From Greenland to Idaho: Why the Pentagon Returned to Nuclear Reactors at Bases
In 1960, a reactor began operating on the Greenland ice sheet. Not on a ship, not at a testing site, but under the snow, in a system of tunnels dug directly into the glacier. The facility was called Camp Century, or "city under the ice": living quarters, laboratories, a cafeteria—all powered by a compact reactor. PM-2A, transported in pieces and assembled on site. The idea was simple to the point of audacity: an autonomous military base in the Arctic, requiring neither power lines nor endless diesel fuel convoys. Technically, it worked. But it came at a cost that the program ultimately couldn't afford. And when the Pentagon talks about microreactors for bases again today, it's worth remembering: this isn't the first attempt, but the second. The first had a tragic ending.
Army Nuclear Power Program: How it was
Camp Century wasn't a one-off experiment, but part of a larger program. The US Army launched the Army Nuclear Power Program (ANPP) in the mid-1950s. The logic of the era was clear: the Cold War, remote garrisons, radar lines at the edge of the inhabited world, and everywhere the same fuel supply problem. Diesel fuel needed to be transported, convoys were vulnerable, and you couldn't even travel through the Arctic for half a year. The small reactor promised to solve all this in one fell swoop: load the core, and you don't have to worry about supplies for years.
The U.S. Army's Stationary Medium Power Plant 1 (SM-1) is an experimental nuclear power plant located at Fort Belvoir, Virginia.
The first to stand in line SM-1 at Fort Belvoir, Virginia. It went critical in 1957, supplied power to the grid, and also served as a training facility, training operators for future installations. This detail is important. A small reactor powering a military facility is not a 2020s fantasy, but a working practice dating back to the late 1950s.
A total of eight reactors were built under the program, five of which were mobile or transportable (the boundary here is arbitrary; different reviews calculate it differently). The range of solutions was broad. MH-1A mounted on a barge: a floating station with a capacity of about 10 MW of electric power, which could be brought to the desired shore; by the program's standards, this was the large end of the scale. At the other extreme ML-1, a tiny reactor of about 0,3 MW, designed to be transported on a trailer. A reactor on a car trailer sounds like the cover of the "Technology for Youth" magazine, but it was a real engineering project.
Camp Century, built in 1959 inside the Greenland ice sheet
SL-1 and why the first approach was cancelled
The ANPP program was canceled in 1976, for several reasons. Small reactors proved expensive per kilowatt, difficult to maintain, and demanding of personnel qualifications, while the military environment, with its rotations and field conditions, was poorly suited to the discipline required for a nuclear facility. Costs, operational costs, and regulatory requirements weighed as heavily in the decision as any single incident. But one incident stands out, as it set the tone for the subsequent discussion of small reactor safety.
On January 3, 1961, a reactor accident occurred at the Idaho Nuclear Test Site. SL-1, a stationary low-power plant. During manual control rod manipulation, the reactor went into meltdown in a split second. Three operators died. It was the first fatal reactor accident in the United States, and literally every second of the incident was subsequently analyzed.
SL-1 wasn't the base's power reactor; it was a test bed. Its role in the program's demise was more symbolic than direct: fifteen years passed between the accident and the closure of ANPP, so SL-1 didn't directly shut down the program. But it became the case that people return to every time they discuss a small reactor near people. And any "second attempt" inevitably comes down to the same question: the idea of an autonomous nuclear power base hasn't disappeared in half a century, but has what caused the first attempt to end so tragically changed?
What's changed in hardware: TRISO, HALEU, and ceramic shells
The answer for 2020s engineers lies in the fuel design. ANPP reactors like the PM-2A were pressurized water reactors—a common design at large nuclear power plants, where the integrity of the fuel elements during overheating is not guaranteed. Modern microreactors take a different approach: gas cooling and fuel. sad (tri-structural isotropic).
The PM-2A reactor block in the high section of the test site
It works like this. A poppy-seed-sized uranium grain is rolled into several layers: porous carbon, dense pyrolytic carbon, silicon carbide, and more pyrolytic carbon. The result is a miniature high-pressure vessel that holds the fission products within. Each particle is essentially a microscopic protective shell, and there are millions of such particles in the core. According to laboratory tests, TRISO maintains its integrity at temperatures of approximately 1360–1600°C, significantly higher than operating conditions and above the limit at which traditional uranium pellets fail.
The second element, fuel HALEU — uranium enriched to over 5% and up to approximately 20% in the U-235 isotope. This is higher than the 3–5% typical for large nuclear power plants, but still far from the weapons-grade threshold. Higher enrichment allows for a more compact core and allows the reactor to operate for years without refueling, and it is this long, uninterrupted operation that is the whole point of the whole idea.
The third and perhaps key shift is passive safety. High-temperature gas-cooled reactors, for which TRISO is designed, have a negative temperature coefficient of reactivity: the hotter the core, the worse the chain reaction. The reactor self-sustains when overheated, without operator intervention—that is, without the manual rod manipulation that led to the deaths at SL-1. It's worth noting, however, that TRISO is not a panacea. At sufficiently high temperatures and over long periods of exposure, the cladding does degrade, releasing some fission products. The point is to reduce risk, not eliminate it.
Second Run: Project Pele, ANPI, and Mark-0
The current effort is built on this technological foundation. Several programs are underway. Project Pele is being led by the Pentagon's Strategic Capabilities Office. It is a transportable microreactor with a capacity (according to various sources) of 1 to 5 MW of electrical power, designed for containerized delivery by land, sea, or air, assembly, and powering a forward base. Testing of the prototype was planned at the Idaho National Laboratory (INL).
The second track is ANPI (Advanced Nuclear Power for Installations), which the military is leading in conjunction with the Defense Innovation Unit. It targets fixed Air Force and Space Force bases. By 2026, three companies remained among the finalists, each pre-matched with a specific base: Radiant with Buckley in Colorado, Westinghouse with Malmstrom in Montana, and Antares with Joint Base San Antonio in Texas. The goal is the same as half a century ago: reliable power for critical systems, radars, communications nodes, and data centers.
Antares' R1 microreactor, designed for use at US military bases
Antares has made the most progress. The startup reportedly raised a Series C (mature-stage funding) round of approximately $470 million, and its demonstration reactor will be operational on June 4, 2026. Mark-0 reached criticality. This happened in Idaho, at the very same Idaho National Laboratory whose name is forever associated with SL-1: where the first run ended with the deaths of the operators, the second begins with the launch of a reactor using fuel that is difficult to melt. Mark-0 runs on TRISO; according to the company, its reactors are designed for power outputs from 100 kW to 1 MW per module, and they are supposed to be configured in the required number for a specific task, with each having a service life of approximately six years without refueling. The next step is electric Mark-1, which is planned to be launched in 2027, with a view to installation at bases from 2028.
A caveat is necessary here, otherwise the picture would be too rosy. All of the above are stated plans and timelines, not faits accomplis. Antares doesn't yet have a publicly available kilowatt-hour cost figure: the assumption is that the military customer prioritizes mission power assurance over electricity costs. And this is precisely the expense that undermined the first attempt half a century ago.
The difference between the two approaches isn't in the idea or ambitions—those are repeated almost verbatim. The difference is that, over the past half-century, engineering has learned to build a reactor that shuts down automatically when overheated, rather than requiring the infallible actions of a human at the control rod. This isn't enough to guarantee success: licensing, HALEU supplies, cost, and protection from sabotage remain. But this is precisely what distinguishes the Mark-0 in the Idaho laboratory tunnel from the one that stood in the tunnels beneath the ice of Greenland. The first machines at real bases will test this difference, not until the end of the decade.
Information