You can hold a rezoning, a conditional use permit, a signed water letter, and 900 contiguous acres, and still not have a data center site. What you have is permission to want power.
A data center interconnection is the utility and grid-operator process that turns requested megawatts into delivered megawatts: service request, load study, network upgrade identification, cost allocation, construction, and energization. Entitlement is a local land-use decision. Interconnection is an electrical and regulatory one. They run on separate clocks, and only one of them decides whether the campus ever turns on.
This is for the people underwriting a site where the power is the assumption rather than the analysis — developers, land funds, econ dev directors, county planners, and the utility engineers who keep getting handed a CUP as though it were capacity.
the two clocks
The powered-land timeline runs roughly like this, and the industry consensus for getting to shovel-ready is 3 to 7 years:
acquire → assemble → utility engagement → interconnection study → entitlements → infrastructure (substation, fiber, grading) → shovel-ready or sale
Note where entitlements sit. Not at the end. In the middle, and often in parallel with a study whose outcome you do not control. A project can win its hearing and then discover that the network upgrade it triggers costs more than the land.
| entitled | energized | |
|---|---|---|
| Who decides | County, city, planning commission, board of supervisors | Utility, grid operator, and in 2026 increasingly a state regulator |
| What it grants | Permission to build a use on a parcel | Firm delivered capacity at a point of interconnection |
| Typical clock | Months to about two years | Years — 5 to 10 in constrained US markets |
| What kills it | Organized opposition, moratorium, comprehensive plan conflict | Queue position, upgrade cost, generation adequacy, a rule change mid-study |
| Moves with the land | Usually, yes | Conditionally, and increasingly with milestones attached |
| What you hold without the other | A zombie | A site somebody else entitles |
three variables, not one
CBRE Investment Management frames powered-land underwriting around three interrelated variables. They are worth stating plainly because most speculative memos are strong on the second, quiet on the first, and silent on the third.
- Power deliverability — can you actually energize on schedule, at the requested load, at a cost the deal survives?
- Development execution — can you entitle, permit, and build the substation, fiber, grading, and stormwater?
- Commercial durability — will the demand you are underwriting still exist at the end of a 3-to-7-year hold?
Notice the asymmetry. You control execution — and entitlement is the part of execution you control most directly. You influence deliverability. You do not control commercial durability at all.
You might be thinking: we have a letter from the utility. Read what it actually says. A study agreement, a queue position, and an executed interconnection agreement with allocated upgrade costs and an energization date are three different documents. Only the third is an asset. The first two are options with carry.
what a zombie actually is
A zombie data center project is a site that is entitled but has no credible power path. Zoning is clean. Fiber is close. The map looks like inventory. The megawatts are hypothetical.
Zombies are expensive in three directions at once:
- For the holder — carrying costs, property taxes, and posted interconnection security that is forfeitable on a missed milestone. Deposits and study fees on large campuses run into serious money. We are not going to publish a per-campus number we cannot source, but assume the security is large enough to be a financing question, not a line item.
- For the grid — a speculative request occupies study capacity and distorts the load forecast every other project is planned against. This is the part regulators finally moved on.
- For the community — the hearing was already spent. Political capital burned, farmland taken out of a lease, a neighborhood's tolerance for the next project reduced, and no jobs, no tax base, no campus. This is the direction almost nobody underwrites, and it is the one we work on.
generation queues are not load queues
Two different queues get discussed as one, and it costs people money.
Berkeley Lab's Queued Up series — the source behind most "the queue is X gigawatts" statistics — tracks generation and storage interconnection: power plants and batteries waiting to connect. A data center is not generation. It is load.
In the US, large load historically had no comparable standardized, public, FERC-jurisdictional queue at all. You requested service from your utility or electric distribution company, and the process was bilateral, local, and largely invisible from outside.
That is changing, which is the whole story of 2026. ERCOT's Batch Zero process and PJM's proposed Large Load Registry are both attempts to build load-side visibility that never existed before. Neither of them is the LBNL figure, and neither is a substitute for it — they are new, separate, load-side instruments. Great Britain is a further case again: it has a national, published demand connections queue, which is why GB load figures are quotable in a way US load figures usually are not.
The practical version: if your investment memo cites a national queue statistic to characterize your load risk, it is citing the wrong queue. Load risk is local — your utility's process, your ISO's emerging large-load rules, and increasingly your state's regulator. That is exactly what 2026 changed.
2026: the year the queue started culling
Four illustrations, all of them structural rather than newsy. Each replaces a place in line with a burden of proof.
Great Britain — Ofgem, 29 July 2026. Contracted demand in the connections queue rose from 41 GW to 125 GW between November 2024 and June 2025 — around 73 GW of it data centres, across roughly 315 projects. Ofgem's "Curate" consultation proposed a refundable Data Centre Commitment Fee on projects of 40 MW and above — roughly £237,500 to £712,500 per MW, or about 2.5% to 7.5% of average project capex — secured from the point a connection offer is accepted until energization, and forfeited if the project exits the queue. Alongside it, evidence milestones for projects with 10 MW or more of IT load: credible compute offtake, genuine long-lead equipment procurement, and financial or technical capability. Responses close 16 September 2026, with a decision expected later in the year. Treat the fee as proposed, not final — but treat the direction as settled.
PJM — the Large Load Registry. PJM proposed a mandatory registry covering every large load of 50 MW or more at a single site, with delivery points inside a one-mile radius treated as one site. It is paired with an Interim Resource Adequacy Service under which new large loads entering service after 1 June 2027 without qualifying new capacity would be curtailed ahead of pre-emergency load management. Filed with FERC on 13 August 2026 in docket ER26-3515.
