Google and Amazon are betting on ‘advanced’ nuclear. This critic warns it’s not ready

As tech companies scramble to find new power sources for AI’s huge energy needs, some of them are turning to startups developing new nuclear technology. Google recently announced that it plans to begin using power from Kairos Power’s small modular reactors by 2030. Amazon is investing in another nuclear startup, X-Energy. Microsoft hasn’t yet announced a similar investment, though it posted a job listing last year that included looking at “optimal integration” of next-generation nuclear reactors.

The newest nuclear tech claims to be safer and more sustainable than traditional nuclear power plants. But some critics argue that “advanced” nuclear technology isn’t necessarily that advanced—and that it’s unlikely to be ready on the timeline that Big Tech wants.

“I think it’s highly unlikely that these reactors are going to perform the way that their developers are promising,” says Ed Lyman, director of the nuclear power safety program at the nonprofit Union of Concerned Scientists.

X-Energy makes a small reactor filled with fuel “pebbles”—each around the size of a billiard ball—that contain thousands of tiny particles of uranium that are each surrounded by layers of carbon. (The type of fuel is called TRISO, or “tristructural isotropic” fuel; the Department of Energy calls it “the most robust nuclear fuel on Earth,” a claim Lyman says is “wildly overhyped.”) The fuel continuously rotates through the core, along with helium that absorbs the heat; the heat turns water into steam to make electricity. Each reactor is about 11 feet in diameter and a little more than 15 feet high, versus a massive traditional nuclear plant with cooling towers that are hundreds of feet tall. The company calls its design “meltdown-proof” and says the particles “retain their integrity under all foreseeable conditions.”

Lyman, who wrote a detailed report about next-gen nuclear reactors in 2021 and closely follows the industry, argues that it’s too early to say that it’s safe. “X-Energy’s specific fuel type has not yet been tested under any circumstances,” he says. When another TRISO fuel was tested in a reactor at Idaho National Laboratory, the experiment had to stop early because it was releasing high levels of radioactive cesium at certain temperatures. X-Energy says its fuel is so inherently safe that a containment building isn’t necessary; Lyman disagrees. The reactors could be vulnerable to air or water leaks, and the fuel has to be made to exacting specifications that haven’t yet been proven. “They’re still kind of basing all their safety analyses on optimistic assumptions,” he says.

X-Energy says the reactors run efficiently, using more than 90% of the available uranium in each pebble. But Lyman says it’s less efficient than traditional nuclear power plants, generating more radioactive waste. The spent fuel will be stored on site for the 60-year life of the reactor, but then the DOE will have to store it in a geologic repository (so far, these repositories don’t exist, and nuclear waste from decades of older nuclear power plants is still piling up).

Kairos also uses TRISO fuel, with a different type of reactor. Lyman argues that it has other challenges. The coolant that the company uses is corrosive, he says, and it could be difficult to find materials for the reactor that won’t be damaged by it. In response, the company says it has done thousands of hours of testing with “very little corrosion” under normal operating conditions; still, Lyman says it’s too early in the process to know how the reactor will truly perform. Construction started on the company’s demonstration reactor in July in Oak Ridge, Tennessee.

The full-scale reactor will also release more tritium, a radioactive isotope, than existing nuclear power plants. Lyman argues it could harm the environment. Kairos says the levels released “do not pose a significant risk to public health or the environment,” noting that some level of tritium naturally exists in groundwater, and that it will “conduct regular monitoring and mitigation efforts to limit any tritium releases.”

The company will also face the challenge of sourcing the fuel it needs, a more enriched form of uranium than current nuclear plants use; right now, not enough of it is produced. But the company argues that customer demand, like that from Google, “sends a meaningful signal” to producers.

For tech companies, small modular reactors are appealing because their size means that they can be built closer to the grid than traditional nuclear power plants. They’re faster to build, and cost less. “We remain committed to renewable energy and have been the largest corporate purchaser for four years in a row,” an Amazon spokesperson told Fast Company. “But we know we’ll also need other sources of reliable carbon-free energy to meet the growing needs of our business. Expanding our energy investment strategy to include other forms of carbon-free energy, including nuclear, is most viable option to help bridge this gap.”

The “‘always-on’ nature of nuclear is valuable,” says a Google spokesperson, noting that the company is continuing to deploy new wind, solar, geothermal, and battery storage projects. (Geothermal power, like nuclear, can also provide the 24/7 energy that data centers need, and Google recently worked with a startup to help deploy a next-gen geothermal power plant that could be used in more locations.)

Kairos says it’s possible to come to market very quickly. “The conventional nuclear development model is long, slow, and capital-intensive,” a spokesperson says. “Kairos Power is disrupting that model by embracing a rapid iterative approach, which accelerates test cycles for innovation and optimization.” For its new demonstration reactor, the company got the first construction permit for a non-traditional nuclear reactor to be issued by the Nuclear Regulatory Commission in more than 50 years. The company aims to begin running the demonstration reactor in 2027, and then have the first reactor for Google running just three years later.

Lyman is skeptical that new plants could be ready to run on schedule by the end of the decade. “Tech companies see themselves as visionaries,” he says. “They don’t like to hear naysayers. They think they’re the masters of the universe. But the fact is that nuclear power is kind of resistant to significant improvements that reduce the cost in time to build safe facilities. There’s just a lot of development and it’s very slow and painstaking work. It’s not building something in your garage with circuit boards.” The projects will also have to get regulatory approval from the Nuclear Regulatory Commision. (Other attempts to reopen older nuclear power plants that have been closed—like Microsoft’s plan to restart a reactor at Three Mile Island—could also potentially face hurdles.)

It makes sense that tech companies with aggressive goals to cut emissions would be looking for multiple new solutions. But given the immediate need for more clean energy for data centers now, it’s not clear why they aren’t filling the gap with even more of the renewable solutions that already exist—and don’t produce nuclear waste.

“I certainly don’t think any of these timelines are realistic,” Lyman says. “At best, they’re going to get a few boutique reactors that will be really expensive to run, and will probably end up being more of a headache for them than anything else.”