AI data centers turn to industrial boilers and steam turbines as gas turbine backlogs grow
Babcock & Wilcox lands a $2.4 billion deal to build boiler-steam plants for Applied Digital as AI power demand outpaces conventional supply chains.
Gas turbines, the traditional workhorses for new power generation, now carry multi-year backlogs. Data center operators are turning to industrial boilers paired with steam turbines as an alternative. Babcock & Wilcox, a 157-year-old company, has signed a $2.4 billion design-build agreement to deliver 1.2 GW of natural gas-fired boiler-steam capacity for Applied Digital’s AI campuses. B&W says it can deliver packaged boiler units in about one year, a fraction of the wait time for large-frame gas turbines.
Why boilers are suddenly cool again
AI-scale data centers are placing orders for hundreds of megawatts to gigawatt-scale gas turbines, and those orders now represent around 20% of some manufacturers’ entire gas power backlog. Natural gas-fired boilers coupled with steam turbines offer a workaround. B&W claims these packaged systems can match the peak-hour efficiency of some gas turbines.
Siemens Energy agreed to supply 20 steam turbine-generator sets, each rated at 50 MW, totaling 1 GW of capacity as part of its arrangement with B&W. That agreement carries an August 2026 notice-to-proceed timeline.
Applied Digital’s North Dakota bet
Applied Digital’s site in Ellendale, North Dakota, had already achieved 50 MW of online capacity by late 2025. B&W’s 1 GW capacity initiative for Applied Digital is focused on natural gas-fired plants expected to be operational by 2028, with a notice to proceed in March 2026.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Industry projections suggest AI-scale facilities could demand tens of gigawatts of new capacity by 2030. A single gigawatt is roughly the output of a large nuclear plant.
Energy markets feel the pull
Natural gas is emerging as the default fuel for AI infrastructure because it offers dispatchable, round-the-clock power that can be sited near the data center itself. As hyperscalers and AI labs secure their own on-site generation, they are effectively bypassing the utility grid for baseload power, taking on the operational complexity of running their own power plants.