Weekly Dispatch

Wherever data centers go, batteries will follow

In last week's Dispatch, I considered the US's evolving battery energy storage map - and had a look at where batteries are being built. The more I think about it, the more I realize there's a deceptively simple answer: they're popping up wherever data centers are.

Follow the (large) load - whether that's in PJM or ERCOT - and you'll see new batteries.

And that load is extremely large. In PJM, data centers aren't just one part of demand growth. They basically are the demand growth.

Strip out data centers and PJM's peak demand would actually be shrinking - at least until 2032. With them, peak climbs about 97 GW by 2046 - and roughly 78% of that growth is large load, overwhelmingly data centers.

To learn more about how Modo Energy expects the US battery energy storage market to evolve nationwide, join our livestream next Thursday.

And this growth isn't even particularly speculative. 8.2 GW of data centers are already operating in PJM, with another 41 GW contracted or under construction - a fivefold jump already locked in.

Geographically, it's pretty intensely concentrated: Dominion's Northern Virginia zone alone accounts for a third of the data-center load PJM has accepted through 2030.

And - surprise surprise - those same zones pulling in the data-center load are exactly where battery developers are now racing to connect. In PJM's latest interconnection cycle, 82 batteries totaling 8.5 GW just cleared a key phase - and Virginia holds the single biggest cohort.

Why cluster there? Increasingly, the capacity market. PJM capacity prices have hit the cap in each of the last two auctions and - with large-load growth showing no sign of easing - are expected to stay there. That hands batteries a large, dependable revenue stream, and it's richest right where the data centers are landing. (The energy-arbitrage case is murkier - big new loads could widen spreads or compress them - so for now it's capacity, not arbitrage, doing the heavy lifting.)

That capacity revenue also shapes what gets built. Accreditation is duration-dependent - a two-hour battery earns almost none of it, which is why there's so little short-duration in the cohort, and it scales up the longer you go. Four-hour systems dominate this cycle, but nine ten-hour projects are now moving through.

But that doesn't mean you should expect to see all those batteries come online.

Batteries drop out of PJM's queue faster than any other tech - a 62% attrition rate by project count, against 50% for non-storage. And that's often down to the cost of connecting, which comes from affected-system reinforcements: grid fixes needed well beyond a project's own point of connection.

Interconnection charges range from about $3,000 to $1.9 million per MW (with a median near $151k), and some projects saw their bill jump ten- to seventy-fold between studies - with one's costs leaping from $0.85 million to $62.7 million.

With those economics, expect to see more queue dropouts.

Frankly, this feels backwards. Batteries can act as a major net positive for congestion - they ease the very constraints those network upgrades are built to relieve - so making it prohibitively expensive to connect doesn't make much sense to me.

If PJM is just starting to heat up for batteries, ERCOT shows what happens to a BESS market as it matures.

Texas has built more grid-scale batteries than anywhere in the US - 16.5 GW and 28.8 GWh by June - but after roughly doubling year-on-year until now, additions are leveling off. In fact, annual fleet growth has fallen from around 100% toward 40%.

The pipeline still looks enormous on paper: ERCOT's queue implies 64 GW of batteries by 2032. But Modo Energy's central forecast foresees 40 GW - although the size of that gap will likely depend on how much new load actually shows up.

Enter data centers. Modo Energy forecasts an additional 35 GW of demand at peak on the ERCOT grid by 2035, driven by new data centers - more than a third of ERCOT's all-time system peak, itself only just set at around 91 GW last week. That flat, round-the-clock load is a major reason stalled queue capacity could move back into the buildable pipeline.

The rules that decide whether that data center load connects are now taking shape. Texas's Senate Bill 6 (signed June 2025) makes large loads pay to play - study fees, non-refundable interconnection fees and deposits - and requires any site energized after the end of 2025 to accept remote curtailment by ERCOT in emergencies.

On top of that, the PUCT's new “Batch Study” process (approved June 2026) replaces one-by-one studies with a single system-wide study of every load of 75 MW or more, with results of the first batch due around spring 2027. The net effect: a higher bar that should weed out speculative entries - but also a hard gate before megawatts move from queued to buildable.

With that uncertainty - plus a 70%-plus slide in battery revenues since 2023 and longer, costlier build timelines - attrition rates are growing for batteries.

A record 13.7 GW left ERCOT's queue in the first half of 2026 - including 2.3 GW that had already signed an interconnection agreement.

That isn't down to demand. It's the difficulty of financing projects while it's unclear when - or whether - merchant revenues will recover to investable levels.

But stronger demand - say, from data centers - could certainly change this picture.