A startup is building a 3D-printed nuclear reactor module to feed power-hungry data centers

The explosive growth of artificial intelligence has created a problem few people outside the power industry saw coming: the data centers that train and run AI models consume staggering amounts of electricity, and the grids meant to supply them are straining to keep up. Into that gap has stepped a wave of nuclear startups, and one has drawn attention for an especially unconventional idea — using 3D printing to manufacture a compact reactor module designed to sit near a data center and feed it power directly. The approach reflects a broader bet that small, factory-built reactors, rather than sprawling traditional plants, are the fastest way to add clean, round-the-clock power for computing.

Data centers need electricity that is not only abundant but constant, running every hour of every day regardless of weather. That requirement is what makes nuclear attractive to the companies scrambling to power AI: unlike solar or wind, a reactor produces steady output around the clock, and unlike fossil-fuel generation, it does so without the carbon emissions that many technology firms have pledged to eliminate.

The 3D-printing pitch

According to a report from The Register, the startup’s central idea is to use additive manufacturing — building components layer by layer rather than casting and machining them the traditional way — to produce a reactor module aimed specifically at data centers. The appeal of that method is speed and cost: 3D printing can create complex shapes in fewer steps, potentially compressing the years-long, capital-intensive process of fabricating reactor parts and making it easier to produce units at scale.

The strategy fits a wider shift in how some companies think about nuclear power. Rather than building enormous one-off plants over a decade or more, the goal is to standardize a smaller reactor design and manufacture it repeatedly, the way a factory turns out any other product. If that model works, each additional unit becomes cheaper and faster to build than the last — the opposite of the cost overruns and delays that have plagued large conventional nuclear projects.

Why AI is driving a power crunch

The demand behind all of this is real and growing quickly. Training a large AI model requires vast arrays of specialized chips running at full tilt for extended periods, and serving those models to millions of users keeps the hardware busy afterward. All of that computing generates enormous electricity and cooling needs, and the largest data centers now draw power on a scale comparable to that of small cities.

Utilities and grid operators, built and planned around slower, more predictable growth, have found themselves confronting sudden demand that outpaces their ability to add capacity. In some regions the wait to connect a large new load to the grid stretches for years, which is precisely why technology companies have grown interested in generating power on site or nearby rather than relying entirely on the existing system. A dedicated reactor next to a data center sidesteps the queue and guarantees a steady supply.

The advanced-reactor landscape

The startup is one player in a much larger field of companies developing what the industry calls advanced reactors and small modular reactors — designs smaller and often simpler than the gigawatt-scale plants that make up most of today’s nuclear fleet. The U.S. Department of Energy, through its advanced reactor technologies program, has backed research into these next-generation designs, which aim to be safer, cheaper and quicker to deploy than their predecessors.

Small modular reactors are meant to be built in a factory and shipped to a site largely complete, reducing the on-site construction that drives up cost and risk in traditional projects. That manufacturing-first philosophy is the same one underpinning the 3D-printing approach, which pushes the concept further by rethinking how the reactor’s components themselves are made.

The obstacles between concept and reality

An ambitious design is a long way from a working power source, and nuclear technology faces some of the steepest hurdles of any industry. New reactor designs must clear extensive regulatory review before they can operate, a process that scrutinizes safety, materials and manufacturing quality in exhaustive detail — and using 3D-printed components in a reactor introduces questions regulators will want thoroughly answered about how those parts perform under heat, pressure and radiation over years of service.

Cost, supply of nuclear fuel, public acceptance and the sheer time required to license and build a first-of-its-kind system all stand between the concept and a reactor actually powering a data center. Many nuclear startups have promised fast timelines and then run into the slow realities of the field. Still, the intensity of the AI power crunch has injected fresh money and urgency into nuclear development, and unconventional ideas like a 3D-printed reactor module reflect how far companies are willing to reach for a steady, carbon-free way to keep their servers running. Whether this particular approach reaches the grid remains to be seen, but the demand pulling it forward shows no sign of easing.

This article was researched and written with the assistance of AI and reviewed by an editor prior to publication.

More from Morning Overview