America’s grid is bulking up for the AI age—fast. In Houston and the Midwest, utilities are forecasting near-50% load growth within a decade. In Iowa, Google is bringing a shuttered nuclear plant back to life. And in Washington, the Department of Energy just made a rare play to expand FERC’s authority over who plugs into the grid. It’s the same story told three ways: demand is no longer hypothetical, capital is moving faster than policy, and the federal reach is growing with every gigawatt.

Peak Demand Is No Longer Theoretical

Across regions, utilities are no longer talking about future demand—they’re engineering for it. From Houston to the Ohio Valley, grid planners are watching data centers turn once-modest load forecasts into full-blown infrastructure booms:

• FirstEnergy (PJM core): expects system peak to jump +15 GW (+45%) from 33.5 → 48.5 GW by 2035, driven by contracted/potential data-center load (3.8 GW contracted, 11.7 GW potential). Transmission capex steps up ~30% to support growth; ~$1B earmarked for large-load interconnections. The utility will also seek ~1.2 GW of new gas in West Virginia for 2031 service.

• CenterPoint (Houston): forecasts +10 GW of peak demand by 2031—~+50% in six years—after connecting 500+ MW of data centers this year and seeing double-digit industrial throughput gains. A $65 B capex plan (with another ~$10 B in view) follows.

Read: the message from both ends of the map is the same. PJM projects ~+48 GW of peak by 2035; ERCOT’s urban core is sprinting. Utilities are shifting capital to wires and firm capacity because the AI/industrial build-out isn’t a forecast anymore—it’s a queue number and a construction schedule.

Google & NextEra to restart Iowa’s Duane Arnold

Google and NextEra will revive the 615-MW Duane Arnold nuclear station, targeting early 2029 operations (regulatory approvals pending). Google will take 24/7 carbon-free output to anchor AI growth in Iowa; the local coop will take surplus. It follows other tech-nuclear deals (e.g., Constellation–Microsoft, Oracle’s SMR-powered design) and underscores the new calculus: if the grid can’t deliver firm, zero-carbon fast enough, Big Tech will help build it.

Why it matters: re-starting proven nuclear capacity sidesteps greenfield risk, stabilizes local reliability, and gives hyperscalers a bankable path to decarbonized baseload at scale.

DOE asks FERC to claim jurisdiction over >20-MW load interconnections

In a rare move, DOE proposed a FERC rule to assert jurisdiction over large-load interconnections (≥20 MW) that tie directly to transmission, aiming to standardize and speed studies with readiness deposits, withdrawal penalties, and expedited treatment for curtailable/dispatchable hybrids. Former FERC Chair Mark Christie flagged core questions: Will mandated interconnections outpace available generation—and who pays? Expect a hard federal–state fight over jurisdiction before any April 2026 finish line.

Upgrade to Grid Brief Premium to get extra deep dives into energy issues all over the world.

Conversation Starters

• WSJ — “Electricity Prices Going Up? Green Policy Is to Blame.”
  A brisk counterpoint arguing that plant retirements + gas constraints in policy-heavy states are the real bill drivers—not the end of federal renewables subsidies. Useful to test your priors on causality and geography.

• C3 News Mag — “Why Are Electricity Rates Rising?” (Michael Giberson)
  A data-driven explainer: delivery costs are soaring while generation costs trend down; states with rising consumption spread fixed grid costs better (rates rise slower). Great framing for separating AI headlines from the wires math.

• Financial Times — “U.S. says $80B of nuclear reactors will be built.”
  Signals a policy-and-capital alignment for AP1000s and BWRX-300 SMRs (with Japanese supply-chain backing). If even half materializes, it’s the supply-side swing the AI era requires.

Today’s Chart

Peak demand isn’t creeping up—it’s sprinting. PJM’s load is projected to climb 45% by 2035, while Houston’s ERCOT region will jump nearly 50% by 2031. Add in the DOE’s push for federal jurisdiction and Google’s 2029 nuclear restart, and you’ve got a grid rewriting itself in real time—faster than policy, and faster than anyone planned for.

Good Bet: Betting on restartable baseload.

From Google’s revived Duane Arnold plant to the Army’s Project Janus reactors, the winning strategy is bringing proven nuclear back online before new tech is ready. It’s fast, firm, and carbon-free — the rare trifecta that can actually power AI’s exponential appetite.

Bad Bet: Assuming the load curve will level out.

Data center forecasts aren’t hype anymore. Utilities from Houston to Pennsylvania are spending tens of billions because the numbers are real — 50% peak growth in under a decade. Betting on “demand moderation” is like betting on dial-up to make a comeback.

Share Grid Brief

We rely on word of mouth to grow. If you're enjoying this, don't forget to forward Grid Brief to your friends and ask them to subscribe!