The reactor core at Idaho National Laboratory wasn't much to look at—roughly the dimensions of a standard shipping container, tucked inside a nondescript facility on the sprawling federal test site. But on June 4, 2026, when that sodium-cooled microreactor achieved sustained nuclear fission for the first time, it marked something larger than its modest footprint suggested.
Antares Nuclear, the startup behind the Mark-0 reactor, had just become the first private company to reach criticality under the Department of Energy's Reactor Pilot Program. And it had beaten the White House deadline by a full month—no small feat when that deadline was July 4, 2026, set by executive order with all the symbolism that date implies.
Energy Secretary Chris Wright called it "historic." What mattered more was the technical proof: a self-sustaining chain reaction using high-assay low-enriched uranium fuel, passively cooled sodium heat pipes, and a design intended for military bases and data centers rather than the electric grid. Zero-power criticality, as nuclear engineers call it. Just enough neutron activity to validate the physics. Not yet generating electricity, but demonstrating that the core worked as modeled.
For a nuclear industry that has spent decades over-promising and under-delivering, the milestone carried weight precisely because it was incremental. One reactor. One test. One month ahead of schedule.
The Executive Order and the Urgency Behind It
The timeline that Antares beat originated in Executive Order 14301, issued May 23, 2025, which tasked DOE with demonstrating at least three advanced test reactors by Independence Day 2026. The order was explicit about bypassing the Nuclear Regulatory Commission's traditional licensing path for experimental reactors on federal land—a move framed as "streamlining" but understood by everyone involved as an end run around regulatory review that could stretch years.
DOE launched the program in June 2025 and by August had named participants: Antares, Oklo, Radiant, Last Energy, Terrestrial Energy, and several others. The offer was straightforward—access to federal test infrastructure, allocations of HALEU fuel (still scarce after the Russian uranium ban took effect), and a regulatory pathway that could move faster than the NRC's legacy frameworks. For startups that had burned through Series A and B capital sitting in pre-application meetings, it was a lifeline.
Antares moved with unusual speed. The company had raised over $140 million by the time the reactor went critical in June 2026. It locked down a DOE fuel allocation, signed a multi-year enrichment contract with Urenco in May 2026, and contracted BWX Technologies to manufacture the TRISO fuel compacts—tiny, layered particles of uranium encased in ceramic and carbon coatings, each one a miniature containment vessel engineered to hold fission products even under extreme heat.
The choreography required to get fuel from enrichment to fabrication to transportation to assembly is not trivial, and Antares pulled it off in less than a year. Whether that proves replicable or was a one-time alignment of luck and federal priority remains an open question.
What the Mark-0 Actually Is
Strip away the policy theater and the Mark-0 is a microreactor: small, passively cooled, designed for resilience rather than efficiency. Sodium heat pipes transfer thermal energy without pumps or water—useful if you're powering a remote military installation in Alaska or a data center that can't afford downtime. TRISO fuel can withstand temperatures that would melt conventional reactor cores, a safety feature that simplifies some licensing headaches even as it introduces manufacturing complexity.
The zero-power test validated neutronics—that the fuel configuration could sustain a chain reaction—but didn't generate electricity. That comes next, assuming the company's timeline holds. Antares has said it will transition to a power-producing prototype in 2027, followed by production deployments in 2028. The Department of Defense shortlisted the company in May 2026 for its Advanced Nuclear Power for Installations program, alongside Radiant and Westinghouse Government Services, which would provide early revenue if those contracts materialize.
For BWX Technologies, the test meant something else: proof that TRISO manufacturing could scale beyond research reactors. For Urenco, the enrichment services contract signaled early commercial demand for HALEU in a market still adjusting to the loss of Russian low-enriched uranium supplies. The Russian ban, effective August 2024 with limited waivers through 2027, forced the U.S. to accelerate domestic HALEU production—though "accelerate" is relative when you're starting nearly from scratch.
DOE had extended Centrus Energy's HALEU demonstration contract through June 2026, with production milestones for the company's demonstration program. Antares consumed a fraction of that output, but the test proved the supply chain could function under deadline pressure. Whether it can function at commercial scale is another matter entirely.
