Forget Nuclear: The Old-School Energy Source Quietly Winning the AI Power Race

While the headlines obsess over nuclear restarts and small modular reactors, a very different fuel is actually powering the GPU racks running today's AI boom, and the companies sitting on its delivery network are quietly stacking record backlogs.

Published June 3, 2026, 7:04am ET · 3 min read

A wide-angle photograph of a massive industrial pipeline extending from the right foreground over calm water, supported by pillars, towards a distant, brightly lit industrial complex. In the background, multiple smokestacks emit plumes of smoke, and various cranes and factory buildings are illuminated, reflecting on the water's surface under a warm orange and grey sunset sky.
This expansive industrial complex and pipeline symbolize the increasing demand for natural gas, driven by the energy needs of new AI data centers. Such robust infrastructure is vital for the midstream sector, reflecting significant investment opportunities. © Spooh / Getty Images

The AI power conversation is dominated by nuclear restarts and small modular reactor headlines. The electrons heating GPU racks today come from natural gas. I’ve been tracking the AI infrastructure buildout for over a year, and the picture keeps sharpening: gas is winning because it can be built behind the meter, at gigawatt scale, with build cycles measured in months.

Look at xAI’s COLOSSUS II, one of the world’s largest AI training data center clusters, powered by a self-built behind-the-meter gigawatt-scale natural gas power plant. Operators have explicitly said their ability to scale depends in part on continued access to natural gas supply at economically feasible prices, the availability of gas turbines and related equipment, and the maintenance of a regulatory environment that permits and supports the use of natural gas for large-scale power generation.

The macro setup is wild. U.S. electricity generation was effectively flat from 2008 to 2023 at a 0.1% CAGR, and growth between 2023 and 2025 has only modestly accelerated to under 3% annually. AI compute demand is exploding into a grid that forgot how to grow.

The pipelines are the toll road

Kinder Morgan (NYSE:KMI | KMI Price Prediction) says approximately 70% of future data center power demand markets sit on its assets. The $10 billion backlog is approximately 90% natural gas and nearly 60% supporting power generation. CEO Kim Dang said "total demand for natural gas is expected to grow by 17% through 2030, led by LNG exports."

Williams Companies (NYSE:WMB) is running the same playbook. FY2025 Adjusted EBITDA hit a record $7.75 billion, up 9% YoY. CEO Chad Zamarin announced "Socrates the Younger," lifting power innovation capital to over $7 billion in execution, with the first project online in H2 2026. Shares are up 20% YTD.

The Appalachian producers

EQT (NYSE:EQT) delivered a Q1 2026 stunner: adjusted EPS of $2.33 beat $2.16, revenue of $3.378 billion topped $3.24B, with record free cash flow of $1.83 billion. CEO Toby Z. Rice said "accelerating power demand growth in the United States – particularly in Appalachia – is creating incremental opportunities in our backyard." Fitch upgraded the credit to BBB.

Antero Resources crushed too, with EPS of $1.72 and revenue of $1.95B. Antero sells meaningful volumes along the LNG fairway. CNX Resources beat by a wide margin on EPS, with revenue that topped expectations.

The equipment chokepoint

GE Vernova (NYSE:GEV) is the bottleneck nobody can route around. In Q1 2026 the Electrification segment booked $2.4 billion in equipment orders to support data centers, more than all of last year. Management raised the year-end 2026 combined gas turbine backlog target to at least 110 GW. Shares are up 48% YTD.

The bottom line

Henry Hub spot near $3.10/MMBtu looks soft against the $5-plus realized prices producers tout, but the structural story is about volume growth. You’d want this thesis if you believe AI compute keeps outrunning the grid. If you think nuclear arrives in time, the urgency fades. I’m watching gas turbine lead times more closely than uranium spot.

Contact [email protected] for any questions or corrections.

Jeremy Phillips

I've been writing about stocks and personal finance for 20+ years. I believe all great companies are tech companies in the long run, and I invest accordingly.

All articles →