Whew. Milwaukee, right? Wow. If you don’t know what I’m talking about, reply and we’ll talk MLB playoffs. ⚾

Yesterday we published our first Currence Research piece using the Currence Data Center Power Economics Model. We use the model to assess what the will-they-won’t-they financing situation at SB Energy’s PORTS-Pike does to project economics.  Clients can read it here.

Reply to this email for a demo or to evaluate the cost of a specific project. We are also always improving our assumptions and benchmarks, so if you work on these projects and are open to helping us further verify them, we want to hear from you.

This is Currence’s weekly newsletter on the moves and motives shaping the load growth era. Not a subscriber yet?

FERC short-circuits PJM’s capacity fix

What happened

On 9 October, FERC rejected the bulk of PJM’s plan to manage data center load growth amid its capacity shortfall. The plan responded to FERC’s show-cause order, issued June 18 after a Department of Energy directive last October pushed FERC to take up large-load interconnection reform.

PJM’s answer was a package of six reforms:

1. A definition of “New Large Load” as 50MW or more of cumulative peak demand at a single site, with a one-mile radius rule bundling affiliated (common ownership) facilities together. I.e. a developer, Joe P. Owerstack, shows up hoping to build three 20MW data centers a half-mile apart. This definition bundles them into a single 60MW Large Load.

2. A Large Load Registry to track who’s building, where, how much capacity they need, backup generation, and contract details. I.e. Joe and his data centers’ 60MW get entered into the registry that PJM maintains, so a state commission could look and see what’s coming online in its territory and plan for it.

3. A capacity-designation requirement, so load-serving entities (LSEs) serving a New Large Load have to bring new capacity equal to that load’s peak demand. I.e. Joe’s utility, the LSE that sells him power, has to bring 60MW of new generation, storage, or demand-side resources to match his load, or Joe’s utility faces curtailment risk on his load. (Backup generation alone doesn’t count, BYONC has to clear PJM’s capacity market first.) In practice, that cost and obligation likely gets passed straight to Joe through his contract. 

4. BYONC, Bring Your Own New Capacity, the mechanism to satisfy that capacity designation. New generation, storage, or eligible demand-side resources clear PJM’s gating criteria for newness, then get accredited at a fixed capacity value (it’s ELCC-based UCAP) for the life of the arrangement. I.e. Joe builds his own 60MW gas plant next to his data centers, bids it into PJM’s capacity market, and once it clears, designates it as BYONC to cover his load.

5. IRAS, Interim Resource Adequacy Service, means that any New Large Load that didn’t bring BYONC is subject to curtailment, triggered once PJM declares its first-stage emergency alert, ahead of voluntary demand response. Example: if Joe never builds his gas plant or doesn’t get it accredited, during a summer heat wave, PJM could force Joe’s data center to cut power before it even asks demand response customers to volunteer.

6. Exclusion from the Variable Resource Requirement (VRR) curve, the demand curve PJM uses to determine how much capacity to buy each year. Starting with the 2029/2030 Delivery Year, any New Large Load MW that had not been covered by BYONC or backstop capacity would get excluded from PJM’s demand curve entirely, regardless of whether it was physically still drawing power. I.e. if Joe never built his gas plant and stayed uncovered, his 60MW isn’t in PJM’s planning numbers, even though his data center keeps running every day that isn’t an emergency. PJM only counts the demand it’s procuring firm capacity for, and it won’t procure capacity for load it could just cut off.

FERC accepted the first two tenets and rejected the next four on three grounds: jurisdiction, undue discrimination, and a flawed accreditation rule.

On jurisdiction, FERC said the filing’s language intruded on state authority over retail rates. PJM’s filing directed Electric Distributors to curtail a defined class of retail customers, Eligible New Large Loads like Joe’s data center, preventing states or Electric Distributors from picking differently, despite PJM’s claim that it was their decision. Under IRAS, as well, Joe would have been compensated through charges on other retail customers in the same zone, which functions as a retail rate. But FERC only regulates wholesale rates, the sales utilities make to each other, and states decide who gets cut off and who pays.

On discrimination, the regulator disputed PJM’s reasoning that Joe’s new 60MW load poses a different reliability risk than an existing 60MW load, and that data centers specifically warrant different treatment than any other large load, pointing to lack of evidence.

On BYONC, locking Joe’s gas plant at its first-year capacity credit forever ignores that some resources, especially storage, lose capacity value over time, FERC said. That risks the market counting megawatts years from now that no longer exist. Because IRAS, capacity-designation, BYONC, and the VRR exclusion are all connected, FERC rejected them outright.

