Update: US Power +1.8%/+3% With DCs Still #1 — Queues 2,061 GW (−10%) but Gas +86%; US Prices Flat vs EU/JP +30%

What changed since the Q3 Gartner vintage

In our Q3 update we treated Gartner’s 10 June 2026 newsroom forecast as the near-term headline and IEA Electricity 2026 as the grid-pace restatement: worldwide data-centre electricity at 565 TWh in 2026 (+26.4%), AI-optimised servers at 31% of power and on track to surpass conventional in 2027, a ~250 TWh dual-ledger gap at 2030 (IEA ~950 vs Gartner >1,200), US data centres at ~50% of US demand growth through 2030, and >2,500 GW of projects stalled in connection queues worldwide. That post answered the near-term TWh promotion question.

This August (202608) note answers the next official prints: what moved when IEA Electricity Mid-Year Update 2026 locked a live 2026–27 demand and wholesale-price path with H1 evidence, and when LBNL Queued Up 2026 Edition restated the full US interconnection stock through end-2025?

Three information events rewrite the dashboard. Mid-Year puts global electricity demand growth at 3.6% in 2026 and 3.8% in 2027 (from 3% in 2025), with expanding data-centre capacity named among the structural drivers. For the United States — the market that decides most AI interconnection politics — Mid-Year locks +2.6% in 2025, a soft H1 2026 (~+1% total; services +3% on rising data-centre consumption; residential −1.7%), then full-year +1.8% in 2026 and +3% in 2027, and repeats that data centres remain the main driver of US demand growth. The same report’s price chapter is the non-obvious geography cut: Q2 2026 wholesale prices rose more than 30% y/y in the EU and Japan after the Strait of Hormuz LNG shock, while the United States was largely unchanged — and H2 futures even point ~10% lower. LBNL Queued Up 2026 restates US active queues at 2,061 GW (generation 1,312 GW + storage ~749 GW) — down 10% y/y as withdrawals outran new requests — while natural gas in the queue jumps +86% to 253 GW, and 549 GW already holds a draft or executed interconnection agreement but is not yet online. Median IR→COD for projects built in 2025 remains over five years; only 13% of 2000–2020 requests had reached COD by end-2025.

The dashboard above is built as a vintage delta — Mid-Year demand-path bars, Hormuz price-shock dumbbells, US H1 sector cuts, LBNL tech stacks with YoY, queue Δ meters, negative-price flexibility hours, a carried dual-ledger path, and a pace-mismatch scatter that adds a price cluster. Pair the global path with the concentration cut for where those watts land, and with US data-center power vs grid capacity for the LBNL demand-versus-miles frame.

The headline table: Q3 → Mid-Year + LBNL

The non-obvious cut is not “Gartner revised 565 again.” The June TWh path and the dual-ledger gap are unchanged. The August news is that the system around those campuses printed new meters: a live US demand slope, a Hormuz price asymmetry that favors US AI power costs, and a full LBNL queue restatement where the stock falls 10% even as gas — the bridge fuel for delayed wires — is the only major tech surging.

Mid-Year demand: global accelerates, US dips then rebounds

Key Questions and Electricity 2026 made 2030 scenario fans and five-year US adds the lead frame. Mid-Year pulls the camera to 2026–27 with H1 evidence.

Global electricity demand is forecast to grow 3.6% in 2026 and 3.8% in 2027, up from 3% in 2025, taking world consumption from about 28,600 TWh in 2025 to 30,700 TWh in 2027. Expanding data-centre capacity sits in the structural-driver list beside industry, appliances, EVs, AC, and heat pumps. That does not restate Gartner’s 565 TWh campus bill — it restates how loud the system load is while AI campuses bid into the same grids.

The United States is the sharper AI meter. After +2.6% in 2025, a milder winter cut H1 2026 growth to about +1%: residential electricity fell 1.7% as heating degree days dropped, while services rose over 3% on rising data-centre consumption and industry rose about 1%. Mid-Year’s full-year path is +1.8% in 2026 and +3% in 2027, and it repeats Electricity 2026’s qualitative hierarchy: data centres continue to be the main driver of US electricity demand growth. The Q3 ~50% of US demand growth through 2030 attribution is carried, not replaced — Mid-Year adds the near-term slope and H1 sector split that utilities actually budget against.

