A fast nuclear thermal propelled spacecraft enters lunar orbit.

NASA

The White House memorandum NSTM-3, issued April 14, 2026, establishes the most comprehensive federal commitment to space nuclear power in decades. It directs NASA and the Department of War (DOW) to pursue parallel reactor competitions for fission surface power and nuclear electric propulsion, sets aggressive timelines (FSP ready for a launch to the Moon by 2030, an in-space reactor for DOW use by 2031), mandates firm-fixed-price contracting with milestone payments, and leverages the Energy Department’s national laboratory infrastructure for fueling, testing, and safety analysis. Combined with NASA’s recently announced plan to send the nuclear electric-propelled SR-1 Freedom mission to Mars by December 2028, this guidance - which follows on the heels of the Trump Administration’s Executive Order ("Ensuring American Space Superiority," December 2025) - represents a genuine inflection point for space nuclear capability.

It also contains a significant strategic omission.

The thing that’s missing is nuclear thermal propulsion (NTP). Unlike electric propulsion, which features high efficiency but very low thrust levels (thus very long journeys to destinations such as the Moon or Mars), NTP is high-thrust and high efficiency - or more accurately, high "specific impulse," in rocket terms. That means short trips (perhaps as short as a single day to the Moon or 60-90 days to Mars). This is valuable not just for NASA - think crewed missions which don’t make their astronauts suffer extended radiation exposure, as well as bone and muscle loss on long trips - but also for DOW and the Space Force, which are scoping out ways to deter our adversaries’ menacing actions on orbit. That means rapid response, and nuclear thermal propulsion is a way to get that, cheaply.

Despite this, NSTM-3 directs that NTP funding be focused “to the maximal extent possible” on just about everything but a nuclear thermal engine: shielding, communications, reactor control, and radiation-hardened instrumentation, whatever is “common” to NTP and nuclear electric propulsion. The memo frames NTP as a future option for development, but provides no programmatic pathway to build the NTP engine itself. DARPA’s joint effort with NASA, the Demonstration Rocket for Agile Cislunar Operations (DRACO), was cancelled last year. NASA’s NTP project page is archived. The practical effect is that the single nuclear propulsion technology with flight-relevant heritage (dating all the way back to NASA’s ROVER/NERVA program, which ended in the early 1970s), the Strategic Defense Initiative’s Timberwind program (1980s-90s), the half billion dollars’ investment in a DRACO flight (planned for 2027 or 2028), and active congressional appropriations is being reduced to a technology reservoir for a different propulsion architecture.

This is a mistake, and the reasoning behind it does not survive scrutiny.

The “cheap launch” fallacy
DARPA Deputy Director Rob McHenry’s rationale for canceling DRACO was straightforward: falling launch costs, driven by the promise of an operational SpaceX Starship, have eroded the return on investment for NTP’s efficiency advantage. If you can launch propellant cheaply enough, the argument goes, chemical propulsion plus (inexpensive) bulk fuel is good enough to get you anywhere you need to go.

But does that logic play out? Starship’s own lunar architecture illustrates a problem it claims to have solved. To land a single Starship on the lunar surface, SpaceX must first launch a “depot” Starship to LEO, then launch somewhere between 10 and 16 additional tanker flights to fill it, each requiring an autonomous rendezvous and cryogenic propellant transfer in orbit - and only then does the lunar-capable Starship tank up from the depot and head off to the Moon. Doable? Yes. Risky? Absolutely. Every one of those rendezvous operations is not simply a schedule risk, it’s a potential opportunity for mission failure. Sure, the architecture closes on paper, but it requires a substantial campaign of missions just to enable one lunar landing. That’s not cheap launch making advanced propulsion irrelevant. It’s cheap launch being overworked by the brute-force chemical propellant demands of a low-performance architecture.

