Dublin's DC pause, four years in

There's a grid connection in West Dublin, at a place called Castlebaggot, that was earmarked for housing. It ended up going to data centres.

Not because anyone decided housing mattered less. Because the hyperscalers had booked the capacity years in advance and Ireland had no rules about priority of access. First come, first connected. The houses are on a waiting list. The CRU's own decision paper lists the case as one of the motivations for tearing up the connection regime. The regulator didn't name it casually.

This is the Irish data centre pause in one anecdote. And people keep missing what's interesting about it.

Four years ago, in June 2021, the Commission for Regulation of Utilities issued CRU/21/124. Not a moratorium, the CRU was careful to say. Just a direction to EirGrid and ESB Networks to only offer connections where the electricity system could handle them, with the criteria set tight enough that inside the Dublin metro area the answer was effectively no. 97% of Irish data centres had clustered in the Dublin ring during the 2010s. The grid had filled up. EirGrid had been warning the government since 2016 and nothing had happened for five years.

Then everything happened at once.

Data centres hit 22% of national electricity consumption in 2024. That figure gets used as a weapon by everyone who picks it up, which is a shame, because it's more interesting than the argument it usually gets pressed into.

I know this is going to annoy people, and I'm partly playing devil's advocate here, but the 22% figure doesn't say what most commentators think it says. It tells you data centres consume a lot of electricity. It doesn't tell you they consume it badly, and it doesn't tell you removing them would free that electricity for housing or EVs. Much of the Irish DC load runs on hyperscaler PPAs that wouldn't exist without the data centres. Some of those PPAs underwrote wind farms that now serve other customers too. The causal arrows are messier than the headline.

Castlebaggot is different. That one deserved the criticism.

What the government did next is where the Irish case gets properly interesting. Not the 2021 restriction. The redesign.

In February 2025 the CRU published CRU/2025/04, the proposed Large Energy User Connection Policy. The final version landed in December. It is not a lifting of the moratorium. It is the terms of a new contract between data centres and the Irish grid, and the terms are demanding.

Any new connection at 1 MVA or above must meet 80% of annual demand with additional renewable generation, proven through a new corporate PPA or directly-developed assets. Six-year glide path from energisation. Dispatchable on-site or proximate generation and/or storage matching the facility's Maximum Import Capacity, participating in the Single Electricity Market. Miss the performance targets, and the SOs reduce your MIC.

If you read that and thought "this rewards sophisticated operators", you read it correctly. Passive colo hosts who don't want to think about the grid won't find Ireland welcoming. Operators who can bid into SEM, run dispatchable assets, and run a serious renewable procurement function will find it easier than they expected, because the competition for sites got smaller.

The Dutch, Germans, and UK DNOs are reading these documents carefully. The combination — locational prioritisation, dispatchable on-site capacity, renewable matching with a glide path — is a template other regulators will reach for once their own grids get similarly full. Which is already happening.

A detail worth naming that doesn't get enough airtime. During the constraint period, the Irish government and system operator commissioned over €1 billion of emergency gas generators near the Port of Dublin, in the Midlands, and at a handful of other sites. The purpose was to get through tight winters without rolling blackouts. It worked. EirGrid got through winter 2025/26 without a widespread blackout. The cost landed on Irish households, adding around €100 per family to 2024 bills.

Constraint isn't free. It just shifts who pays.

What is explicitly not in the new regime is a general demand-flexibility mandate. The CRU considered it, declined. The SOs retain the ability to impose flexibility case-by-case, which in practice means any applicant showing up with credible flexibility capability will clear faster than one who doesn't. Written rules and applied rules rarely match in regulated markets. Anyone who's worked on an Irish connection application in the last eighteen months already knows this.

The Irish story is genuinely four years ahead of the continent. It took four years to produce and will take another six to work through the pipeline. Most European markets are still at the 2019 position on their own version of this timeline.

Anyway. The documents are public. Google "CRU2025236". You'd rather read them now.