China’s services sector (internet/digital services including data centres, plus EV charging) is up 8.0% y/y in H1 2026 on the same Mid-Year chapter — useful context that the AI power race is not only a US interconnection story, even when US clusters dominate theme politics.

Price geography: Hormuz hits EU/Japan; US AI power stays cheap

Q3 did not feature wholesale-price asymmetry. Mid-Year’s LNG/Hormuz chapter should change how planners read where AI campuses want electrons.

In Q2 2026, average spot wholesale electricity prices in the European Union and Japan rose more than 30% year-over-year as Strait of Hormuz disruptions removed nearly 20% of global LNG supply and spiked gas prices. The United States was much less affected, with Q2 average wholesale prices largely unchanged from a year earlier. Australia printed about 45% lower on strong renewables and battery shifting; India rose less than 10%. H2 2026 futures as of mid-July point to roughly +25% in the EU and nearly +40% in Japan, versus about −10% in the United States.

That is an AI power-cost geography story, not a tourist footnote. Gartner’s US slice — 204 TWh, 36% of world data-centre electricity, 68 TWh dedicated AI — already concentrated load where interconnection binds. Mid-Year adds that those same US nodes are comparatively insulated from the LNG price shock that is raising EU and Japanese power costs. Onsite gas and turbine-order scarcity (Key Questions’ 15–27 GW bridge, carried) still matter for connection timing; Mid-Year’s fossil-spend aside — US spending on fossil-fired generation set to outpace China in 2026 for the first time in decades as the data-centre boom drives gas-turbine orders — rhymes with LBNL’s gas-queue surge below.

LBNL queues: stock −10%, gas +86%, 549 GW with IA not online

Q3 widened the aperture from LBNL’s 1,312 GW US generation queue to IEA’s >2,500 GW stalled worldwide. August features the full Queued Up 2026 Edition stock.

As of end-2025, about 8,200 projects were actively seeking US grid interconnection, representing 1,312 GW of generation and roughly 749 GW of storage — 2,061 GW total. High withdrawals (over 750 GW) alongside fewer new requests produced a 10% decrease in active queue volume versus 2024. That is reform-and-market hygiene, not “the bottleneck is gone.” Median duration from interconnection request to commercial operation for projects built in 2025 remains over five years. Historically, only 13% of capacity that requested interconnection from 2000–2020 had reached COD by end-2025; 75% had been withdrawn.

Composition is the AI-relevant cut. Natural gas is the only major technology rising: 253 GW active (+86% y/y). Solar (773 GW, −19%), storage (749 GW, −16%), and wind (220 GW, −19%) all fell. Roughly 549 GW already has a draft or executed interconnection agreement but has not reached commercial operations — including 45 GW of gas among the IA-backed stock. Withdrawal rates remain high even after signing an IA. Read that carefully: the queue is shrinking at the top line while the bridge fuel for delayed wires is the slice growing fastest — and a large IA-backed stock still has not energised.

The worldwide >2,500 GW stalled stock and 1,200–1,600 GW unlock range from Electricity 2026 are carried; Mid-Year did not restate them. Use LBNL for the US generator/storage book and IEA for the global congestion narrative — do not collapse them into one number.

Flexibility: negative prices and the battery case

Key Questions’ 20–25 GW of batteries inside data centres and AI load swings exceeding 50% of rated capacity within a second remain carried. Mid-Year adds the market half of the flexibility case.

Negative wholesale prices occurred during about 20% of hours in South Australia and California in H1 2026 (similar to 2025). Spain’s share rose to 17% from 10%. Sweden and Finland fell from about 6% to 2% as flexibility measures improved. During European heatwaves in June, midday-to-evening price spreads reached USD 600/MWh in several markets. Those prints do not move the dual-ledger TWh path, but they raise the option value of campus batteries, demand response, and flexible non-firm connections — the same unlock toolkit Electricity 2026 quantified at 750–900 GW (flexible connections) and 450–700 GW (grid-enhancing technologies).