We can do better than that. An NTP stage with roughly twice the fuel efficiency (the aforementioned “specific impulse”) of chemical propulsion doesn’t need any of those extra flights. It picks up its payload and propellant in LEO (say, from Starship or Blue Origin’s New Glenn rocket), delivers it to lunar orbit, and returns for the next load. The handoffs are singular and straightforward. The logistics chain is short. The failure modes are fewer. The cheap launch that Starship or its competitors provide is not wasted on hauling propellant to feed itself; they haul payload, which is the point.

This is where DARPA’s argument inverts. The agency framed NTP’s efficiency as a diminishing asset in a world of cheap launch. The opposite is true. Cheap launch makes a reusable NTP tug more valuable, not less, because the tug’s economics improve as its traffic volume increases. A high-reuse orbital transfer vehicle operating at two or three times the specific impulse of a chemically propelled rocket can amortize its capital cost across dozens of missions. As flight rate climbs and refueling costs drop (thanks to the very same cheap launch McHenry cited), the cost per kilogram for the middle leg converges toward LEO access costs: a few thousand dollars per kilogram rather than the tens (to hundreds) of thousands that characterize today’s chemical upper stages when they’re employed to go anywhere but low Earth orbit. That's not a marginal improvement. That's a transformation of cislunar economics, enabled precisely by pairing cheap launch with efficient in-space propulsion.

The relevant question was never "is NTP worth it if launch isn’t expensive?" It was always "what’s the right mix of delivery systems to get things anywhere we want to go?" Starship is a near-optimal answer to getting yourself to LEO in the first place. NTP is a wonderful answer for the middle/final leg, getting things from LEO to other orbits. But no amount of cheap airlift to a forward operating base changes the fact that you still need to get from the base to the objective. That’s the middle/last leg, and pretending cheaper first-mile costs eliminate the need for it is wishful logistics.

All this being said, the most important advantage is not cost; it is time. You can’t respond to a cislunar contingency with orchestrated sequences of launches or a slow, low-thrust electrically-propelled spacecraft. Sometimes you have to go fast.

Defense needs responsive maneuver
NSTM-3 directs DOW to pursue a mission-enabling “mid-power” (20 kilowatts or more) in-space reactor by 2031, including an analysis of operationally relevant use cases for space nuclear systems. The memo wisely avoids over-specifying the mission, but its exclusive focus on reactor power (not propulsive thrust) implicitly favors NEP for the propulsion architecture.

For national security space operations, this creates a clear capability gap. DOW’s emerging cislunar mission set includes space domain awareness, inspection, logistics resupply, and responsive repositioning of assets - and in the future, defensive and offensive operations ("maneuver warfare"). These missions demand transit timelines between key orbital regimes measured in hours to days, not months or years. A satellite that can reposition from GEO to, say, one of the Earth-Moon Lagrange points - and do it in just a day or two - using NTP has operational relevance. One that takes months or longer to spiral between orbits using NEP simply doesn’t.

The memo’s own language calls for an analysis of “operationally-relevant use-cases.” If that analysis is honest, it will have little choice but to conclude that many defense missions require thrust levels that NEP cannot provide. By the time that conclusion arrives (90 days from the memo’s issuance), the NTP industrial base will have lost another quarter 's worth of momentum. Engineers will migrate to other programs. Test infrastructure planning will stall. The analysis will recommend a capability whose development pipeline has been disrupted by the very policy that commissioned the analysis.

Crew safety demands fast transit
The memo references NTP’s future role for “crewed missions to Mars." This is worth examining carefully.

Galactic cosmic ray exposure is the binding constraint on Mars crew transit. NASA’s permissible exposure limit translates to a practical ceiling on total mission duration. Chemical propulsion yields Mars transit times of 7-9 months each way, when planetary alignment is favorable (roughly every two years). An NTP-enabled vehicle can reduce outbound transit to as little as 60-90 days using higher energy trajectories that are effectively out of reach to chemical propulsion. NEP, despite its high specific impulse, cannot reduce transit time for a crewed vehicle because its low thrust-to-weight ratio prevents it from performing the impulsive burns needed for fast transfer orbits. An NEP vehicle carrying crew would have to spiral slowly out of Earth’s gravity well, traverse the Van Allen belts at leisure (adding to radiation dose, not reducing it), and then spend months in heliocentric cruise.

