In partnership with
⚡ THIS WEEK'S SIGNAL
Enverus now projects 62 GW of new U.S. data center capacity through 2030, with roughly 40% of it going behind the meter. The number everyone will quote is the 40%. The number worth sitting with is the fuel bill attached to it — about 1.3 Bcf/d of incremental gas demand by 2030 — which implies far less generation actually running than "a quarter of the buildout goes off-grid" would lead you to expect. Either the bridge is thinner than advertised, or the forecast is conservative. How that resolves decides a lot.
━━━━━━━━━━━━━━━━━━━━━━━
📰 MAIN STORY
Enverus Intelligence Research finally put a number on something the industry has been describing anecdotally for two years. Its 2Q26 data center capacity forecast projects approximately 62 GW of new Lower 48 data center capacity through 2030, supported by roughly $5 trillion in cumulative hyperscaler capital expenditures from 2026 through 2030. About 40% of installed capacity lands behind the meter. Interconnection queues and permitting timelines are doing what constraints do — routing load around themselves. The geography is unsurprising: PJM, ERCOT, MISO and WECC together account for more than 80% of forecast additions, and near-term visibility is unusually firm, with roughly 96% of forecast 2028 additions already tied to projects under construction or in advanced development.
Everyone in this picture is optimizing for something different from what the press releases say. Hyperscalers are not optimizing for cost, and at this point they are not optimizing for carbon either — they are optimizing for schedule certainty, because a data center that energizes eighteen months late is a competitive loss that no PPA discount recovers. BTM developers like VoltaGrid are optimizing for contracted offtake that lets them finance equipment ahead of demand; the company reported roughly a 7.5 GW order book through 2030 alongside its $1 billion equity raise from Blackstone Tactical Opportunities and Halliburton. Upstream gas operators are optimizing for fixed-price offtake that strips out basis risk and lets them capture margin on power rather than molecules. And utilities are optimizing to keep the load on their books, because load that leaves the system does not stop needing the system.
That last point is where the arithmetic turns uncomfortable. GE Vernova closed the second quarter with 116 GW of gas equipment backlog and slot reservations, up from 100 GW three months earlier, against annualized output tracking toward 20 GW this quarter and 24 GW by 2028. The turbines going behind the meter come out of the same queue as the turbines utilities need for the grid everyone else is still connected to. Meanwhile RRA forecasts $1.295 trillion in aggregate energy utility capex across 46 companies for 2026–2030 — spending that enters rate base and gets recovered from the customers who remain. One side buys schedule certainty. The other side receives it as a line item.
Here is the number worth screenshotting. Run 1.3 Bcf/d through a heat rate between 7 and 8.5 MMBtu/MWh and you get somewhere near 6.5 to 8 GW of continuous output. Forty percent of 62 GW is roughly 25 GW. Data centers run close to flat. Those two figures do not obviously reconcile, and the ways they might reconcile matter: BTM plants running as bridge and peaking rather than full baseload, hybrid configurations with retained grid backup, nameplate-versus-average-draw effects, or "incremental" excluding gas that would have burned somewhere anyway. Each explanation implies a different decade.
Scenario A: The RTO and transmission owner responses landing at FERC by August 17 produce credible fast-track large-load frameworks in PJM and MISO — minimum financial commitments, flexible service, studies inside 60 to 90 days. Grid interconnection becomes schedulable again, BTM share plateaus near the Enverus base case, and the 1.3 Bcf/d figure holds as a genuine bridge. Gas developers keep the order book but lose pricing power.
Scenario B: The August 17 filings mostly defend existing tariffs, and more states follow New York's Executive Order 62 with permitting pauses that make grid-connected siting slower than building your own plant. BTM stops being a bridge and becomes the primary delivery mechanism. In that world 1.3 Bcf/d is a floor, not a forecast, and the stranded-cost question moves from theoretical to the front of every rate case in PJM.
What to watch: August 17, 2026 — whether PJM files a Section 205 tariff revision or defends its existing tariff. That single choice tells you which scenario you are in.
━━━━━━━━━━━━━━━━━━━━━━━
Reply to everything. Edit nothing.
Your inbox is full. Slack is piling up. Client messages need a response yesterday. Typing thoughtful replies to all of it takes hours you don't have.
Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Speak like you would to a colleague — tangents and all — and get polished output. Emails, Slack, LinkedIn, WhatsApp, whatever's open.
89% of messages sent with zero edits. Used by teams at OpenAI, Vercel, and Clay. Works on Mac, Windows, and iPhone.
━━━━━━━━━━━━━━━━━━━━━━━
⚡ QUICK HITS
New York Became the First State to Pause Hyperscale Permitting: Governor Hochul signed Executive Order 62 on July 14, halting state environmental permits for data centers at 50 MW and above for up to one year while the Department of Public Service produces a Generic Environmental Impact Statement. The legislature's broader Responsible Data Center Development Act, which uses a 20 MW threshold, remains unsigned — meaning NYISO developers are currently navigating an executive pause and an unresolved statutory one at the same time. [Governor's Office]
GE Vernova's Backlog Grew Another 16 GW in a Quarter: Gas equipment backlog and slot reservation agreements moved from 100 to 116 GW in Q2, with the company guiding to at least 125 GW by year-end and total backlog at $176 billion. If you are planning a project that needs a heavy-duty frame after 2029, the slot is now a harder-won asset than the site. [SEC Form 8-K, July 22]
The FERC Resource Adequacy Reports Are Already In: The show cause orders required each RTO and its transmission owners to file a separate resource adequacy report within 30 days — a deadline of roughly July 20 — describing how they will ensure sufficient generation for new and existing large loads. Those filings landed weeks before the August 17 tariff responses and are the closest thing to a utility-native admission of where the capacity actually is. [Bracewell]
Twenty-Seven States Are Writing Their Own Large-Load Rules: MultiState counts 27 states with active large-load legislation, with California, Ohio and Utah already enacted and thresholds ranging from 10 MW in South Dakota to 150 MW in Alabama. A developer running a multi-state siting analysis is now pricing regulatory divergence, not just power cost — and the 10 MW floors capture facilities that no federal framework touches. [MultiState]
Public Opposition Now Runs Ahead of Nuclear: Gallup's March 2026 survey found 71% of Americans oppose an AI data center in their local area, including 48% strongly opposed, against 53% opposing a local nuclear plant. Water and energy use were cited most often at 18% each — which puts community resistance squarely on the two inputs that behind-the-meter gas increases rather than resolves. [Gallup]
━━━━━━━━━━━━━━━━━━━━━━━
🔧 TOOL / RESOURCE OF THE WEEK
FERC eLibrary — the six large-load show cause dockets: FERC opened a separate EL docket for each jurisdictional RTO when it issued the June 18 show cause orders, and every response, protest and intervention lands there in public view. Between now and August 17 this is the single highest-signal place to watch regional divergence form in real time, because the filings arrive before the trade press writes them up. → [FERC eLibrary — search by RTO name]
━━━━━━━━━━━━━━━━━━━━━━━
💬 CLOSING THOUGHT
The honest read is that we have a very confident capacity forecast attached to a comparatively modest gas demand forecast, and the industry is about to spend a decade and several hundred billion dollars in the space between them. For the optimistic case to hold, the August filings need to make grid interconnection schedulable again. That is a narrow thing to be hoping for.
So I'm curious what you're seeing. Is behind-the-meter still being sized as a bridge on the projects you're touching? Or has it quietly become the plan?
