The new wave of DERs

The arc Guernsey describes has three stages. Before aggregators there were utility interruptible programs: manual phone calls to very large industrial customers, who would cut their usage during a crisis. She calls it a fire extinguisher, often not used at all. Then aggregators arrived, EnerNOC among the first, and with them the idea that a technology platform plus a portfolio could do something no single site could. Then came rules, in particular Federal Energy Regulatory Commission orders 719 and 745, which she says started to value a megawatt of demand reduction the way a megawatt of supply is valued. The end state is the one she leads with, and she is only half joking that it makes the topic boring: on her platform at any given moment there are somewhere between one and twenty dispatches running, and it has been dispatched on 365 of the last 365 days. When she posted that, the response was surprise, which is the tell that the outside view of this market is still the fire extinguisher.

What broadened was not one thing. There is more equipment with an automatic on-off switch and an internet connection; more use cases, having graduated from capacity emergencies to economic dispatch, balancing renewables, ancillary services (the fast-response products grid operators buy to keep supply and demand matched), local congestion management and even carbon-driven dispatch; and more need, because prices are rising and load is growing. The participant list now runs to more than 50 verticals within commercial and industrial plus residential, from an electric vehicle in a driveway and a smart thermostat through retail, big box stores, school districts, wastewater treatment plants and commercial real estate, up through paper and steel mills, crypto miners, and AI data centers at the top. The heterogeneity is the point rather than a complication: assets with different operating constraints get paired so that the aggregate can answer what the grid asks, which she describes as Tetris. The underlying resource is still largely load control on conditioned space, now increasingly alongside behind-the-meter assets, with battery storage one of the fastest-growing categories as its costs fall.

Ask what that earns and she answers “it depends,” twice, and then gives a trend rather than a rule. Value is smoothing out. The hottest, peakiest summer day as the whole ballgame belongs to the earlier era; there are winter peaks now, and shoulder-season tightness caused by generator outages, and reserve prices can actually be higher in the shoulder seasons precisely because other resources are short. Capacity and reserves are both effectively call options, paid for being available rather than for energy delivered, and reserve markets clear daily. Customers who participate economically curtail whenever prices cross their own threshold, or strike price, which clips price spikes rather than calendar peaks. On top of that, traditional capacity programs are now being called over and over; she has lost count of how many times. She likes the frequency for two reasons. It spreads the payments, and it exercises the muscle, because the thing that actually matters to her is that when the grid calls, the resource shows up.

That leaves the question of what is holding it back, and here she declines to name a single binding constraint: it is all of the above. Higher prices widen the addressable set because more things pencil. Deployment simply takes time. Market rules move, and the capacity accreditation changes in PJM meant more megawatts showed up but were counted differently. And the one she would fix with a magic wand is data access: smart meters were deployed and put into rate base, and in a shocking number of territories the data is still hard to get, even for the customer whose meter it is. The new product follows from the same logic. If the constraint is capacity in a particular zone, define the constraint geographically, build an incremental virtual power plant inside it, run the resulting megawatts through the market’s own accreditation process, and transfer the accredited capacity to the utility with no money changing hands there, because the data center funds it. On her account that is a win all round: the utility gets capacity it can count, the load gets permission to site, and ratepayers do not see the cost on their bill and can opt into earning from the program themselves. The reframe underneath it is that co-location need not mean the same land parcel; unless the physical constraint is something like a substation upgrade, co-locating at the constraint is what solves the problem, which opens up the whole toolkit. And her closing argument is explicitly one about market design rather than economics: this is a utilization problem, capacity built for a sliver of hours, so even if you could buy gas plants off the shelf tomorrow, using what is already there would be the better outcome.