Ireland is quietly abandoning decades of nuclear opposition as artificial intelligence data centers consume nearly a quarter of the country's electricity supply. The Irish government is reportedly considering overturning its 1999 prohibition on domestic nuclear generation, a dramatic policy shift driven entirely by the explosive growth of AI infrastructure from companies like Meta, Microsoft, Google, and Amazon. Data center electricity consumption reached 7,012 gigawatt-hours in 2025, up from just 5% of metered consumption a decade earlier.

Why Is Ireland Suddenly Reconsidering Nuclear Power?

For nearly three decades, Ireland maintained one of the world's strictest nuclear bans, a legacy of public opposition following the Three Mile Island and Chernobyl disasters. But the economics of powering AI have changed the calculation entirely. The Central Statistics Office reports that data centers accounted for 23% of Ireland's metered electricity consumption in 2025, up from just 5% in 2015. This explosive growth has transformed electricity from a commodity utility into a strategic bottleneck for the nation's digital economy.

Taoiseach Micheál Martin has stated the government is "open" to nuclear power, while Fianna Fáil TD James O'Connor has introduced legislation to lift the nuclear veto. The shift reflects a broader recognition that renewable energy alone cannot reliably power the continuous, high-intensity operations that AI data centers require.

What Do Energy Experts Say About Nuclear's Timeline?

The debate has exposed a fundamental tension between political urgency and engineering reality. Dave Kirwan, Managing Director at Centrica Power, argues that smaller modular reactors could be deployed alongside major electricity users like data centers, potentially accelerating deployment compared to conventional nuclear plants. "I worry that we think we have the luxury of 10 to 12 years," Kirwan stated. "I don't think we do".

Dave Kirwan, Managing Director at Centrica Power

However, Hannah Daly, Professor of Sustainable Energy at University College Cork, counters that nuclear cannot address Ireland's immediate electricity challenges. "Nuclear is certainly not something that can address any of those challenges in the next 15 to 20 years," Daly explained. Small modular reactors could theoretically reduce construction times, but the technology remains relatively immature, and regulatory, licensing, and supply chain obstacles could add years to development timelines.

Ireland already generates substantial renewable electricity. Wind supplied 32% of electricity generation in 2025, while gas accounted for around 40%, and electricity imports represented approximately 16%. Yet even this renewable-heavy mix cannot reliably meet the demands of hyperscale AI infrastructure operating 24/7.

How Are Global Tech Companies Securing Power for AI?

Ireland's nuclear debate reflects a much wider shift among data center operators worldwide. In January 2026, Meta announced agreements with Vistra, TerraPower, and Oklo that could support up to 6.6 gigawatts of new and existing nuclear capacity by 2035. The company explicitly linked nuclear power to its AI infrastructure strategy, signaling that reliable baseload electricity has become as critical as silicon chips.

Across the Atlantic, the pattern is even more pronounced. Four major technology companies have contracted roughly 18.8 gigawatts of firm generation capacity. Meta leads at 7.72 gigawatts across multiple nuclear commitments. Amazon follows at 7.24 gigawatts, dominated by a small modular reactor commitment of up to five gigawatts by 2039, plus 1.92 gigawatts from Talen's Susquehanna plant running to 2042. Alphabet holds 2.3 gigawatts through Elementl and Kairos. Microsoft's 1.53 gigawatts includes the single most symbolically loaded deal in the sector: an 835-megawatt twenty-year agreement to restart Three Mile Island.

"State-of-the-art data centers and AI infrastructure are essential to securing America's position as a global leader in AI. Nuclear energy will help power our AI future, strengthen our country's energy infrastructure and provide clean, reliable electricity for everyone," said Joel Kaplan, Chief Global Affairs Officer at Meta.

Joel Kaplan, Chief Global Affairs Officer, Meta

What makes these contracts remarkable is not just their scale, but what they reveal about corporate strategy. These are not energy companies. They are buyers of artificial intelligence, reaching two stations back down the supply chain to secure the one input that cannot be improvised or rented. You can lease compute capacity. You can license a model. You can hire researchers. You cannot conjure five gigawatts of firm baseload electricity on short notice.

Steps to Understanding the Power-Compute-AI Supply Chain

• Power Generation: The first station in the AI production line, measured in terawatt-hours per year or gigawatts of capacity. This includes turbines, reactors, gas plants, dams, and solar panels that generate electricity.
• Compute Infrastructure: The second station, where electricity is converted into calculation through data centers, accelerators, high-bandwidth memory, cooling systems, and switchgear. Nothing downstream can exceed what this station releases.
• AI Capability: The third station, where compute is converted into actual intelligence through training runs, model weights, and inference serving. This is where the excitement lives, but it depends entirely on the two stations before it.

United States data centers now draw roughly 29.2 gigawatts of continuous power. Against a national generation base of about 4,520 terawatt-hours in 2025, this works out to roughly 256 terawatt-hours per year, or about 5.6% of all electricity the country generates. That figure has a dramatic history. American data centers consumed roughly 176 terawatt-hours in 2023, about 4.4% of national electricity. Credible projections for 2028 range from 325 to 580 terawatt-hours, representing an increase of 80% on the low end to more than tripling on the high end. The gap between those two numbers is roughly the annual electricity consumption of a mid-sized industrial economy, and no one can predict which end the market will land on.

The International Energy Agency projects global data-center electricity consumption roughly doubling from about 485 terawatt-hours in 2025 to around 950 terawatt-hours by 2030, with the AI-focused portion growing even faster. That trajectory is already visible in interconnection queues and turbine order books. Efficiency improvements in AI models have reduced power consumption per task, but absolute electricity demand continues rising because the volume of inference and the size of training runs have grown even faster.

For Ireland, the question is therefore not simply whether nuclear energy is politically acceptable. It is whether the country can build enough reliable electricity infrastructure to support the data centers that have become central to its digital economy, and how quickly any new generation could arrive. Finland has become a recurring comparison in Ireland's nuclear discussion. The country has a similar population but generates a significant share of its electricity from nuclear power and has some of the European Union's lowest non-household electricity prices.

The stakes extend beyond Ireland. As AI infrastructure demands continue to accelerate globally, every nation with ambitions in artificial intelligence faces the same fundamental constraint: electricity. The companies that secure reliable, low-carbon power will determine the geography of AI development for the next decade. Ireland's nuclear debate is not an outlier. It is a preview of the energy politics that will define the 2030s.