3,000 GW in the US Grid Queue. 128 Weeks for a Transformer. France Delivers in 18 Months.

Three data points, published independently this week, tell the same story from three different angles. Together, they constitute the clearest quantitative case for European AI data center deployment that has appeared in a single news cycle.

THE QUEUE THAT EXPLAINS EVERYTHING

The US grid interconnection queue now holds more than 3,000 GW of active requests against approximately 1,280 GW of installed generation capacity. That ratio — more than 2.3x — has been building since 2005. Lawrence Berkeley National Laboratory data shows that median interconnection study duration has risen from under 20 months in 2005 to over 50 months in 2023. The process takes longer, costs more, and succeeds less often than it did a decade ago.

The cost data is particularly revealing. Total interconnection costs for projects completing in the 2019–2023 period reached $200–400/kW — compared to under $50/kW pre-2014. For a 100 MW data center, that translates to $20–40 million in interconnection costs alone, before construction begins. And for projects that withdraw — which represent the majority of applications — those costs are largely unrecoverable.

THE TRANSFORMER CRISIS IN NUMBERS

The grid queue is a process problem. The transformer crisis is a manufacturing problem. They compound each other in a way that most developers only discover after committing capital.

Current benchmark data from industry procurement sources shows power transformers at 128 weeks average lead time — up from roughly 52 weeks in 2019, representing a 146% increase in delivery time alongside a 77% price increase. Generator step-up transformers, the specific type required for large data center grid connections, are at 144 weeks — close to three years.

The White House Presidential Determination of April 2026, invoking Section 303 of the Defense Production Act to declare transformers essential to national defense, was the formal federal acknowledgment of what the procurement data had been showing for two years: domestic manufacturing cannot meet demand, and the shortfall is a national security issue.

WHY THESE TWO PROBLEMS COMPOUND

The US grid queue and the transformer shortage are not independent constraints. They interact in a way that makes the total timeline worse than either alone would suggest.

The standard US interconnection process requires a developer to complete a Facilities Study — the final study before the interconnection agreement — before transformer specifications can be confirmed. That study typically takes 12–24 months after the System Impact Study, which itself takes 12–24 months after the Feasibility Study. By the time a developer knows exactly what transformer they need, they are already 4–6 years into the process. At that point, they place a transformer order with a 144-week lead time.

Total timeline from application to first power: 8–12 years for a greenfield US site. This is not an edge case. It is the median outcome for projects that complete the process at all.

THE FRANCE COMPARISON — WHAT 18 MONTHS ACTUALLY MEANS

France's RTE operates a deterministic connection process with published timelines. A standard HTB connection runs 12–24 months from application to energisation. But the more important comparison is the brownfield case — sites with existing HV infrastructure from former industrial use.

Nebius Group acquired the former Bridgestone tire factory in Béthune, Hauts-de-France in early 2026. The site retained its HTB substation connection from previous industrial use. First power is targeted for end 2026: 18 months from signing to 240 MW of operational AI compute capacity.

The EU transformer advantage is structural, not incidental. European second-tier manufacturers — Efacec in Portugal, Pauwels in Belgium, TMC in Italy, Schneider in France — deliver at 20–32 months. More importantly, in France, transformer procurement can begin at month one of the project — before RTE approval, alongside site acquisition. The specifications are known early enough to place the order before the queue resolves.

This is the operational insight that changes the economics: in France, the transformer lead time runs in parallel with the grid connection process. In the US, it runs sequentially after it.

WHAT DEVELOPERS SHOULD DO NOW

The data published this week makes two things clear for developers evaluating European deployment.

First, the US constraint is structural and worsening. The queue has been growing faster than the grid for 15 consecutive years. The transformer shortage has not peaked. The Presidential Determination accelerates domestic production, but domestic manufacturing capacity cannot close a 128-week lead time gap in less than 5–7 years. Developers who are waiting for the US market to normalise are waiting for a horizon that recedes as they approach it.

Second, the France window is finite. RTE's 5 fast-track sites represent 4,800 MW of near-ready capacity. The brownfield HTB stock that makes 18-month timelines possible — former industrial sites with existing substation infrastructure — is a depleting asset. Nebius saw this and moved in 2025. The sites available today will not be available in 2028.

The numbers this week are not new in kind. They are new in clarity. 3,000 GW in queue. 128 weeks for a transformer. 18 months in France. The gap between those three figures is where AI infrastructure strategy is being won and lost right now.

→ Related: Grid data center hub · France site selection · $55M to save 4 months — FERC filing · Goldman Sachs $700B grid analysis

Sources: Lawrence Berkeley National Laboratory (LBNL) Interconnection Queue Study 2024 · FERC interconnection data · Industry transformer procurement benchmarks June 2026 · White House Presidential Determination April 2026 · RTE / Choose France 2026 · GridReadiness field intelligence. Updated June 2026.