Newsletter · · Ashutosh Agarwal

Utilities Plan a $1.4 Trillion Grid Buildout as the Affordability Backlash Builds - Powering AI: Grid, Gas, Generation & Nuclear - Week of September 11, 2026

Powering-AI podcast roundup for the week of September 11, 2026. US utilities now plan $1.4 trillion of capital spending through 2030, grid equipment costs have roughly doubled, and a bull case for a multi-decade supercycle met a bear case built on interconnection queues and a ratepayer affordability backlash.

Powering AI: Grid, Gas, Generation & Nuclear

Week of September 11, 2026: Utilities Plan a $1.4 Trillion Grid Buildout as the Affordability Backlash Builds

This week the buildout stopped being a story about demand and started being a story about a bill. Utilities are now telling their investors they plan to spend $1.4 trillion by 2030, and in the same seven days, the pundits, the regulators and even the President all started asking who pays for it. The supercycle is real. So, suddenly, is the pushback.

TL;DR

  • The number of the week is $1.4 trillion. That is what US utilities are now planning to spend through 2030, up 21% in a single year. It landed the same week the White House declared a power-equipment emergency and the same week the public turned, hard, against data centers.
  • Everything that carries electrons got roughly twice as expensive. Conductor, transformers, switchgear: all up about 2x; gas plants up 2-3x; transformer lead times out to three to four years. Whoever owns the bottleneck owns the pricing power.
  • The bull and bear cases both showed up this week, and both were good. One side sees a multi-decade demand super-cycle. The other counts 2.5x more interconnection applications than the grid can actually build, and says a lot of this never gets constructed.

What's new

The capex number just jumped again, and it's now a political problem. On Open Circuit, a guest from the research shop Power Lines debuted a new study (it was picked up by the Wall Street Journal that morning) putting planned US utility capital spending at $1.4 trillion by 2030, a 21% jump over last year's roughly $1.1 trillion figure. The comparison they reached for: "three interstate highway systems" and "thousands of Hoover dams." Why it moves the thesis: for regulated utilities, capex is the earnings engine: spending goes into the rate base and comes back as regulated profit. More approved spend means more earnings. But the same episode flagged the catch that runs through everything else this week: utilities requested a record $31 billion of rate increases last year, power prices are up 40% over five years, and roughly one in three Americans now struggle to pay a utility bill. As the guest put it, the whole regulatory system "structurally rewards CapEx over OpEx": it pays utilities to build, not to optimize. That is great for rate-base growth right up until voters decide they have had enough.

Washington called it an emergency. On the Energy News Beat Podcast, host Stuart Turley, the writer known as Doomburg, and Jackery Americas CEO Stephen Wang walked through the Trump administration's new state-of-emergency order aimed at onshoring transformers and grid components. The hard numbers are the point:

"We have a five-year wait for large natural gas generators, we have a two- to three-year wait on transformers... If you lose a transformer, it may be 18 months."

Wang added that 60 to 80% of electrical components are currently imported, alongside a new ban on foreign-made inverters. Doomburg pointed to Project Kilby in Texas (Chevron, Microsoft and GE Vernova teaming up to run a big data center on behind-the-meter natural gas) as the template for getting load online without waiting on the grid. Why it matters: lead times this long are a moat for anyone who already makes the equipment, and a hard ceiling on how fast anything actually gets built.

"Every component of the grid" is now roughly 2x more expensive. Catalyst with Shayle Kann put specific multiples on the inflation:

"Conductor is like twice as expensive... Transformers are two plus times more expensive. Switchgear is two times more expensive. Gas power plants are two to three times more expensive."

The under-appreciated point: about two-thirds of all utility spending is just maintenance and hardening (replacing poles and wires at the end of their life), which has nothing to do with data centers but is getting more expensive because of them. On rates, Kann's read is nuanced and worth holding onto: at the local level, utilities now "hold all the cards," so most new data-center deals keep customer bills flat or even push them down (an EPRI study found data centers have been about 6% beneficial to bills so far). It's at the global level, where data centers are the single biggest driver of demand, that prices grind higher for everyone.