This also lands ~5 weeks after FERC knocked down PJM's backstop capacity auction, so PJM has now lost both of its main tools for the capacity shortfall. PJM is now staring down the barrel of its next base residual auction. It’s still fixable, as FERC basically said the concept in PJM's transmittal letter might be fine, but the actual tariff language overreached, so PJM can refile.

Mark’s take

I keep coming back to one project.

Aligned Data Centers has this project in Hillsboro, Oregon, where it paid for Calibrant Energy to add a 31MW/62MWh battery, that it runs to Portland General Electric’s peak schedule. So, effectively, the grid just sees less of the load. And if the data center needs to curtail, the battery is there. It’s elegant. It’s clever. It’s one of those things that helps a utility and regulator get to yes.

PGE agreed, and approved the 100MW connection years ahead of the transmission upgrades. And PGE’s customers got a reliability asset without grid upgrades, an operator program, or a FERC filing. To me, this model is something that should be rinse-and-repeat across the country, anywhere the utility owns the chain.

Given that, I’m not all that convinced losing IRAS is such a terrible thing. To deal with capacity shortfalls in PJM, players across the board are getting creative, and going around the ISO.

This week, the bypass has shown up in at least three places:

  • Utilities buying capacity directly. AEP signed up for as much as 1.6GW of Mainspring linear generators (see Meter Reading below). To me, this is fascinating – that the pull for power is so great that companies with once exotic techs, like Mainspring’s linear engine, that had been hanging out in the bleachers are being pulled down to the field to play in the playoffs. But it could work – it skips the turbine queue, the wait for the PJM market to send a price signal, and it can even get past some environmental hurdles that hold back other gensets.

  • Data centers paying others to flex. Voltus, who we interviewed here a few months ago, announced a $225m raise this week to sell data centers flexibility pooled from other customers in their region. The data center takes those MW’s and hands them to its utility to offset its own load.

  • Data centers flexing themselves. This is the Hillsboro model mentioned above. Okay, maybe this didn’t happen out in the market this week (at least not announced yet!), but on Tuesday we put a size on this market. In our analysis we put this market for storage for load-flex-for-interconnect at 285GWh across ERCOT, SPP, PJM, and MISO by 2035.

Hillsboro worked because PGE owns the chain – the generation, the wires, the customer relationship. In PJM, that model translates to vertically integrated utilities like Dominion in Virginia. But in Pennsylvania and Ohio, utilities own only the wires, so a battery can solve the transmission bottleneck while capacity still goes through PJM.

So, when it comes to load, the result of all this is that it’s pushing PJM more in the direction of being a scorekeeper or umpire. The two pieces FERC approved were the large-load definition and the registry. PJM tracks who connects, declares the emergencies, and accredits what everyone brings. And it still runs the auction that pays Constellation, NRG, Talen, and Vistra, but it’s no longer the one solving the shortfall. If I’m at PJM, I really can’t decide if I’m breathing an exhale of disappointment, or a sigh of relief – maybe a bit of both.

Going forward, I’m watching for two things. The first is the shape of the PJM’s refile. Does it look like Hillsboro? i.e., a voluntary option with firm service as the default. The other is whether Dominion and the Virginia’s State Corporation Commission offer flexible connections first, like PGE did in Oregon. If they do, we’ve got Hillsboro in PJM, in Data Center Alley, refile or not.

Who this helps

Storage developers with sites near planned data centers. If the Hillsboro model goes national, ready-to-build battery projects become the scarce asset, and our analysis puts the prize at 285GWh by 2035.

Vertically integrated utilities in PJM, like Dominion. Owning the generation, the wires, and the customer relationship means they can run the Hillsboro playbook (data center pays for a battery, utility nets the load) without waiting on PJM’s refile.

Existing IPPs, like Constellation, NRG, Talen, and Vistra. Jefferies said the ruling points toward higher capacity prices for longer in PJM, since there’s no quick fix for the shortfall now that IRAS is dead, and that’s a direct tailwind for generators already selling into the market.

Who should be nervous

Wires-only utilities in Pennsylvania and Ohio. They can’t replicate Hillsboro since they don’t own generation, so capacity still has to clear PJM’s market, leaving them stuck with the shortfall FERC just told PJM to find another way to handle.

PJM itself. It lost its backstop auction five weeks ago and now its large-load capacity tools too, leaving it to track registrations while states decide who actually gets cut off.

Meter reading (9 - 15 Oct)

A quick read on the numbers shaping the market. The capex, the contracts, the regs, all anchored in the so-what.

