In early July, when the PJM capacity auction cleared at another record high, it intensified the ongoing conversation about whether power markets are prepared for the coming surge of demand from data centers.
The auction, which ensures power availability to meet the coming load several years in advance, settled at the capped price of $329 per megawatt-day. That outcome, explained former FERC commissioner Allison Clements, is a reflection both of surging demand, and of the limits of existing supply amid a data center boom.
But it’s not an “emergency,” she clarified. Instead, the auction results are more of a warning sign: markets are tightening, and the region’s long-term reliability strategy remains unsettled. This is complicated by PJM’s hybrid market structure, which mostly comprises competitive retail states with a capacity market on top.
Clements, who served at FERC from 2020 until 2024, currently advises data center developers, energy companies, and utilities. Last week, she joined a Latitude Dispatch to unpack the implications of the PJM capacity auction for the era of artificial intelligence. Below, find excerpts of our conversation, edited for brevity and clarity.
How unique is what we’re seeing in PJM? Is it a leading indicator for other markets?
Alison Clements: There are some unique challenges in PJM in that you’ve got restructured states and you’ve got a wholesale market. That wholesale market is not just energy and reserves; it also has a capacity construct, and you have 13 states involved, each with really different state level energy policies.
The combination of those things is a real recipe for difficulties. But we’re in a context around the country of tightening energy supplies and of rising costs and of community concern about data center development and what it’s going to bring to a given state. So everything is changing around the country. PJM certainly has its own unique set of challenges and issues to be dealt with, though. People want to build in PJM and they can’t right now, as easily as they should be able to.
What are they key takeaways from the auction for data centers in particular?
Alison Clements: It’s not emergency conditions. There needs to be new generation investment in PJM, and therefore prices are going up.
If the capacity market existed next to a high functioning and efficient interconnection process in PJM — which it does not today — and interconnected into a highly efficient and functioning local and regional transmission system planning process, then you’d have all the three legs of the stool working. In that case these prices coming out of PJM should ultimately balance out when new generation gets built, comes online, and there’s grid infrastructure to support it.
But in terms of getting speed to power…it’s getting harder and harder. It used to be that a data center company, even two years ago, could go up to your favorite utility in PJM and sign an energy services agreement. Now that simply isn’t the case. There are processes being developed in real time. Some utilities are leaning into the opportunity to facilitate the load growth. Some are perhaps overcorrecting in fear of costs being stranded.
So things are getting harder. The glimmer of hope for data centers is that some of the actions being taken on the load side are helping to clear out some of the speculative load interconnection requests. We might be over that time of pure gold rush, and getting down to more serious players. We’re still going to have a lot of challenges. It’s still a lot of growth, but at least we might be moving into that phase of the timeline.
How should developers be thinking about PJM’s proposed non-capacity backed load initiative, or NCBL?
Alison Clements: There’s a lot of concern about rising prices, about costs being shifted from data center developers over to residential customers, around stranded costs for the infrastructure that needs to get built. So the PJM management and staff put forward a proposal: NCBL.
Now, it is nothing if not a controversial proposal. On top of the capacity market construct, it would overlay an administrative process by which new large loads — 50 megawatts or above — could volunteer to curtail during emergency conditions. If there aren’t enough volunteers around the region, then PJM can require mandatory backing down of these loads. So it puts the new large loads at low priority relative to firm power availability in the case of an emergency.
The concept, with nuance, might be okay. Trying to figure out a way to decrease the capacity obligations of new large loads, conceptually is a good thing. But the way NCBL is being proposed — and of course there’s a lot of details we’re still waiting on — there is a lot of concern about it.
It’s meant to be temporary, just until the system planning and interconnection processes catch up and we have enough space and supply on the grid to serve demand. But you can’t finance a new power plant based on uncertainty around whether you might have to back down or not in any given year, and how much you might need to spend to ensure that if you do get backed down, your data center operations are only interrupted in an acceptable way for your business model.
Ultimately, NCBL is an uncertain, temporary administrative process to ensure that the lights don’t go out. It’s not ideal from an efficiency or an economic perspective.
How are data centers shifting their strategies for securing power? Where are we in that evolution?
Alison Clements: They’re certainly getting hip to the challenge. You’re seeing a lot of the data center developers who’ve been doing this for a long time hire C-suite-level energy roles, so people are coming in with power sector experience.
I think there’s a real opportunity for forward-leasing data center developers and customers to find a partner utility and start working on [flexibility options]. Flexibility can mean a lot of things. It can mean curtailing, like the NCBL, or it could mean finding ways to get a better sense of the utilization rate of an existing data center, such that any excess availability is shared.
Developers are certainly considering lots of backup power options and bridging options. So if we want to be attached to the grid with five nines reliability, what can we do for the next three years? You can’t run diesel generators all the time. So what kind of gas generation can be run in combination with batteries so I can get 70 MW online?
And then in three years the utility promises me another 150 MW, but then the question becomes, what happens to that 70 MW? Are you still using them? Can you sell them into the grid? Can you move into another site in the market? So it’s an exciting time relative to the opportunity for innovation in the market.
