The Electric Grid

EIA's July 2026 Short-Term Energy Outlook (via Reuters, 7 July 2026) put US consumption at a record 4,195 billion kWh in 2025, rising to 4,269 billion kWh in 2026 and 4,399 billion kWh in 2027, led by data centres and electrification. Commercial demand is expected to exceed residential demand in 2026 for the first time on record. An earlier EIA outlook (February 2026) put baseline load growth at 1.9% in 2026 and 2.5% in 2027.

• LBNL's Queued Up 2026 edition (June 2026) counted over 2,060 GW of generation and storage seeking connection at the end of 2025, a 10% fall from a year earlier.
• High withdrawals and fewer new requests drove the fall; the third and fourth quarters of 2025 saw record withdrawal volumes (Interconnection.fyi, citing LBNL data).
• Median time from request to signed interconnection agreement was 45 months in 2025, and from agreement to operation 31 months, up from 21 in 2008 (LBNL, June 2026).

After DOE asked FERC in October 2025 to standardise large-load (20 MW and up) interconnection, FERC on 18 June 2026 issued show cause orders to CAISO, ISO New England, MISO, NYISO, PJM and SPP, finding their rules appeared inadequate (Utility Dive, 2026). The orders cover study processes, cost shifts, co-location, flexible loads and alternative transmission technologies such as advanced conductors. FERC extended the response deadline to mid-November 2026 (Utility Dive, August 2026). States keep authority over retail sales and siting.

Wood Mackenzie's Q2 2025 survey, as reported by POWER Magazine in 2026, put average lead times at about 128 weeks for power transformers and 144 weeks for generator step-up units, with a roughly 30% supply deficit for power transformers in 2025. Announced North American factory investment reached nearly $1.8 billion by September 2025, including Hitachi Energy's $457 million Virginia plant (targeted for 2028) and Siemens Energy's $150 million North Carolina plant (production expected early 2027). Press reports in 2026 cite GSU waits above 160 weeks, but those come from secondary sources.

Not refreshed: data centre share of US electricity (2026-2030 estimates), utility capex (about $180B a year), the $2T+ grid investment estimate, the 2026 transformer lead-time survey, the ICF 2030 demand forecast, and asset-age and miles-of-line figures; no newer primary source was read for these.

1. 1882Pearl Street: Edison lights lower Manhattan
2. 1930sRural power: the New Deal wires America
3. 1965Northeast blackout: interconnection's growing pains
4. 1990sDeregulation: markets replace monopoly in half the US
5. 2020sThe strain: new demand meets old iron

1. Generate: plants and farms make power at scale.
2. Transmit: high-voltage lines move it across states.
3. Distribute: local wires step it down to neighbourhoods.
4. Balance: operators match supply to demand every second.

1. Generation~20 kV at the plant (the power plant)
2. Transmission230-765 kV highways
3. Substationstepped to ~13 kV
4. Distributionthe local feeders
5. Your home120/240 V (your outlet)

The grid was largely built in two bursts, the 1960s-70s and the 1990s. The bill for that aging arrives at exactly the moment demand growth returns.

The chart marks two flat decades, then +25% by 2030. Values are read approximately from the axis; the 2026 point is EIA's July 2026 forecast of 4,269 billion kWh (2025 was a record 4,195 billion kWh).

For twenty years the grid's problem was too little growth. Overnight, the problem became too much. An industry staffed, financed and regulated for stagnation is being asked to sprint.

Against data centres that want power in two years and factories in three. The mismatch between demand's clock and the grid's clock is the single most important number in American energy.

Transmission, distribution, transformers and storage: the third great electrification build. For scale: that approaches the inflation-adjusted cost of the entire Interstate Highway System, twice.

1. Bills rise
2. Rich customers defect to solar
3. Fewer payers left
4. Fixed costs concentrate
5. Bills rise faster (the loop repeats: the death spiral)

Electricity used to be boring infrastructure. It is now the binding constraint on the two biggest builds of the era, AI and the energy transition. That promotion, from plumbing to strategic asset, is the grid's whole story this decade.

1. Find the pinch: a corridor that congests a few hundred hours a year.
2. Site batteries: at each end of the constrained line.
3. Shift the peak: charge off-peak, discharge through congestion.
4. Defer the build: the $1B line waits a decade, or forever.

• Big load routes around. AI campuses co-locate with plants or build their own power.
• Batteries flood in. Storage is the fastest-permitted asset on the grid.
• Reconductoring spreads. The quiet doubling of existing corridors.
• Reform grinds forward. FERC and states inch queues toward sanity.

Resilience looks expensive until you price its absence.

1. Do queues start shrinking? Reform's first measurable win.
2. Do transformer waits peak? New factories vs surging demand.
3. Does big load go behind-the-meter? The workaround becoming the norm.
4. Does a state pioneer AI tariffs? The template others will copy.
5. Does a major outage reset politics? One bad week changes everything.

The 20th century's grid was built for a nation of consumers. The 21st century's must serve a nation of computers, factories and batteries, all at once. Same wires, new civilisation. That is the engineering brief of the next two decades.

Every technology of the decade (the model, the robot, the EV, the factory) ends its supply chain at the same place: a transformer that takes three years to arrive. The grid is where the future queues.

1. Demand growth is structural: five megatrends, one set of wires.
2. Process, not physics, is the constraint: queues and permits set the pace.
3. Hardware scarcity is real: transformers are the new GPUs.
4. The money flows to the build: utilities and equipment makers feast.
5. Power is strategy now: energy access decides where the future gets built.

• LBNL: 'Queued Up' interconnection data; data-centre demand study (2024-25).
• Wood Mackenzie: transformer supply-deficit analysis (2025).
• EPRI: data-centre load forecasts to 2030.
• ICF: US electricity demand growth outlook (2025).
• DOE: National Transmission Needs Study; grid reports.
• S&P Global: data-centre power demand tracking (2025).
• Utility filings: capital plans and rate cases (2025-26).
• Figures marked 'estimate' are compilations.