The U.S. nuclear industry is learning a hard lesson: announcing an ambitious SMR project is easy. Building one that actually works and makes money is another thing entirely.
China is already constructing large-scale reactors at scale and is projected to surpass U.S. nuclear capacity by 2030. The strategic gap is widening.
AI data centers are hungry for power, and major tech companies are betting on small modular reactors to feed them. That corporate demand has accelerated funding and timelines. But demand alone cannot compress the engineering work required to move from concept to commercial operation.
Qualifying materials, building supply chains, running prototypes, and managing the physics at scale takes years—and cannot be rushed simply because electricity needs are rising. The current crop of reactor designs is more diverse than anything seen since early civilian nuclear power: sodium-cooled fast reactors, molten salt designs, high-temperature gas reactors, and pressurized water variants are all in development. Different applications demand different solutions. A small, transportable reactor for a remote site is not the same engineering problem as a 300-megawatt plant feeding a data center.
The Energy Department's Advanced Reactor Demonstration Program and the National Reactor Innovation Center have funded experimental facilities and demonstration projects. Several developers have achieved criticality—genuine physics milestones proving their designs work in theory.
But engineering triumph in the lab is not a ticket to the market. Vogtle Units 3 and 4 in Georgia used a modular design already approved by the Nuclear Regulatory Commission. Both still suffered major schedule delays and cost overruns. SMR designs remain largely unproven at commercial scale, leaving real questions about power output, fuel consumption, operating costs, and downtime that only deployment will answer.
The first U.S. SMRs expected to deliver power to commercial loads are targeted for around 2030. Realistically, these will be pilot projects, not the start of mass deployment.