Dual ledger carried: 950 vs >1,200 is still the 2030 fork

Mid-Year is silent on a restated data-centre TWh path. That is itself a vintage fact. We carry Gartner’s 447 → 565 → 702 → >1,200 TWh electricity path and 104 → 132 → ~290 GW capacity path, and Key Questions’ IEA central ~950 TWh by 2030. The ~250 TWh dual-ledger gap remains the honest 2030 fork. Do not average 447 with IEA’s 485, and do not invent a Mid-Year “consensus” TWh for data centres that the report did not print.

Capacity (GW) remains the interconnection constraint; electricity (TWh) remains the energy bill; Mid-Year’s price and LBNL’s queue composition are the new clocks on the same race.

What would rewrite the next vintage

1. An IEA or Gartner restatement that moves the 2030 dual-ledger endpoints by more than ~50 TWh towardor away from — each other.
2. US full-year 2026 demand printing well above or below Mid-Year’s +1.8%, especially if services/DC growth diverges from the H1 +3% signal.
3. LBNL mid-year queue file showing gas still surgingor reversing — after the +86% end-2025 print, or IA→COD conversion accelerating past the 549 GW backlog.
4. Energisation prints proving flexible connections and GETs cleared hundreds of GW from the >2,500 GW global stalled stock.
5. EU/Japan power prices staying elevated long enough to redirect AI capacity announcements toward US (or other insulated) gridsor US prices finally catching the LNG shock.
6. Onsite gas connecting at the high end of 15–27 GW and cutting measured wait timesor failing in the turbine queue Mid-Year says is already bidding US fossil spend past China.

Until those print, the live frame versus the Q3 update is: Mid-Year locks US +1.8%/+3% with DCs still #1 and H1 services +3%; US wholesale flat vs EU/Japan +30%+; LBNL restates 2,061 GW active (−10%) while gas jumps +86% to 253 GW and 549 GW sits IA-backed but not online; dual-ledger ~950 vs >1,200 and Gartner +26% to 565 TWh remain carried.

Caveats and methodology

1. This is a vintage delta, not a second encyclopedia. For the full IEA scenario fan, fuel mix, and regional concentration, use the research post, the Key Questions update, the Q3 Gartner update, and the concentration companion.
2. IEA TWh ≠ Gartner TWh/GW ≠ LBNL queue GW. Different scopes; never average into a consensus forecast.
3. LBNL Queued Up covers generation and storage interconnection, not large-load (data-centre demand) queues. Use it as the supply-side congestion book that AI load still depends on.
4. Gas queue prior for Δ charts uses an implied ~136 GW baseline from the disclosed +86% to 253 GWtreat the prior as estimated where LBNL’s slide deck emphasises the percentage move.
5. Mid-Year did not restate Gartner’s DC TWh path, the dual-ledger gap, worldwide >2,500 GW stalled, or unlock rangesthose are carried from Q3 / Electricity 2026 / Key Questions.
6. Physical generation mix ≠ contractual PPAs. Carry that research caveat forward.
7. Negative-price hours measure flexibility scarcity in wholesale markets; they are not a direct campus PUE meter.
8. US 68 TWh dedicated AI remains a Gartner composition slice from Q3, not a Mid-Year reprint.

Primary sources: IEA Electricity Mid-Year Update 2026; LBNL Queued Up: 2026 Edition (end-2025 queues); Gartner newsroom 10 June 2026 and IEA Key Questions on Energy and AI (April 2026) as carried meters; prior baseline in AI power-grid update 2026q3.

The shareable takeaway

Versus our Q3 Gartner/Electricity print, the newest vintage does not move the 565 TWh or dual-ledger 2030 fork — it rewrites the system around them. Mid-Year locks US electricity at +1.8% in 2026 and +3% in 2027, with data centres still the main growth driver and H1 services at +3% even as a mild winter flattened the total. Wholesale prices stayed flat in the US while the EU and Japan printed +30%+ after the Hormuz LNG shock — a power-cost geography that reinforces US cluster concentration. LBNL restates active queues at 2,061 GW (−10%) while gas jumps +86% to 253 GW and 549 GW sits with an IA but not yet at COD. Electricity can still be built for AI. The open question is whether wires, gas bridges, and flexibility tools clear where the GPUs plug in — on a clock that matches +26% campus years and >5-year interconnection medians, not only 2030 scenario slides.

Related reading: Q3 Gartner + Electricity update, Key Questions vintage update, IEA global research frame, power-grid concentration, and US data-center power vs grid capacity.