Newsletter · · Ashutosh Agarwal
FERC Puts Six Grid Operators on a 60-Day Clock - Powering AI: Grid, Gas, Generation & Nuclear - Week of June 30, 2026
FERC declined to write one national data-center rulebook and instead fired show-cause orders at six RTOs with a 60-day fuse, the bull case for grid, nuclear and uranium got its fullest articulation in years, and the sharpest pushback landed on the demand forecasts the whole trade rests on. Our synthesis for the week of June 30, 2026.
Powering AI: Grid, Gas, Generation & Nuclear
Week of June 30, 2026: FERC Puts Six Grid Operators on a 60-Day Clock
Powering AI: Grid, Gas, Generation & Nuclear, Tuesday, June 30, 2026
For two years the bottleneck story has been a slow grind: too few transformers, too many years in the interconnection queue, everyone waiting on someone else. This week the regulators stopped waiting. FERC put six grid operators on a 60-day clock, a marquee NextEra transmission line is mired in a fight over who pays, and the uranium crowd is quietly saying the part they usually whisper out loud. The supercycle didn't get louder this week. It got more contested.
TL;DR
- FERC declined to write one national rulebook for data centers and instead fired show-cause orders at six RTOs, a real, dated catalyst with a 60-day fuse, not another think-tank white paper.
- The bull case (grid equipment, nuclear, uranium re-rating together) got its fullest articulation in years, but the most specific pushback this week was on the demand forecasts the whole trade rests on.
- Uranium's fuel-cycle insiders now say the bottleneck has moved to the front end. The long-term price is $94/lb and climbing.
What's new
FERC stops dithering and starts the clock. The single biggest development of the week was regulatory. On the POLITICO Energy podcast, FERC Chair Laura Swett walked through the decision her commission made June 24: rather than impose one national co-location rulebook, FERC issued show-cause orders to six regional grid operators, PJM, CAISO, NYISO and SPP among them, to justify or rewrite their data-center interconnection and cost-allocation rules. SPP is the standout, already in with novel proposals; PJM is "under the 206 proceeding for co-location." The deadlines are tight, 60 days, or a 45-day abeyance if an RTO files a 205. Two tells that this is a serious operator and not a headline-chaser: she timed the release after market close specifically so it wouldn't "tank a stock," and she stayed deliberately inside federal jurisdiction. This is the framework that decides whether hyperscalers can co-locate behind the meter and how fast, the thing the entire PJM merchant complex has been waiting on.
The line that explains why none of this is easy. On Catalyst with Shayle Kann, Latitude's Maeve Allsup dissected NextEra's Mid-Atlantic Reliability Line, ~100 miles, just under $1 billion, approved by PJM back in 2022 when ChatGPT didn't yet exist. It is now stuck across four state commissions, with consumer advocates wielding the hyperscalers' March 2026 White House Ratepayer Protection Pledge as a cudgel and Maryland filing a fresh May 2026 FERC complaint to dump the data-center transmission cost onto Dominion's zone. Kann's framing is the one to internalize: "power is the rate limiter to AI growth." The proof point is brutal: the US built nearly 4,000 miles of high-voltage transmission in 2013; lately we average hundreds, and need thousands. The wire, not the chip, is the governor.
Washington writes a check for big nuclear. On Motley Fool Hidden Gems Investing, the team unpacked fresh federal loan support for five large nuclear projects, funneled through special-purpose vehicles. The AP1000 is "really your only viable option... the only licensed option," and its parent is Cameco (CCJ), which owns Westinghouse 50/50 with Brookfield. Constellation (CEG), the largest US nuclear operator, already supplying Walmart, is the frontrunner for new units. GE Vernova (GEV) gets the turbine pull, though its gas-turbine backlog already runs "out until 2031." The framing number, courtesy Goldman: data-center power demand grows 160% by 2030, with ~$1.3 trillion of hyperscaler capex going to power generation and related infrastructure.
