Picture two applications sitting in the same grid operator's in-tray. One is for a new estate of 400 homes on the edge of Dublin. The other is for a 60MW data hall a few kilometres away. Both need capacity from the same overloaded part of the network, and there's only room for one of them this year.
Which one do you sign?
Grid connections don't actually work like that. But the real question is close enough that, last year, the most senior civil servant in Ireland's energy department asked it out loud.
The question nobody wanted to ask
In May 2025, Oonagh Buckley, then secretary general of the department responsible for energy, told an audience that data centres had used up the grid's spare capacity. Policymakers, she said, were "having to even think about prioritising what is the social need of the demand: is it housing or is it AI?"
The Taoiseach quickly rejected that framing as "not true", and a fellow minister called it an oversimplification. They had a point: a grid isn't one queue, and homes and data halls rarely compete for exactly the same megawatt. But the numbers explain why the question got asked.
According to the CSO, data centres used 23% of all metered electricity in Ireland in 2025, up from 5% in 2015. That's 7,663GWh, a 10% rise in a single year. All of Ireland's urban homes together used 18%. EirGrid expects data centres to reach 31% of national demand by 2034. No other EU country comes close.
Ireland has turned into the world's test case for the question every grid will eventually face: when there isn't enough power for everyone, who gets it first?
What Ireland decided
For several years the answer in Dublin was, in effect, "nobody new". EirGrid stopped offering new data centre connections in the greater Dublin area, citing constraints that could last until around 2028. The industry body Digital Infrastructure Ireland puts the value of projects caught in that gap, permitted and on purchased land but without power, at €5.6–5.8bn depending on which of its estimates you read.
In December 2025 the regulator, the CRU, replaced the freeze with something more interesting: a bargain. Under its Large Energy Users Connection Policy, a new data centre of 10MVA or more can connect if it:
- Brings its own firm power. It must provide new dispatchable generation or storage, on site or nearby, matching 100% of its maximum import capacity. That plant has to trade in the wholesale market, and the data centre can't ramp up to full load until it's running.
- Pays for new renewables. At least 80% of its annual demand must be matched by new Irish renewable generation within six years of energisation.
- Passes a location test. The grid operators now assess each application against local constraints. Greater Dublin is currently treated as fully constrained.
In return, compliant sites are exempt from mandatory demand curtailment during grid emergencies.
So Ireland's official answer to "homes or data centres?" is neither. It's closer to: both, provided the data centre brings its own power and pays for its own green electrons.
The case that data centres are grid-hogs
It's worth taking the critics seriously, because their argument is about money, not ideology.
Someone pays for the wires. The CRU has approved up to €18.9bn of investment in Ireland's grid and network for 2026–2030, the largest in the State's history. It estimates the cost at roughly €12–21 a year on the average household bill. A recent academic report from Trinity College argues that data centres don't pay the full cost of the reinforcement their demand requires. The CRU disputes this, saying network investment isn't being undertaken "solely or primarily" for data centres, that each customer pays in proportion to use, and that charges for the largest users have gone up.
Housing feels the squeeze. The Construction Industry Federation has reported that housebuilders face "significant and ongoing delays" getting electricity connections. Whatever the cause, that's the kind of fact that turns a technical debate into a political one.
The US shows what happens at scale. PJM, the grid operator covering 13 states from Illinois to Virginia, runs an auction to secure future generating capacity. Its independent market monitor attributes around $6.3bn, or 38%, of the cost of the most recent auction to data centre demand. That cost ends up on everyone's bills.
The case that data centres are anchor customers
The other side has a serious argument too.
Big, steady demand pays for infrastructure. A grid is mostly fixed cost. A customer that draws a flat load 24 hours a day for 15 years is, in principle, the ideal customer to spread that cost across, if the tariff makes it pay its share. Ohio offers one model: its regulator approved a tariff requiring new large data centres to pay for at least 85% of their contracted capacity for up to 12 years, whether they use it or not. Consumer advocates signed up to it as well as industry.
