chief operating officer at 804 Advisory. For supply procured in the auction rather than in bilateral deals, PJM plans to pass the wholesale costs down to local utilities based on standard allocation formulas. State regulators will then sort out whether and how to assign those costs to specific data centers.
“If the states don’t do that, PJM is essentially saying, ‘Well, we’re just going to allocate it to everybody, including retail ratepayers, and you guys deal with the political fallout,’” said Yang, an energy attorney who previously served as regulatory counsel for PJM. States will also decide whether to compensate these loads for ramping down, she added.
This comes at a time when states themselves are asking for a greater role in PJM’s governance. But the delegation structure also recognizes the current political landscape of anger over high electricity prices and local opposition to data centers, said Hoos.
States and ISOs were once extremely wary of alienating data centers by suggesting they’d have to curtail. A year or two ago, developers were even able to demand the removal of certain regulatory barriers because the potential tax revenue and other benefits for a region were so high; in that environment, asking data centers to change their operations for the sake of the grid was untenable.
But things have changed rapidly, and the choice to just “go elsewhere” doesn’t really exist anymore, Hoos said.
“We’ve gotten down to the real brass tacks of…what’s truly non-negotiable to keep moving ahead,” she added. “Originally data centers were saying there’s no sort of flexibility that’s acceptable, but when it’s between that and not getting connected at all, some flexibility, in some instances, suddenly becomes acceptable.”
Persistent price problems
That said, the combination of last week’s plan and PJM’s separate, FERC-directed co-located load proceeding “is designed to push everybody to bring their own generation,” Yang said.
It’s a major win for the emerging industry of specialized onsite generation services, including technologies like long-duration battery storage, gas turbines, and distributed resource aggregation. PJM now expects that data centers will be working with these kinds of providers, Yang said: “If you’re a large load without on-site generation in the queue and teed up, you’re already kind of behind at this point,” she added. “I really expect data centers to bring their own capacity as the default at this point.”
The biggest barrier, Hoos said, is how challenging it still is to build the new generation these large loads require, even for projects that have guaranteed offtake.
“I would expect to see…a lot of things get built unless we really can’t solve the physical building part, which I think is a real concern,” she added. The challenges of supply chains, construction, and grid interconnection are bigger than anything PJM can solve on its own.
But even if all of the procured new generation comes online as planned, it will still be “a drop in the ocean” of what will ultimately be needed to keep up with load growth and retirements, Yang said.
Meanwhile, “people are at odds” about whether, and how soon, PJM’s new approach might bring down prices for consumers in the region, she added; some observers argue that it might take between three and five years to really see an impact.But those disagreements are part and parcel of the reality of this procurement process. “These are quick proposals,” Yang said. “They’re ones that don’t have a lot of precedent.”
As Hoos put it: “We’re at the start of a very big experiment.”