AI's Insatiable Appetite and Nuclear's Second Act
The Antares milestone arrived at a peculiar moment in American energy policy. On January 13, 2026, the Energy Information Administration forecast commercial electricity sales rising 2.2% that year and 5.3% in 2027—the sharpest demand growth since the early 2000s, driven almost entirely by data centers. U.S. power generation hit a record 4,430 terawatt-hours in 2025, with Texas and the West South Central region leading the surge as hyperscale campuses proliferated.
A Lawrence Berkeley National Laboratory report from December 2024 had quantified the scale: data centers consumed 176 terawatt-hours in 2023, roughly 4.4% of U.S. electricity. By 2028, that figure could reach 325 to 580 TWh—as much as 12% of national demand, depending on how aggressively AI training and inference workloads grow. The International Energy Agency's 2026 outlook confirmed the trend wasn't uniquely American; data centers were reshaping electricity demand projections globally.
For nuclear startups that had spent the better part of a decade pitching aging coal plant replacements or remote mining operations, the narrative shift was clarifying. AI needs power. It needs it constantly. Wind and solar can supply daytime megawatts, but training a frontier language model at three in the morning requires firm capacity. Natural gas can fill that gap, but corporate net-zero pledges from the tech giants financing these facilities have made nuclear look, if not cheap, then at least strategically defensible again.
Antares isn't building grid-scale reactors. The Mark-0 and its commercial successor, the R1, target on-site power: military bases, industrial complexes, data centers that want captive generation and don't mind paying a premium for reliability. Whether the economics work at scale remains unproven, but the DOD interest suggests someone in the Pentagon believes distributed microreactors solve problems that centralized baseload plants don't.
Regulatory Reform and the Licensing Maze
While Antares tested under DOE authorization, the NRC was rewriting the rulebook for everyone else. On April 29, 2026, the agency's Part 53 licensing framework took effect—an optional, risk-informed, technology-inclusive path that promised lower costs and faster reviews than the legacy Part 50 and Part 52 frameworks designed for light-water reactors in the 1970s.
No one has completed a Part 53 application yet, so the cost and time savings remain theoretical. In April 2026, NRC proposed an even faster track: Part 57, a microreactor-specific framework targeting six- to twelve-month reviews for construction permits and operating licenses. The proposal, still under consideration as of mid-June 2026, aimed to enable "high-volume microreactor deployment," according to legal analysis from Orrick—language suggesting regulators were preparing for manufacturing-scale production rather than one-off prototypes.
The DOE pathway that Antares used offers speed but limited commercial upside. Reactors built under DOE authorization can't be sold into the merchant power market without separate NRC licenses, which negates much of the time advantage. Still, for startups racing to demonstrate technology and secure DOD contracts before their capital runs out, the pilot program provided a faster on-ramp than NRC review.
Other companies were threading the needle differently. Radiant Nuclear entered NRC review for a Part 70 materials license in May 2026, planning fueled tests at INL's DOME facility. Oklo's subsidiary, Atomic Alchemy, secured an NRC materials license in March 2026 for isotope production—a parallel revenue stream while its Aurora microreactor works through licensing. TerraPower received an NRC construction permit for its Natrium sodium-cooled fast reactor in March 2026 and broke ground in Wyoming the following month. Kairos Power broke ground on its Hermes-2 fluoride-salt-cooled demonstration in Tennessee on April 17, 2026, targeting operations by 2030. X-energy went public April 24, 2026, advancing a pebble-bed high-temperature gas reactor under a DOE cost-share for Dow Chemical's Seadrift industrial heat project.
The licensing landscape is fragmenting—some companies building under NRC permits, others testing under DOE authorization, a few attempting both simultaneously. The common thread is urgency, though not everyone is moving at Antares' pace.
Skepticism and Safety Questions
Not everyone celebrated the June milestone. The Union of Concerned Scientists had criticized DOE's rewrite of safety and security directives governing DOE-authorized reactor tests back in January 2026, alleging the agency had "slashed" detailed requirements in favor of performance-based standards that left too much discretion to operators.
