Energy

27 Gigawatts Of Requests Met A 9 Gigawatt Grid

Israel froze new data centre grid connections for 140 days after requests hit 27 gigawatts against 9 gigawatts of average national demand. Denmark already wrote the fix.

EnergyBy Servola Tech Desk2026-07-255 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

27 Gigawatts Of Requests Met A 9 Gigawatt Grid

Key takeaways

Israel's Electricity Authority ordered its system operator to stop processing new data centre connection requests for at least 140 days after applications reached 27 gigawatts, three times the country's average consumption of about 9 gigawatts, which means a connection queue is now the binding constraint on compute rather than chips or capital.
Treat your own connection application as a document that has to prove maturity, not reserve a place: Denmark abolished first-come, first-served for large loads on February 1 and now judges projects in pools on development progress and grid-friendliness.
If you have a live application in any congested market, assemble the evidence file now - signed offtake, planning consent, equipment orders, a firm grid study - because that file, not your queue position, is what survives a re-ranking.
Budget for a connection date that is set by a regulator's allocation rule rather than by your build schedule: waits in Frankfurt, London, Amsterdam, Paris and Dublin already average seven to ten years against an 18 to 24 month construction window.

A regulator counted the queue and stopped the clock

Israel's Electricity Authority instructed the system operator Noga in late July to stop processing new server farm connection requests for at least 140 days, and the reason is arithmetic rather than politics. Requests on the desk total roughly 27 gigawatts. Israel's average consumption is about 9 gigawatts, and its all-time record, set during the August 2025 heatwave, is around 17 gigawatts. Some 19 gigawatts of that pile arrived in about two months, on top of an existing 8 gigawatt backlog. The Authority's own language is blunt: 27 gigawatts dramatically exceeds the entire planned generation and transmission capacity for the decades ahead.

Why it matters: without the freeze Noga would have had to issue refusals covering more than 25 gigawatts of requests, and a refusal is a harder thing to unwind than a pause. The grid has already committed about 1.5 gigawatts to server farms, which on its own would take data centres to roughly a tenth of national electricity by the start of the next decade. That is the whole story in one line: the committed pipeline is already material, and the requested pipeline is three times the country.

Most of 27 gigawatts was never a project

A queue that is three times national demand is not a demand forecast, it is an options market. Filing an application is cheap, holding a place is valuable, and so the rational move for any developer in a tightening market is to apply early and often for more capacity than they will build. That is how 19 gigawatts appears in eight weeks. Noga described the task ahead in exactly those terms, as how to correctly allocate a limited public and infrastructure resource among all economic needs, which is the language of rationing rather than of forecasting.

Yes, but: the speculative applications are not the villain, and treating them as fraud misses the mechanism. They are the predictable output of a rule that rewards being early rather than being ready. That is why the freeze on its own solves nothing. During the 140 days the Authority says it will re-examine its assumptions on supply reliability, system redundancy, competition in the electricity market, future prices and gas consumption. Every one of those is a reason to change the allocation rule, and none of them adds a megawatt.

Denmark already ran this experiment and published the answer

Energinet paused all new large-load connection agreements in March after its own queue reached about 60 gigawatts against Danish peak demand of roughly 7 gigawatts, with data centres accounting for around 15 gigawatts of it. The pause was not the remedy. The remedy was that from February 1 Denmark abolished the first-come, first-served principle for large loads and replaced it with collective assessment in pools, ranking projects on maturity, development progress and how friendly they are to the grid. The pause was then lifted on June 3 with a stricter case-processing model in place.

Read those two countries together and the pattern is clear. Neither regulator found more capacity. Both changed who gets it. Ireland, the Netherlands and Spain each arrived at the same junction and each wrote a different rule, and Texas answered with batch screening rather than project-by-project review. So a business planning capacity across several European markets is not facing one grid constraint. It is facing four or five separate rationing regimes, each with its own definition of a serious project.

