DOE backs AI-driven reactor design with $293 million as federal baseload push widens
Federal funding for Prometheus and a West Virginia coal plant reveals Washington's dual bet on advanced nuclear and clean coal to close a widening supply gap.
The Department of Energy has committed $293 million through a competitive program to Prometheus, a partnership between Idaho National Laboratory and NVIDIA that uses artificial intelligence to accelerate nuclear reactor development. The project targets cuts in reactor build times of up to 50% and reductions in operating costs by a similar margin.1
US electricity demand is accelerating faster than new baseload supply is arriving, driven by data centers, manufacturing reshoring and electrification. The DOE's commitment to Prometheus signals the agency sees advanced nuclear as a viable answer, not a distant one.1
The project addresses both the design and operational phases of small modular reactors and microreactors. By deploying AI to simulate reactor physics and streamline regulatory filings, INL and NVIDIA aim to collapse a timeline that has historically stretched a decade or more. Whether they can do it is a different question.1
The same week the Prometheus commitment was disclosed, the DOE awarded $18.5 million in development funding to the TerraSpark Energy Campus, a 1.6-gigawatt greenfield coal plant in Grant County, West Virginia, pairing Babcock & Wilcox supercritical boilers with Mantel Capture's molten borate carbon capture system, according to Power Magazine. Combined with $21.5 million in non-federal cost share, the scoping and design phase carries a total value of roughly $40 million.2
TerraSpark's award sits inside a $525 million DOE coal program launched in September 2025 under President Trump's executive orders on clean coal and grid reliability — distinct from the nuclear funding stream. The developer targets a 2030 startup date and a 95% to 98% CO2 capture design, with four pathways outlined for the captured carbon.2
The two awards together reveal the breadth of Washington's baseload bet. One agency is simultaneously funding SMR design via AI and greenfield coal through separate appropriations streams. Both respond to the same demand-side pressure renewables have not closed on utility timelines.
But TerraSpark's slice is small. Its $18.5 million is the smallest of the four Topic Area 1 awards in the coal program — well behind AES Puerto Rico's $164.5 million for a retrofit and life extension of its 510-megawatt Guayama plant and AES Warrior Run's $78 million for a carbon capture FEED study.2
The Prometheus commitment carries its own caveat. SMR development has been set back repeatedly by supply chain disruptions and technical difficulties since the early 2020s, and no commercial SMR is operating in the United States. The DOE's $293 million is a substantial research commitment, yet it falls well short of the capital required to bring a single reactor to financial close.3
Canada is moving along a parallel track. Ottawa released a national nuclear energy strategy on June 29 (2026-06-29) and announced the following day that seven Williams Treaties First Nations would take minority stakes in the Darlington SMR project through the Indigenous Loan Guarantee program — a structure that links project financing to Indigenous equity participation rather than grant funding alone.4
For gas markets, the Prometheus timeline is the variable worth tracking. If AI-assisted design can deliver a construction-ready SMR in roughly half the traditional timeframe, baseload gas demand growth in the mid-2030s looks softer than current utility integrated resource plans imply. If Prometheus stalls — as earlier SMR programmes have — coal with CCS becomes the only scalable dispatchable alternative outside of gas. TerraSpark's 2030 target gives the market a concrete milestone against which to judge that outcome, though a scoping-and-design award is a long way from a shovel in the ground.2,13