Edwin Lyman, a physicist with UCS, told reporters in June that the Mark-0 achievement was "a first step" requiring far more testing before anyone could claim the design was safe for deployment. DOE and Antares maintained that the reactor met all applicable standards, but Lyman's point was that the applicable standards had been rewritten to move faster—perhaps faster than prudence allowed.
The criticism isn't entirely unfair. Nuclear safety culture developed over decades for good reason, much of it written in the aftermath of accidents that today's startup founders weren't alive to witness. Whether passive safety features and walk-away designs genuinely reduce risk or simply shift it in ways we haven't fully considered won't be known until these reactors operate at scale. And scale, for microreactors, remains aspirational.
The Fuel Bottleneck
Even if Antares transitions to commercial deployment on schedule, fuel supply remains a hard constraint. The company secured Urenco enrichment services and BWX fabrication for TRISO compacts, but HALEU availability is still gated by incremental DOE allocations and a domestic enrichment capacity that won't reach commercial scale for years. Centrus produced kilograms in its demonstration program; fleet deployment will require metric tons annually. Until U.S. enrichment capacity expands or international suppliers step in—complicated by sanctions, export controls, and geopolitics—advanced reactor timelines will collide with fuel bottlenecks no matter how fast the licensing process moves.
The global nuclear landscape is shifting in ways that make the U.S. timeline look both ambitious and late. Ontario Power Generation's BWRX-300 construction at Darlington is advancing toward a 2030 target, likely the first Western small modular reactor to reach operation this decade. GE Hitachi signed a February 2026 contract with Poland's Orlen Synthos to design two BWRX-300 units for deployment by 2035. France confirmed plans for six EPR2 reactors with final investment decisions expected before year-end 2026. The U.K. contracted with Rolls-Royce SMR on April 13, 2026, to deliver the nation's first small modular reactors, though timelines remain vague.
The U.S. goal articulated in the 2025 executive orders is to expand nuclear capacity from roughly 100 gigawatts today to 400 GW by 2050. Reaching that target would require not just new builds but license renewals for existing plants. California approved a 20-year extension for Diablo Canyon on April 2, 2026. Michigan's Palisades plant is undergoing a restart with DOE loan guarantees, targeting operations in early 2026 after regulatory reviews that stretched into spring, though that timeline has already slipped more than once.
Whether microreactors contribute meaningfully to the 400 GW target depends on economics, licensing speed, and whether the DOD and industrial customers provide enough early demand to drive down costs. Lazard's 2025 levelized cost analysis placed new nuclear at the high end of generation options, though the Nuclear Energy Institute's March 2026 brief argued that system-level reliability and decarbonization value justified the premium. Perhaps. Or perhaps the industry is still hoping policy support will paper over economics that don't yet pencil.
What Comes Next
Antares hit its deadline. The reactor went critical June 4, one month early. Two more DOE pilot reactors are supposed to reach criticality by July 4, 2026, if the executive order's targets hold. If they do, the administration will have demonstrated that federal urgency—paired with regulatory optionality and fuel allocations—can move advanced nuclear from PowerPoint slides to sustained fission in under thirteen months.
If they don't, the Antares milestone starts to look less like a turning point and more like an outlier. A well-funded startup with the right technology at the right moment, beating a deadline that may have been as much political theater as industrial policy. The difference between a proof of concept and a reproducible model.
The next twelve months will clarify which version of the story holds. Whether the supply chain can support more than one demonstration at a time. Whether DOD contracts materialize or get caught in acquisition bureaucracy. Whether Part 53 and Part 57 licensing actually deliver faster, cheaper approvals or just create new procedural labyrinths. Whether HALEU production ramps or remains a chokepoint. Whether other pilot program participants cross the finish line or stumble in the final stretch.
For now, Antares has a functioning reactor core in Idaho and a deadline it beat. That's more than most nuclear startups can claim. Whether it's enough to power AI's future—or even a single military base—depends on everything that comes after the champagne and the press releases. The hard part, as always, is what happens when the demonstration ends and deployment begins.
