The nuclear renaissance is real, but it's not playing out the way Silicon Valley expected. While companies like Oklo and Helion capture investor attention with cutting-edge reactor designs, a less glamorous competitor is actually delivering the goods: established nuclear manufacturers with decades of experience, steady revenue, and the financial stability to see long-term projects through.

The problem is stark. Oklo, despite its 704% stock surge last year and high-profile Air Force contracts, generated zero revenue in 2025 and posted a $139.3 million operating loss. The company is entirely dependent on government grants and investor funding to operate. Its Aurora reactor uses innovative liquid sodium cooling and the company is pioneering nuclear fuel recycling, but neither technology is generating income yet.

Meanwhile, Microsoft faces an urgent deadline. The tech giant is constructing a sprawling data center campus near Malaga, Washington that could demand as much electricity as a medium-sized city. The company has partnered with Helion, an Everett-based fusion startup, to deliver clean power by 2028. But nuclear experts are deeply skeptical. Helion has achieved fusion reactions before, but no fusion device has ever produced more electricity than it consumes. The company's latest prototype reached 150 million degrees Celsius, short of the 250 to 300 million degrees needed for commercial operations.

Why Established Nuclear Companies Are Winning the AI Power Race?

BWX Technologies tells a different story. The company has built over 400 naval nuclear reactors since the USS Nautilus, giving it unmatched experience in compact reactor design. Unlike Oklo, BWX generates substantial revenue: $3.19 billion in 2025, up 18% from the previous year. The company is profitable with a 10.3% net profit margin and grew earnings per share by 20% year-over-year.

BWX is developing its own small modular reactor called the BANR (BWXT Advanced Nuclear Reactor), which can generate 75 megawatts of electricity. Because the company has diverse revenue streams beyond SMRs, it can develop this technology with far less financial risk. This matters enormously for data center operators who need reliable power, not speculative bets on unproven technology.

The broader nuclear landscape is shifting rapidly. TerraPower broke ground on Kemmerer Unit 1 in Wyoming, America's first utility-scale advanced nuclear power plant. The 345-megawatt Natrium reactor uses sodium cooling paired with molten salt energy storage and can ramp output to 500 megawatts for over five hours to handle peak demand. TerraPower has also signed an agreement with Meta for up to eight Natrium plants by 2035 to support data center power needs.

Kairos Power broke ground on its Hermes 2 Demonstration Plant in Oak Ridge, Tennessee, marking the first power-producing Gen IV reactor to receive a construction permit from the Nuclear Regulatory Commission (NRC). This 50-megawatt facility will feed power directly into the Tennessee Valley Authority grid under an agreement with Google to help decarbonize the tech giant's data centers in Tennessee and Alabama.

How to Evaluate Nuclear Power Solutions for Data Centers

• Revenue and Profitability: Companies with established revenue streams and positive profit margins can weather delays and technical challenges without collapsing. Oklo's zero revenue makes it vulnerable to funding disruptions, while BWX's $3.19 billion in annual revenue provides stability.
• Proven Track Record: Experience building reactors in constrained spaces matters. BWX's 400-plus naval reactors demonstrate the company can deliver in demanding environments, whereas Helion's fusion technology has never achieved net-positive electricity generation anywhere on Earth.
• Regulatory Progress: Construction permits from the NRC signal serious progress. TerraPower and Kairos Power have already broken ground on actual facilities, while Helion is still racing to build its first commercial plant by 2028.
• Contractual Commitments: Real agreements with major tech companies matter more than press releases. Meta's commitment to eight Natrium plants and Google's deal with Kairos Power represent genuine demand, not speculative partnerships.

The skepticism about Helion's timeline is widespread among nuclear experts. Paul Wilson, chair of the nuclear engineering department at the University of Wisconsin-Madison, stated: "I am skeptical of their timeline. It would be surprising to me if they're delivering electricity on a routine basis in 2028".

Tom Lyon, a corporate sustainability researcher at the University of Michigan, goes further, calling Microsoft's deal with Helion "a remarkably clever new form of greenwashing. It's this really creative way to look like you've made a commitment without a commitment".

The data center power crisis is real and urgent. Nearly half of the world's proposed data center projects are delayed largely due to power constraints. Growing local opposition blocked or delayed about $156 billion worth of projects in the United States in 2025 alone. In Seattle, a proposal to add five large data centers generated such strong backlash that the mayor announced she was considering a moratorium on new projects.

But the solution isn't necessarily the flashiest technology. In Chelan County, Washington, utility staff warned that continued data center growth could increase the risk of outages and voltage instability. A wave of proposed data centers following Microsoft could together demand nearly as much electricity as Seattle uses on an average day.

Advanced reactor development is accelerating across the industry. The Air Force selected three microreactor developers for its Advanced Nuclear Power for Installations program, with goals to operate at least one advanced nuclear reactor on an Air Force installation by 2030 or sooner. Radiant Nuclear will deploy a portable 1-megawatt helium gas-cooled microreactor at Buckley Space Force Base in Colorado. Westinghouse will deploy its eVinci microreactor at Malmstrom Air Force Base in Montana, producing about 5 megawatts with eight-plus years of fuel life. Antares Nuclear will build at Joint Base San Antonio in Texas with its R1 microreactor, delivering 100 kilowatts to 1 megawatt with over six years between refueling.

Blue Energy raised $380 million to build standardized, prefabricated nuclear power plants that can be assembled largely offsite, much like ships or offshore platforms. The company targets up to 1.5 gigawatts for its initial Texas site, with early works planned for later in 2026 and a final investment decision in 2027.

The lesson for investors and data center operators is clear: proven execution beats speculative innovation when the stakes are this high. Oklo and Helion may eventually deliver breakthrough technology, but established players with revenue, experience, and regulatory approval are the ones actually powering the AI revolution today.