2.6GW // Mainspring and Enerflex landed two large power equipment orders, and neither is a frame turbine. AEP signed Mainspring for up to 1.6GW of linear generators through 2034, with the first 190MW firm for 2027. Enerflex won a contract for over 1GW of packaged gas-fired units for an undisclosed data center developer, pushing its behind-the-meter backlog past 1.75GW through 2028. GE Vernova, Siemens Energy, and Mitsubishi Power’s frame turbines are sold out for years, and both buyers get to skip the line.

$72bn // Emera and Canadian Utilities agreed to merge as equals, creating a Top 20 North American utility with roughly $45bn in rate base and six million customers across Canada, the US, and Australia. The combined company plans $32bn in capital spending through 2030, with 80% of post-merger earnings concentrated in Florida and Alberta. It is the largest merger in Canadian corporate history, and it runs on the same logic driving NextEra’s pending Dominion acquisition: that scale in a couple of fast-growing jurisdictions funds AI-era load growth.

75% // FERC accepted Duke Energy Carolinas’ large-load tariff, 4-1. New loads of 100MW or more pay at least 75% of their scheduled monthly demand for 15 years after network upgrades go into service, whether or not the load materializes. This is the first FERC-accepted take-or-pay floor for large-load transmission. Duke didn’t have to post matching financial security from the loads themselves, so the floor protects the utility’s capital, not ratepayers.

2.1GW // PJM approved its first three Expedited Interconnection Track projects, and four-fifths of it is storage. The track takes projects above 250MW that can reach commercial operation within three years and caps at 10 requests a year. Engie has 860MW in Crawford County, Pennsylvania and 800MW in Morrow County, Ohio, both batteries. LS Power is adding a 455MW uprate to its 810MW Hunterstown combined-cycle plant in Adams County, Pennsylvania. All three are expected online by mid-2029, 18 months after the track itself sunsets at the end of 2027. 

$225m // Voltus raised a Series D to scale bring-your-own-capacity for data centers. Generation Investment Management led, with Activate Capital and Vitol co-leading, and it’s targeting 20GW under management by 2030, up from 8.5GW today. It’s Voltus’s first raise since a $31m Series C in 2021, and it comes with Google already signed as a BYOC customer in PJM, worth watching who signs next.

Explore more Signals on Currence here.

On the docket

The policies, rulings, and company moves worth watching.

DOJ asks a court to send Ørsted’s Revolution Wind’s approval back to Interior. A new filing has the  Interior agreeing to a voluntary remand of its 2023 construction approval for further review, even though the project is already delivering power to the grid. The filing is from an existing suit from anti-wind group Green Oceans, but the Interior’s position is opening up new questions.  

The Senate’s permitting bill heads for a post-midterm floor vote. The Bipartisan American Affordability and Jobs Act, a 417-page package from Sens. Capito, Whitehouse, Lee, and Heinrich, sets a 150-day deadline for legal challenges to environmental reviews and promises already-permitted projects stay permitted, with a vote planned for early November. But Revolution Wind’s new filing is a test for pro-clean energy Senators.

New & upcoming at Currence

The latest research, features, and data drops on the Currence platform.

What speed-to-power is worth. A new case study runs SB Energy’s PORTS-Pike campus in Ohio through Currence’s Data Center Power Economics Model, exploring what the activation date does to $/MW. It shows why a credible early activation date beats a chip upgrade. Read the case study.

Load-flexibility-for-interconnection is a 285GWh storage opportunity. A new analyst take argues short-duration batteries will beat backup generators as the default tool data centers use to meet curtailment obligations under load-flexibility-for-interconnection rule. Currence models flexible-load storage deployment by 2035, with upside for pre-positioned developers and OEMs. Read the analyst take.

Events

Where the market is meeting, and where to find us

📅 Norrsken Impact/Week // Barcelona, Oct 14-15 // Founders, funders, and policymakers gathering on climate, energy, and impact investing. 

📅 RE+ // Las Vegas, Nov 16-19 // North America's largest clean energy event, covering solar, storage, hydrogen, and grid edge tech.

The B-side

I’m guessing if you’re reading this, you have a favorite Tiny Desk Concert. I do. Or at least a top five. Today, it’s one of those – Valley Queen. I love how our song this week, Hold on You, just moves forward like a train and just keeps on moving. So good.

But I must admit, my favorite in-studio performance of all time is this one.

Full Powerstack B-Sides on Spotify.

Interested in diving deeper? Talk to our team and learn how the teams that finance and build the energy system use Currence to stay ahead in the energy and AI buildout, including Southern Company, Tokyo Gas, Jefferies, Galvanize, B Capital, and others.