The fastest way to power is a fuel cell or a generator, not a wire. On Motley Fool Hidden Gems Investing, analysts Rachel and Matt framed Bloom Energy's edge in one line: "Bloom's moat is providing power quicker than the alternatives. It can deploy solutions in 90 days or less", against multi-year waits to connect to the grid. Bloom sits on a $20 billion backlog (only $6 billion of it product, the rest long-tail service revenue), trades at about 80x forward earnings after a ~500% run, and was just added to the S&P 500. Their honest caution: this is a company priced for the boom to continue. They contrasted it with Enphase's data-center push (an ~11 GW addressable market by 2031, but no volume until 2028 and a core home-solar business shrinking 20% a year), which Matt called "a 2028 option on a shrinking company." For anyone who doesn't want to pick the winning technology, Rachel pointed at the arms dealers: Eaton and Schneider Electric, "companies manufacturing the high voltage cables, transformers and switchgear... They don't depend on which specific chip architecture wins."

The debate

Both sides genuinely spoke up this week. Here's the steel-man of each.

The bull case: a multi-decade super-cycle that lifts the whole complex together. The clearest articulation came from the consultants and bankers at FMI, unpacking their first-ever energy & power outlook. The framing:

"We are in a pretty phenomenal moment in the history of the energy and power sector... over the next five years about a trillion dollars of capital planned to go into various energy and power subsectors."

Their numbers do the arguing. Annual transmission-and-distribution spend is heading from about $158 billion a year to roughly $255 billion, a 10% compound growth rate, "about double... what we've had over the last four decades." Their five-year-forward peak-demand projection has gone from 24 gigawatts in 2022 to 166 gigawatts, a 7x jump. Data centers move from about 3% of US electricity demand to roughly 9% in five years, and about a third of new data-center projects will "bring their own power" behind the meter, three-quarters of it gas. The investment read-through, per their capital-markets partners: money flows to the pinch points (engineers, skilled craft labor, and long-lead equipment), and because there are "very few scaled independent platforms today," those assets command premium takeout valuations. This is the bull case in one sentence: the constraint is the opportunity.

The bear case: 2.5x more applications than the grid can build, and a bubble nobody wants to name. On Interchange Recharged, Wood Mackenzie's Sylvia Leyva Martinez and Clean Power Hour's Tim Montague made the skeptic's case cleanly: there are "two and a half times more data-center interconnection applications than there is expected build-out of grid energy." Montague cited former Microsoft executive Ramez Naam's figures: roughly 230 GW of data-center interconnection applications over five years against a Bank of America estimate of only ~93 GW of new power plants actually getting built in that window. Martinez, who forecasts this for a living, was blunt about the gap between applications and reality: WoodMac routinely discounts its own ~90-100 GW/year project pipeline down to "around 30 to 40" that actually gets built. And Montague said the quiet part out loud: "Is there a bubble? For sure... When will it burst? I have no idea." The bear point isn't that AI power demand is fake; it's that the interconnection queue, the labor pool and the equipment lead times mean a large share of these announced projects never break ground, which caps the revenue that the whole supply chain is being priced for.

And a third voice worth separating out (the affordability backlash), because it hits regulated utilities specifically. The Open Circuit discussion above is really the bear case for the regulated names: if the political system decides $1.4 trillion is too much, the mechanism that turns capex into earnings starts to wobble. Two early signs this week: Indiana's governor signed a law paying utilities based on performance and outcomes "not just on how much they're spending," and New Jersey's governor opened with an executive order on utility business-model reform. As the Power Lines guest warned, if bills go up another 40% or even double, "I don't think any of us can imagine what that would do... not just the new politics of electricity, but also the new markets around electricity."

The names in play

GE Vernova shows up twice this week as the picks-and-shovels winner of the behind-the-meter shift, anchoring Project Kilby's gas play alongside Chevron and Microsoft (Energy News Beat), and as one of three SMR contenders on The 7investing Podcast, where the hosts framed it as the low-drama option: "GE is very good at iterating things," working with Hitachi off proven reactor designs rather than starting from a blank sheet.

Bloom Energy is the week's clearest "moves-now" thesis: real backlog, real deployments, but a valuation (80x forward, up 500%) that already assumes the boom holds (Motley Fool). Eaton got a direct name-check as the chip-agnostic way to own the buildout.