ERCOT — the Batch Zero pause, 3 August 2026. Following a directive from Governor Greg Abbott, ERCOT paused Batch Zero classification and study activity and said it would not authorize large computational loads — data centers and cryptocurrency mining facilities — at 75 MW and above to energize until an eligibility verification is complete. The PUCT granted good-cause exceptions to the missed 7 August classification deadline. The audit covers roughly 250 to 300 projects at 75 MW and above, with a report due 10 December 2026. ERCOT's general counsel said in public that the Batch Zero study will miss its 9 April 2027 deadline, and no replacement date has been set.
FERC — RM26-4 and the show-cause orders. The Department of Energy directed FERC in October 2025 to consider a rule on interconnecting large loads, generally defined as demand above 20 MW. On 18 June 2026, rather than issue one national rule, FERC opened six Federal Power Act section 206 show-cause proceedings requiring PJM, MISO, SPP, CAISO, ISO-NE, and NYISO to justify or reform their large-load and co-located-load terms. RM26-4 remains open.
What these have in common is not severity. It is evidence. Post money. Show an offtaker. Show procurement. Show credit. Show that you are actually building. The reform is a milestone cull, and its stated target is the speculative application.
Which means the honest read is not "regulators are the risk." If a site's power case is a queue number and a good relationship, 2026 repriced it. If the power case is a completed study, an allocated upgrade cost, a commitment fee you can actually post, and a phased load you can defend, 2026 helped — the culls clear the projects that were standing in front of you.
And on the obvious workaround: behind-the-meter gas, solar, or storage is a real speed-to-power path, and it is also an air permit, a fuel contract, and a hearing of its own. It relocates the constraint. It does not delete it. FERC's show-cause orders are actively rewriting co-location, netting, and ancillary service terms, so the hybrid economics underwritten in 2024 are not necessarily the ones a site will operate under.
who pays for the upgrade
The Ratepayer Protection Pledge (March 2026) put the industry on record that developers and hyperscalers, not households, should fund the network upgrades their load causes — and PJM cited it directly in its own filing. Cost allocation is now an underwriting variable and a permitting variable at the same time: who pays for the upgrade largely decides how the hearing goes. We wrote the household side of that separately, in data centers and your electric bill.
what ensurance can and cannot do here
Start with the cannot, because it matters. Ensurance cannot energize a site. We hold no queue position, we build no substations, and nothing in this protocol moves anyone up a list. Interconnection reform, deposits, and cost causation are utility, ISO, FERC, and state-regulator problems. Anyone selling you a shortcut through them is selling you something else.
What can be funded is the other half of the wait.
The watershed, the working farmland, and the living cover on and beside a power corridor exist whether or not the interconnection agreement ever executes. Those systems are the actual asset — the aquifer recharge, the stream buffer, the pasture, the tree line that absorbs the noise. They are also, in practice, what the hearing is about. Enforceable, measurable, funded commitments on water, noise, and farmland improve the odds of approval. They do not guarantee it — funded commitments have been denied — and the honest claim is a shift in probability, not a rule. Promises alone do measurably worse. We worked the hearing side of this in the data center opposition series, and covered the three commitments themselves in depth: water, noise, and farmland.
During a 3-to-7-year hold, entitlement is the part of execution a developer most directly controls. Funding the living systems on and around the parcel is a present-tense action available on day one, while the study is still pending. In this protocol that takes the form of a certificate — a purchase tied to one named natural asset, where proceeds route to the steward of that asset rather than into a general fund. It is not an insurance policy, and it is not a credit retired against damage somewhere else.
Live doors, at their real scale: rural-open-space.ensurance and inland-wetlands.ensurance are two of the accounts through which those systems are funded today. Volumes are small, and we would rather say so than imply a market that is not there yet. On our own stage: we find, broker, option, entitle, and subdivide infrastructure land and route the natural remainder. We do not bid pads, build halls, move queues, or run a powered-land fund.
frequently asked questions
what is a data center interconnection?
A data center interconnection is the process of connecting a large electrical load to the grid: the service request to the utility, the load and system impact studies, identification of required network upgrades, allocation of who pays for them, construction, and finally energization. It is separate from land entitlement and separate from generation interconnection. In the US it is governed by the serving utility and, increasingly, by ISO large-load rules and state regulators; in Great Britain it runs through a national, published demand connections queue.
why do data center projects stall in the queue?
Four common reasons. The requested load triggers network upgrades whose cost or construction schedule breaks the deal. There is not enough firm capacity in the local service territory to serve the load 24/7. The application was speculative and cannot satisfy the evidence milestones — offtake, procurement, financial capability — that regulators began adding in 2026. Or the rules changed mid-process, as with ERCOT's Batch Zero pause, and the schedule simply stopped. Note that none of these are land problems, which is why entitlement progress is a poor proxy for power progress.
what is a zombie data center project?
A zombie data center project is one that is entitled but has no credible power path — zoning approved, site controlled, megawatts hypothetical. It carries cost for the holder, occupies study capacity and distorts load forecasts for the grid, and spends a community's political tolerance on a campus that never energizes. The 2026 wave of first-ready rules, commitment fees, registries, and milestone culls exists specifically to find them.
next steps
Don't buy a CUP and call it power. Screen the interconnection first, underwrite all three variables, and treat entitlement progress as evidence about entitlement only.
Then, while you wait on watts: fund the place the hearing is about.
see how this works for data center developers →
who pays when a campus lands: data centers and your electric bill →
the screen that comes before this one: a map line is not powered land →
the series
- what powered land actually is — dirt with a solved electricity equation
- a map line is not powered land — proximity is not deliverability
- the acre next to the substation — the million-dollar bid is a conversion clock
- who buys powered land — hyperscalers want shovel-ready
- entitled is not energized — you are here
- keep the living acres — the pad side and the living side