The memo’s position that NTP development can be deferred while NEP components are matured first is internally inconsistent. If crewed Mars missions are an end goal (as the memo states), then an NTP engine sits right in the critical path. Shielding, control systems, and rad-hardened instrumentation are of course necessary but are hardly the long pole in the technical tent. The focus should be squarely on building a near-term reactor with already-available fuel that offers propulsion-grade thermal output (~2,750 K hydrogen exit temperature or better), cryogenic propellant storage, and vehicle integration. None of these things are advanced by building a 20-kWe Brayton-cycle electric power system.

The right framework is intermodal, not unimodal
The strongest version of a national space nuclear strategy recognizes that different propulsion modalities are optimized for different mission segments, just as terrestrial transportation uses trucks, trains, ships, and aircraft for different legs of a supply chain.

Chemical propulsion handles the first mile: Earth surface to LEO. It is mature, high-thrust, and increasingly cheap and reliable.

NTP handles the middle miles: LEO to GEO, LEO to cislunar space, LEO to Mars transfer orbit. It provides the thrust needed for reasonable transit times and the specific impulse needed to tamp down on propellant mass requirements. It’s the only nuclear technology that can close both requirements simultaneously for time-critical missions.

NEP handles the long haul: efficient cargo pre-positioning to Mars, outer solar system exploration, station-keeping for high-value assets where time is not the constraint but total delta-V is. SR-1 Freedom promises to be an excellent demonstrator for this mission class.

FSP provides the destination infrastructure: continuous (high-density) power for lunar and planetary surface operations independent of sunlight.

NSTM-3 funds three of these four capabilities. The missing leg (NTP) could use some help. Congress has recognized this gap for years, appropriating $110-175 million annually for NTP development even when NASA did not request it. The FY2026 enacted appropriations (P.L. 119-74) continued this pattern. The White House should work with this congressional consensus rather than redirecting the funds toward an architecture that, however valuable, does not address the same operational need.

Recommendations
First, preserve NTP engine development as a distinct program element within the overall initiative. The common components identified in OSTP’s guidance (shielding, control systems, instrumentation) are genuinely shared between NTP and NEP, and funding their co-development is sensible. But the NTP-unique elements (reactors and the components that turn them into rocket engines) must also be funded, or the “path toward NTP” referenced in the memo leads nowhere.

Second, include NTP in DOW’s 90-day use-case analysis as a propulsion option for responsive cislunar missions, not just reactor power for payloads. The analysis should compare NTP and NEP maneuver timelines for operationally relevant scenarios and let the physics speak for itself.

Third, direct NASA to accelerate and focus the NTP fuel element and reactor development efforts that Congress has supported for many years, separate from the NEP demonstrator. Incremental, inexpensive test flights - perhaps of something short of a fully-fueled reactor - will allow us to exercise the space nuclear regulatory regime and identify friction points for the interagency to address. These “NTP-supporting efforts” are exactly the long-lead items that will determine whether NTP is available in the 2030s, when they’ll be needed. Stopping and restarting this work adds years and costs more than sustaining it.

Fourth, preserve the NTP industrial base. National labs and private companies developing advanced fuel elements and engine components represent a fragile ecosystem. Policy signals matter. Engineers (and irreplaceable skill sets) depart when programs face uncertainty or cancellation. Reversing that attrition is much harder than preventing it.

NSTM-3 is the most significant space nuclear policy document since SPD-6 (a policy prescription authored by the first Trump Administration’s National Space Council, intended to streamline space nuclear regulatory requirements and get reactors on orbit). Its framework for FSP and NEP is sound, its contracting approach is disciplined, and its timelines are ambitious but achievable. Extending that same discipline and ambition to NTP would give the nation a complete space nuclear capability, not two-thirds of one.

Disclosure: The author is CEO of Dark Fission Space Systems, which is developing nuclear thermal propulsion technology under a U.S. Space Force SBIR contract.