On the utilities and midstream, Telltales read the week's capital-allocation decisions like an X-ray. NextEra: shareholders approved the $66.8 billion all-stock merger with Dominion that would create the largest US regulated electric utility, and separately committed $13 billion to a Pennsylvania gas complex, but against ~$82 billion of net debt and negative free cash flow, "stock is the currency you use when your own cash flow can't fund the plan." ONEOK is buying Brazos Midstream's Permian assets for $4.4 billion cash, funded by a $9 billion Apollo minority-equity injection ("somebody else is buying a piece of the company so the company can buy the assets"), a tell on how stretched midstream balance sheets are at ~10x net-debt-to-free-cash-flow. And Cheniere, the cleanest of the four: Corpus Christi Stage 3 hit substantial completion, the 5,000th Gulf Coast cargo went out, and management raised the 2026 EBITDA guide to $7.9-8.4 billion. "They finished the thing they said they'd finish and raised the number."

On SMRs, the 7investing reality check is the one to keep. Today's demonstration reactors (Oklo, GE Vernova, NuScale) are quoting a levelized cost of $80-130 per megawatt-hour (and the hosts think it lands closer to $150+), versus roughly $51-70 for wind, solar and hydro. Their line: "we have not seen the switch flipped." The bull path is real but it's a second-generation story: costs are hoped to fall to $40-90/MWh once these move from demos to repeatable builds. NuScale drew the most interest for its Romania utility-scale project and its Tennessee Valley Authority optionality; Oklo for its sodium-cooled, fuel-recycling approach ("Elon Musk with a napkin and a pencil going after the nuclear industry") and its near-zero revenue today.

Read-throughs

How this week's threads connect across the complex:

  • On-site gensets (Cummins, Caterpillar). The bridge-power trade got a vivid case study on The S2G Podcast: ANA's eBoss pairs a battery with a diesel generator so the engine only runs to recharge, cutting fuel use up to 80%, on ~$400 million of annual sales. CFO Alicia Wainio on demand: customers "just can't wait for the grid... if you can get temporary power on site right away, then you do that." With ~160,000 mobile generators already in US rental fleets, this is a large, boring, real end-market for engine makers, and Electric Perspectives added that reciprocating engines are increasingly replacing diesel backup and selling peaking capacity back to the grid.
  • Copper, electrical steel, conductor and cable. If conductor, transformers and switchgear are each ~2x more expensive (Catalyst), that inflation flows straight up to the raw inputs: copper (Freeport), grain-oriented electrical steel, and the cable/conductor makers stringing new lines. The Pipeliners Podcast added a wrinkle: utilities are reconductoring existing towers to double capacity "miles and miles a day," often by helicopter, a fast, copper-and-conductor-hungry alternative to building new corridors.
  • Gas-weighted E&Ps and turbine OEMs. Roughly three-quarters of behind-the-meter power is gas (FMI); Project Kilby, NextEra's $13B Pennsylvania complex and Cheniere's rising LNG guide all point the same way. The Pipeliners episode's INGAA figure (+37,000 miles of gas pipeline needed through 2052) sizes the midstream tail. Gas turbine capacity remains the choke point, with gas project costs up 4-5x since the late 2010s per Black & Veatch on Electric Perspectives.
  • Enrichment and fuel cycle. The SMR designs quietly raise a fuel problem: Oklo runs on high-assay low-enriched uranium (HALEU), and Black & Veatch flagged the nuclear supply chain (long geared to maintain the existing fleet) as a binding constraint on any new-build wave. That is the structural bull case for enrichers and converters, and for physical-uranium holders, even before spot prices enter the conversation.
  • Hyperscalers are now funding the grid directly. The most important structural shift this week: the load customers are writing the checks. Electric Perspectives catalogued it: AEP Ohio's upfront deposits and take-or-pay capacity; the first operationalized ratepayer-protection tariff between NV Energy and Microsoft; Google funding ~$50 million of meter-sited batteries in Xcel's Minnesota territory and another ~100 MW of demand response with Voltus. When the buyer de-risks the utility's capex, the whole rate-base machine spins faster.