AI data centers are planning to install massive natural gas power plants on their own sites while they wait to connect to the national grid. The total capacity projected across the US and Europe reaches 181 gigawatts — roughly four times the United Kingdom's peak electricity demand. The research lays bare the gap between the industry's on-paper clean energy commitments and what's actually happening on the ground.

What happened?

According to the research, AI data center projects in the US, the UK, and continental Europe are proposing to build gas-fired power stations on-site to get around grid connection delays. The total proposed capacity sits at 181 gigawatts — a scale that shows just how long connection queues have grown, and that companies are choosing to turn to fossil fuels rather than delay their growth.

Why it matters

The root of the problem is that grid infrastructure can't keep pace with the speed of AI investment. In the US in particular, applications for new power plant and transmission line connections at major regional grid operators sit in queues that stretch for years — in some regions, more than five. A GPU cluster can be built and brought online in a matter of months, while transmission line or substation investments require far longer planning and permitting processes. That mismatch is pushing data center developers toward a logic of generating their own power first and figuring out the grid later.

This picture stands in stark contrast to the clean energy deals big tech companies have announced in recent years. Deals to restart shuttered nuclear plants, investments in small modular reactors (SMRs), and long-term nuclear power purchase agreements were all presented publicly with a zero-carbon emphasis. But most of these projects are still in construction or permitting, and bringing them online will take years. A gas turbine, by contrast, can be installed and running in a much shorter time. In the end, companies get to preserve their long-term clean energy promises while using gas as a bridge in the short term — but there's no guarantee that a 181-gigawatt bridge won't turn into permanent infrastructure.

Who does this affect, and how?

This choice has a direct impact on climate goals. On-site gas plants create an additional emissions source that falls outside companies' carbon-neutral commitments, because these plants are generally not counted against a company's own electricity consumption figures but instead show up in local air quality and regional emissions inventories. For local residents, the impact is more tangible: a gas plant means adding a source of air pollution and noise right next to a data center that's already a heavy water user for cooling.

SourceTime to bring onlineScale potentialCarbon impact
On-site gas turbineMonths to 1-2 yearsMedium-highHigh
Nuclear SMRYearsHighVery low
Grid connection (including renewables)Years (queue)HighVariable

The table sums up why companies are choosing gas: it's the only option that offers speed. But that speed carries the risk of growing regulatory and social costs over the long run.

What's next?

In Europe, the carbon cost of these plants could eventually come back to companies as a bill under the emissions trading system. In the US, state-level zoning and environmental permits are already running into local opposition; anti-data-center movements have become visible in regions with heavy data center concentration. The moment regulators take notice of this 181-gigawatt capacity, additional carbon taxes or stricter permitting processes could be on the table.

The resulting picture is this: AI companies are promising clean energy while operational reality is pushing them toward gas. This isn't a one-time choice — it's a pattern that will repeat every year that grid investment fails to keep up with AI demand. The gap between climate commitments and on-the-ground energy decisions will keep widening.