Uranium's "day in the sun." Two uranium heavyweights surfaced the same thesis from different angles. On The KE Report, Uranium Insider's Justin Huhn relayed UXC president Jonathan Hinze's comment at the World Nuclear Fuel Markets conference: capacity grows from ~400 GW to ~500 GW in five-to-six years on what's already known, and "I'm concerned about uranium," the bottleneck has shifted off conversion and enrichment and onto the front end. The long-term price sits at $94/lb, up $14 over the year, with ~150 million pounds contracted in roughly ten months. On Money of Mine, Sachem Cove's Mike Alkin made the structural version: price from $17 to over $100 in the contract market, ~430 operating reactors plus ~80 under construction, and "not enough economically viable uranium to feed that base," with major mines running dry in five to seven years. He doesn't even need the SMR or data-center story; the existing fleet does the work.
Cameco's CEO on why utilities won't go first. On Energy Evolution, Grant Isaac gave the most honest read on the new-build logjam. Vogtle's first AP1000 cost ~$18 billion, the second ~$12 billion: "the problem wasn't building the first two units... the problem was we stopped." In the US, where there's no energy minister to simply decree it, Constellation's Joe Dominguez building two reactors means telling shareholders "we're going to build the next two most expensive ones," a "race for the bronze medal." The fix in the strategic partnership: have the government order long-lead kit for up to ten reactors up front. Watch Duke (DUK), which already holds two AP1000 operating licenses, with North Carolina advancing a bill to bar baseload retirements until a 1+ GW nuclear plant is approved.
The debate
Bull. The cleanest articulation came on Animal Spirits, where First Trust's grid strategist argued we are in the "early innings" of a buildout that adds, per Bloomberg, ~17 million miles of transmission and distribution over 25 years. The list of beneficiaries is the supercycle in one breath: Quanta (PWR), Eaton (ETN), Schneider, ABB, Johnson Controls, the cable maker Prysmian. His best point: this isn't purely a data-center bet, reshoring (his hometown Micron fabs), grid hardening and plain old aging infrastructure carry a multi-year backlog even if AI cools. Pair that with the uranium deficit and you have grid, generation and fuel re-rating together.
Bear. And yet the most specific argument this week was aimed at the foundation. On the Energy Capital Podcast, a PJM-steeped guest noted Dominion projects ~70 GW of new large-load capacity against a ~24 GW system peak, numbers an independent market monitor flatly called "fantasy." The mechanism is "phantom load": developers bid the same project into many jurisdictions, utilities feed unvetted supplementals into PJM, and the aggregate triple-counts. Reinforcing the bear: on The Energy Gang, Wood Mac's Ed Crooks laid out $1.4 trillion of planned utility capex against $31 billion of rate-increase requests and a warning that bills could climb ~40% over five years, exactly the political backlash that delays rate-base recovery and stalls lines like MARL. And on SunCast, the pitch that virtual power plants can deliver capacity 20-40% cheaper and ~3x faster than gas peakers is the efficiency-caps-the-move argument in a sentence. Even the Motley Fool bulls asked the right question on their own names: does a few reactors actually move the needle for CEG or GEV, or does "the nuclear industry need these companies more than these companies need nuclear"?
Read-throughs
If the wire takes 7-10 years and a gas turbine takes two, the load goes to gas, Barry Cinnamon spelled that out on The Energy Show, and it's the read-through to on-site gensets (CMI, CAT) and the turbine OEMs whose backlogs (GEV to 2031) are both the bull case and the bottleneck. The same delay funds the copper, electrical-steel and cable complex (FCX, Prysmian) that the grid build consumes regardless of which generation source wins. On the financing side, Tech Disruptors priced it cleanly: investment-grade hyperscaler off-take lets these deals fund at roughly SOFR + 250-350 bps, while neoclouds pay S+450-500, the off-taker's credit is the deal, and private equity is pouring into substations, switchgear and the cabling underneath. The uranium read-through is the longest-dated: an ex-Russia fuel cycle takes two-plus years end to end, which is why utilities are contracting 2032-2033 pounds today and why the converters and enrichers, not just the miners, sit on the critical path.