Data centres fund renewables. Microsoft announced power purchase agreements for around 900MW of new Irish wind and solar back in 2022. The IEA expects renewables to meet nearly half of the growth in data centre demand globally to 2030, much of it underwritten by tech company contracts. As Equinix Ireland's Peter Lantry put it: "If we're given the ability to grid connect... we will be buying and paying for wind farms or solar farms onshore and offshore."
Their generation can help the grid. The Irish rules mean every new data centre adds dispatchable capacity to the system, not just demand. In a tight winter evening, that could be the plant that keeps the lights on.
The honest answer is that both cases are true, and which one wins depends on the rules. For contractors, that makes this a business risk rather than a debate: the rules decide which projects go ahead, where, and in what order. (How local communities feel about all this is a story in itself, and that's Issue #6.)
Three grids, three answers
Ireland: bring your own power. Connection is possible, but the price of admission is a power station.
The US: join the queue. PJM's interconnection queue holds more than 260GW of requests, and waits reportedly average around five years. Some developers are building their own gas generation rather than wait, and more states are writing large-load tariffs like Ohio's.
The Nordics: use the surplus, while it lasts. Hydropower gives the region some of Europe's cheapest, cleanest electricity. Sweden's grid operator, Svenska kraftnät, puts the Nordic surplus at around 53TWh this year. But it expects that to fall to 29TWh by 2030 as data centres and electrification take it up. Even Norway's grid operator has reportedly tightened the rules for large connections. Every surplus turns into a queue eventually.
Where the power can't come to the project, the project builds the power. Pure DC's Dublin campus is the clearest European example: a 110MW microgrid built with AVK, using Wärtsilä dual-fuel engines across three 30MW energy centres plus a 20MW battery. It runs on natural gas, can switch to HVO, and is designed to work independently of the national grid. Whether large tenants will accept gas long term is a net zero question, and one we'll come back to in Issue #5.
The labour market on the other side of the fence
Here's what this means for the people side of the job.
A data centre that has to bring its own power is really two projects: a data hall and a small power station. The second needs people the first never did: HV substation engineers, protection and control specialists, gas engine and battery commissioning engineers, and people who can synchronise generation with the grid and run it safely.
Meanwhile, the grid itself has become a labour market of its own. Ireland's €18.9bn network programme has to connect up to 300,000 new homes, a million EVs and 680,000 heat pumps by 2030, and bring in offshore wind. Every one of those projects wants the same HV jointers, cable crews and testing engineers that a data centre energy centre needs. The grid is now the data centre industry's supplier, its competitor for labour, and sometimes its employer.
What this means on site
- Power strategy now drives the programme. On a "bring your own power" project, the critical path runs through the energy centre, not the data hall. Get HV and generation commissioning people involved at bid stage, not after the hall is topped out.
- Hire for power generation, not just building services. Engine, battery and protection commissioning engineers, and HV authorised persons, come from a small pool that utilities and renewables developers are also chasing. A strong M&E team without them is a team waiting on someone else's programme.
- Treat grid-side programmes as competitors for labour. When a utility mobilises a substation or transmission programme near your site, your jointers and testers will get calls. Know who is exposed, and think about retention before the calls start.
Three numbers worth knowing
23%: the share of Ireland's metered electricity used by data centres in 2025, according to the CSO. All urban homes together used 18%.
80%: the share of annual demand that a new Irish data centre must match with new Irish renewable generation within six years of switching on.
29TWh: the Nordic power surplus forecast for 2030, down from around 53TWh in 2026. Even the hydro surplus has a limit.
So, who gets the power?
Increasingly, whoever brings their own.
The "homes or data centres" framing makes a good headline, but the real question is narrower and more practical: who pays for the wires, who builds the generation, and who has the people to do both. The projects that answer those questions first are the ones that get built.
One question back to you: is power strategy changing who you hire on your current project, or when? Reply and tell me. I read every answer.
Alex The DC Build Network