Via interestingengineering.com
Applied Atomics has licensed the dormant mPower small modular reactor, betting a decade-old design can meet the electricity demands of the AI era.
A nuclear reactor design that burned through roughly $400 million in investment before being quietly shelved in 2017 is getting a second life, courtesy of two engineers who used to build rockets for Elon Musk.
Ben Kellie and Paul Keutelian, both former SpaceX engineers, founded Applied Atomics in 2025 with a specific goal: take a proven but abandoned reactor concept and route its output directly to the data centers that AI companies are desperately trying to keep powered. On June 17, 2026, the company signed a licensing agreement with BWXT, giving Applied Atomics exclusive commercial rights to the mPower small modular reactor design across the US and Canada.
What exactly they’re reviving #
The mPower reactor is a Generation III+ light-water small modular reactor rated at 195 megawatts electric, or 575 megawatts thermal. For context, a single large conventional nuclear plant typically generates around 1,000 MW electric, so mPower sits in a middle tier: smaller and faster to build than traditional nuclear, but meaningfully more powerful than the micro-reactors that often dominate SMR headlines.
BWXT originally developed mPower with partner Babcock and Wilcox, drawing significant federal and private support before the project stalled under the weight of cost overruns and a difficult regulatory environment. The design did not disappear entirely. Test facilities in Lynchburg, Virginia were preserved, which means Applied Atomics is not starting from a blank page. The prior $400 million in development work is effectively a head start baked into the licensing deal.
Kellie has pointed to the team’s aerospace background as a genuine advantage. SpaceX built its reputation on treating rocket hardware with manufacturing discipline that the traditional aerospace industry had long abandoned. The implicit pitch at Applied Atomics is similar: apply that same rigor to a reactor design that already works on paper but historically struggled to make it from paper to grid.
Why data centers, and why now #
Small modular reactors occupy an appealing theoretical position in this landscape. They can, in principle, be sited closer to load centers than giant conventional plants, built in modular increments as demand grows, and manufactured with more factory standardization than one-off megaprojects. No SMR design has reached full commercial operation in the US.
Applied Atomics is targeting first commercial operation around 2031. In partnership with Core Power, Applied Atomics is conducting a feasibility study on floating nuclear plants, essentially mounting mPower reactors on marine platforms. The logic is primarily about permitting: international waters and purpose-built offshore zones may offer regulatory routes that are faster than the standard US land-based NRC process. Russia has operated a floating nuclear plant, the Akademik Lomonosov, since 2019.
The skeptics have a point #
The SMR sector has seen a wave of enthusiasm over the past several years, and Applied Atomics is entering a field that is simultaneously crowded with ambition and thin on completed projects. Critics of the broader SMR push tend to raise three recurring objections: cost, regulation, and waste.
On cost, the concern is structural. Nuclear projects in the West have a persistent tendency to run over budget by multiples of the original estimate. The mPower design’s prior abandonment was partly a cost story. Licensing a design that previously stalled does not automatically solve the underlying economics, especially when the construction cost environment has only become more expensive since 2017.
On regulation, the NRC process is thorough by design, and thoroughness takes time. A reactor design that was partway through licensing review nine years ago will not simply pick up where it left off.
On waste, civilian nuclear power in the US still lacks a permanent storage solution. That is a political and infrastructure problem that no startup can resolve unilaterally, and it remains a long-term liability for any operator.
What Applied Atomics has going for it is specificity. Rather than designing a novel reactor from scratch, the company is licensing a design with hundreds of millions of dollars of prior engineering work embedded in it, supported by preserved physical test infrastructure in Lynchburg, Virginia.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our