What actually gets you connected now

Under a maturity test, the scarce input stops being an early application and becomes evidence. Signed offtake or a named tenant, planning consent in hand, ordered long-lead equipment such as transformers and switchgear, a completed and funded grid study, and a construction programme with dates that a regulator can check: that is the file that survives a re-ranking. A placeholder application with none of it attached moves from being a free option to being a liability, because pooled assessment is where thin applications get visibly sorted from real ones.

The bottom line: if you hold a live application in a congested market, build that evidence file before the rule changes rather than after, because the re-ranking happens on the regulator's timetable and you will not be given a window to improve your submission. And if you are choosing between sites, weigh the allocation rule as heavily as the tariff. A cheaper megawatt-hour in a queue you cannot clear is worth less than a dearer one in a market that will actually connect you inside your build window.

The second-order effect lands on everyone else's bill

The Authority is not only asking whether the grid can carry the load. It listed competition in the electricity market and future electricity prices among the things it will reassess, which is the point at which this stops being a data centre story and becomes an industrial cost story. When a single category of new load can request three times national consumption, the allocation rule that follows sets the marginal price of power for every factory, cold store and workshop on the same network.

In Israel that argument is already institutional. Avi Simhon, who chairs the National Economic Council, is pressing legislation to classify server farms as national infrastructure so their planning moves faster. That is a coherent position and it collides directly with the Authority's freeze, because faster planning consent for a load the grid cannot yet carry only moves the bottleneck one step downstream. Which of those two wins is the thing worth watching over the next 140 days, and the answer will be copied.

Frequently asked questions

Does a 140-day freeze in Israel matter to a European operator?

It matters as a template rather than as a local event. Israel is the fourth jurisdiction in roughly a year to stop taking large-load connection requests, after Denmark, and alongside restrictions in Ireland and the Netherlands and batch screening in Texas. Each one has responded by rewriting how capacity is allocated rather than by adding capacity, and the rules they write get copied between regulators. If you plan compute or industrial load anywhere in Europe, the allocation rule is now a planning input on the same level as the tariff.

Why would developers request three times the country's electricity?

Because applying is cheap and holding a place in a queue is valuable. Under first-come, first-served, the rational move is to apply early, in several places, and for more capacity than you expect to build, which is why Israel received about 19 gigawatts of requests in roughly two months. That is not necessarily bad faith; it is the predictable result of a rule that rewards speed of filing over readiness to build. It is also why regulators are moving to maturity-based pools, where a speculative application no longer works.

What should I do if I already have an application in a queue?

Assemble the evidence that proves the project is real, and do it before the rule changes. That means signed offtake or a named tenant, planning consent, purchase orders for long-lead items such as transformers and switchgear, a completed grid study, and a dated construction programme. Under pooled assessment your queue position may be reset, but a documented, funded, consented project ranks well in almost every version of the new rules. Do not assume you will get a chance to strengthen the file after the re-ranking is announced.

The constraint on compute has moved from silicon to the substation, and the countries hitting the wall are not building their way out of it but writing rules about who gets served first; the operators who treat those rules as a design input will be connected while their competitors are still holding queue tickets.

grid connectiondata centresIsrael Electricity AuthorityNogaEnerginetinterconnection queueindustrial electricity

FERC Just Made Data Centres a Grid Entity

On 16 July 2026 FERC voted 5-0 to direct NERC to write mandatory Reliability Standards for data centres and to revise its registry criteria. Both are due 31 December 2026. What it means for owners with US compute.

4 min read

Energy

Google's Power Use Jumped 37% in a Single Year

Google's electricity use rose 37 percent in one year, its largest jump ever. The part it cannot offset - supply-chain carbon and grid access - is the cost every AI buyer now carries.

3 min read

Law & Regulation

The EA Buyout Now Hinges on Brussels

The largest take-private in history has cleared EA shareholders and US antitrust. The last real gate is the EU - and not classic merger review, but the Foreign Subsidies Regulation, facing its first blockbuster test.