Why location is the bottleneck of the energy transition.

The Dutch energy transition has stopped being a forecasting problem and started being a logistics problem. Capacity targets are set. Hardware orders are placed. Finance is queued. The question isn't whether we can build it. It's whether we can locate it fast enough.

That sentence sounds like marketing until you sit through a grid-operator investment-plan review. The numbers are unambiguous: lead times on substations and high-voltage connections sit at five to nine years, and the largest single chunk of that calendar — frequently three of those years — is permitting and stakeholder consultation. The construction itself is a fraction of the total.

This is the part of the energy transition almost nobody talks about. Almost everybody is solving the wrong problem.

The stack of acceleration

Three things stack on top of one another at the same time, each of them pulling at the planning calendar from a different direction.

First: transformer density. Where one mid-voltage transformer once served roughly a hundred residential connections, the rule of thumb is now closer to twenty-five. Electrification of heat and transport pulls peak load up and changes its shape. The grid that worked for 2010 doesn't work for 2030.

Second: housing. The Netherlands has committed to roughly nine hundred thousand new homes by 2030. Every cluster of homes needs a transformer location, every new substation needs an HV connection, every HV connection needs land acquisition and a permit.

Third: net congestion. The grid operators are publishing congestion-warning regions at a cadence that would have been alarmist five years ago and is now routine. The structural answer to congestion is more substations in the constrained regions — exactly the regions where every parcel is already contested.

Layer those three and you get a planning calendar that doesn't fit inside a quarter, a year, or sometimes a political cycle.

Where the quarters go

Strip a five-to-nine-year substation timeline down to its components and you get something like this:

• Year 1: candidate selection, internal screening, initial site studies.
• Years 2-3: permitting, public consultation, environmental review, objection handling.
• Year 3-4: land acquisition, contractor procurement, design freeze.
• Years 4-7: construction.
• Years 7-9 in contested cases: re-permitting after objection, alternative-site studies, council escalation.

The bulk of the calendar that actually slips is in years one through three. Construction tends to land roughly on schedule once it's started. Permitting and stakeholder consultation are where uncertainty enters the system, and the uncertainty cascades.

What makes those early years so expensive is the kind of mistake that's possible. A site that looks viable in the first screening round can fail eighteen months later in public consultation — not because the site is technically wrong, but because the screening round didn't surface the political reality of the neighbourhood, the prior opposition history of the gemeente, the visual impact assessment, the proximity to a particular residential street. The cost of catching it late is enormous: the calendar restarts, and the next iteration carries the political memory of the previous one.

The counterargument and why it fails

The standard objection to "site selection is the bottleneck" runs like this: Site selection has always been part of the work. What's actually constrained is permitting capacity at the gemeenten, or land prices, or contractor availability.

The objection isn't wrong, but it's incomplete. Permitting capacity is constrained — yes — but every poorly-screened candidate eats from the same fixed pool of permitting capacity. Land prices are tough — yes — but the candidates that turn out to be expensive and infeasible are eating from the same budget as the candidates that pay back. Contractor availability is real, but it shows up after the design freeze, not before.

The constraint that compounds is the constraint at the top of the funnel. If site selection narrows from a thousand candidates to a hundred in three weeks instead of three months, the downstream permitting effort runs against a better-qualified set. Less rework. Fewer surprises in week sixty. Cleaner dossiers for ACM. The bottleneck doesn't go away; it stops being the binding constraint.

What changes the shape

What changes the shape of the calendar is not faster permitting — that's a function of the gemeenten, not the operator. What changes the shape is better screening earlier, applied to every candidate consistently, with the public-opposition signal surfaced in week three rather than month thirty.

This is precisely what AI scoring at scale enables. Three things have to come together:

1. Comprehensive enrichment: per-candidate data from Kadaster, BAG, BGT, BRO, environmental contours, demographics, traffic, prior opposition patterns. No gaps, no inconsistent depth between candidates.
2. Auditable reasoning: every score traces back to its inputs, every input is citable. A score without citation is just a guess that learned a new vocabulary.
3. Ensemble scoring: three models, scored independently, disagreement signalling locations that need a human read. No single model's confidence carries the decision.

Apply those three to a thousand candidates and the funnel narrows on the right axes. Apply them to one candidate at a time and you're back where you started.

The energy transition is going to be built. The question is whether the calendar bends from nine years to seven, or from nine years to five. The screening round is where that bend happens.

That's why location is the bottleneck — and why getting the screening round right is the highest-leverage thing a grid operator can do in 2026.