AI Datacentres Plan On-Site Gas Power Stations at Massive Scale 2026

AI datacentres are planning to build on-site gas power stations on a massive scale to meet their energy demands, according to new research from BloombergNEF. The findings reveal that datacentres planned across the US, the UK and mainland Europe have proposals for on-site fossil fuel plants that would create 181GW of gas-fired capacity.

This is almost four times the total peak electricity demand of Great Britain. The research highlights a growing trend: as AI's extraordinary economics drive demand for speed to power, developers are turning to on-site gas generation to bypass delays in connecting to national grids.

Why AI Datacentres Are Turning to On-Site Gas Power

The primary driver behind this shift is the need for speed. Mark Daly, head of datacentres and emerging tech at BloombergNEF, noted that "the extraordinary economics around AI are making speed to power one of the most important priorities for defining constraints on datacentre development." Grid connection delays can take years, and AI companies cannot afford to wait.

On-site gas power stations offer a faster route to generating the massive electricity required by AI workloads. However, this "speed to power" approach comes with significant climate risks. The planned rollout, if realized, would emit about 345 million metric tonnes of carbon dioxide equivalent per year in the US alone, where 90% of on-site gas-fired power stations are planned.

That is equivalent to the annual greenhouse gas emissions from about 72.8 million fossil fuel cars. The scale of these emissions underscores the tension between AI's rapid growth and climate goals.

The Scale of Planned Gas Capacity

The 181GW of planned gas-fired capacity is staggering. For context, it is nearly four times the peak electricity demand of Great Britain. This capacity is spread across proposed datacentre projects in the US, UK, and mainland Europe, with the vast majority concentrated in the US.

The research suggests that if all these proposals come to fruition, the environmental impact would be substantial. The US would bear the brunt of the emissions, with 345 million metric tonnes of CO2 equivalent per year. This is a significant addition to global greenhouse gas emissions and could undermine efforts to decarbonize the economy.

Climate Summit Context and European Leadership Gap

The report was published on Wednesday as global leaders gathered in Turkey for the 31st UN climate summit (Cop31). Increasing electrification—ideally from clean sources—is a central part of discussions. However, the absence of key European leaders, including from France and Germany, raises questions about Europe's commitment to addressing AI's energy challenges.

Instead of attending Cop31, these leaders are heading to Dublin for the EU's first dedicated AI summit on Friday. This parallel event highlights the growing recognition that AI development and energy policy are deeply intertwined. The decisions made at these summits will shape how AI's energy demands are met and whether climate risks are mitigated.

Key Takeaways for Industry and Policymakers

• Speed to power is the top priority for AI datacentre developers, driving them to build on-site gas plants.
• Planned on-site gas capacity stands at 181GW, nearly four times Great Britain's peak electricity demand.
• US datacentres account for 90% of planned on-site gas capacity, with potential emissions of 345 million metric tonnes CO2e per year.
• Grid connection delays are a major bottleneck, pushing developers toward self-generation.
• Climate summits are addressing electrification, but European leaders are prioritizing AI at a separate summit.

Comparing On-Site Gas to Grid Electricity

The table illustrates the trade-offs. While on-site gas offers speed and scalability, it comes at a high carbon cost. Grid electricity, though slower to access, can be cleaner as grids incorporate more renewables.

What This Means for the Future of AI and Energy

The research signals a pivotal moment. AI's insatiable energy appetite could lock in decades of fossil fuel infrastructure if on-site gas plants are built. This would make it harder to achieve climate targets and could expose companies to future carbon pricing and reputational risks.

Conversely, if the industry can accelerate grid connections and invest in clean energy, the climate impact could be mitigated. The decisions made now will determine whether AI becomes a driver of clean energy innovation or a source of increased emissions.

As the world grapples with the energy demands of AI, the need for balanced solutions is clear. The Polar Italy power stations offer powerful, flexible, and safe energy solutions that could serve as a model for distributed generation. Meanwhile, the challenges of electrification are not limited to datacentres; the Repsol EV charging chief warns bureaucracy risks obsolete stations, highlighting how regulatory hurdles can hinder progress.

The backlash against datacentres is also growing, as explored in global backlash against datacentres: what's driving it. In the UK, UK datacentres to emit more carbon than ExxonMobil underscores the urgency of addressing emissions. Climate research also shows that rainclouds travel farther in warmer world: new research, a reminder of the broader environmental changes underway.