Lock the payload at 2.60 tonnes. That’s the whole trick.
Not the engines, not the tanks, not the reactor, not the radiators – just the cargo. The crates and instruments and habitat parts that somebody actually wants at the far end of the trip. Everything that exists only to push that cargo gets counted somewhere else, on the transport side of the ledger.
This sounds like bookkeeping. It is, in the sense that the most important decisions in any comparison are the ones you make before the comparison starts. Loosen that one definition – let a spacecraft fold its own reactor into the “payload” column – and a heavy nuclear ship can claim a mass advantage it never earned. The comparison stops measuring propulsion and starts measuring who’s better at creative accounting.
I came into this expecting the answer to be obvious. Nuclear is more efficient than chemical, so nuclear wins, so we should build the nuclear thing and go to Mars. That’s roughly the shape of every confident take I’d read. The shape is right. It’s also too fast, and the place where it’s too fast turns out to be the entire story.
The equation that doesn’t take “please” for an answer
Before any of the engineering matters, there’s a law underneath it that doesn’t care about your budget or your schedule.
It’s the rocket equation. The version Tsiolkovsky wrote down more than a century ago, and the one that has quietly killed more spacecraft concepts than any other single fact. Here’s the part that bites: to change your velocity by some amount, you throw mass out the back. And the propellant you need doesn’t climb in a straight line as the mission gets harder. It climbs exponentially.
That word gets thrown around loosely, so let me be precise about what it costs. Each extra kilometer per second of speed you need costs more propellant than the last one did. Push the mission hard enough and the propellant stops being a part of the spacecraft and becomes the spacecraft, with the actual cargo riding along as a rounding error on top of a mountain of fuel.
The one lever against this is exhaust velocity – how fast you fling the stuff out the back. Engineers measure it as specific impulse, or Isp, but the plain-language version is all you need: the faster your exhaust, the less of it you have to carry. Higher Isp means a smaller fuel mountain for the same trip.
That single idea sorts the three contenders before anyone builds a thing.
Chemical rockets burn fuel and oxidizer. They hit hard and they’re proven, but the energy comes from molecular bonds rearranging, and chemistry sets a ceiling on how fast that exhaust can move. No clever engineering breaks through it. It’s a wall made of physics.
Nuclear thermal propulsion – NTP – walks around the wall. Run hydrogen through a fission reactor, heat it far past anything a combustion chamber could survive, and the exhaust comes out roughly twice as fast as the best chemical stage. Same idea as a chemical rocket, hotter source, double the efficiency.
Nuclear electric propulsion – NEP – takes the strange path. The reactor doesn’t heat the propellant at all. It makes electricity, and that electricity accelerates ions to exhaust velocities that make even NTP look sluggish. The least propellant wasted of anything on the list.
So the ranking is set, and it’s set by the physics, not by anyone’s preference. For a fixed payload on a fixed trip, NEP wastes the least propellant, NTP wastes less than chemical, chemical wastes the most. That order will not flip no matter what hardware anyone builds.
What the fuel mountain actually weighs
Now you can put numbers on it. And the numbers are where the obvious answer starts looking right – which is exactly when I should have gotten suspicious.
Run one consistent model – same assumptions about tanks and structure and reactor mass applied evenly to all three – and ask what it takes in low Earth orbit to send that fixed 2.60 tonnes toward Mars.
Mars Cargo Propulsion Comparison

Comparison of chemical, NTP, and NEP propulsion systems across mass, propellant fraction, and transit characteristics.
Chemical comes in around 15.4 tonnes. NTP around 10.6. NEP around 7.5. Both nuclear options beat chemical on launch mass, and NEP beats everything.
Look at the propellant fractions and the reason jumps out. Chemical is roughly three-quarters propellant by weight. The cargo is a passenger on a fuel truck. NTP cuts that to about half. NEP cuts it to about a quarter.
Here’s the thing I had backwards going in. I assumed nuclear wins because reactors are some kind of magic – compact, powerful, light. They are not light. Reactors and shielding and radiators are heavy and complicated. Nuclear wins for a completely different reason: propellant is brutally heavy, and the higher your Isp, the less of it you haul. The reactor isn’t the prize. The fuel you don’t have to carry is the prize.
I want to be careful about one number in that table, because it’s the one I’d trust least. Whether NEP actually beats NTP on mass is not a law – it’s a model result. Make the reactor shielding heavier, make the radiators bigger, make the power conversion hardware clumsier than assumed, and NEP’s edge over NTP shrinks. For a small payload it could vanish or flip outright.
So the durable claim – the one I’d stake my name on – is narrower than the table looks. Both nuclear options beat chemical. The exact gap between the two nuclear options is a guess dressed in a decimal point. Hold those two claims at different confidence levels, because they deserve different confidence levels.
These tonnages aren’t constants of the universe either. Change the tank technology, change the structural fractions, and the digits move. The pattern – nuclear under chemical – is far sturdier than any specific number. That’s the part worth carrying around in your head.
The bill the mass table never shows
If you stopped at launch mass, you’d build the NEP ship tomorrow. Lightest on the pad, end of story.
Stopping there was the frame I tried first, and it’s wrong, and being wrong about it taught me the actual lesson of this whole exercise.
Because efficiency and thrust are two different things, and the physics keeps them in separate accounts.
A chemical or NTP stage can dump a huge amount of power into its exhaust for a few minutes. One hard burn, and it’s off Earth and onto an interplanetary path. A clean, decisive shove.
NEP can’t do that. Its thrust is a whisper. Efficient as anything ever built, but feeble – the force at any instant is nowhere near enough to break Earth’s grip in a single push. So the NEP ship can’t leap away. It has to fire almost continuously for months, spiraling outward through one slightly higher orbit after another, slowly clawing its way up the gravity well before it’s even properly on its way to Mars.
This is called a low-thrust Earth-escape spiral, and the name makes it sound like a malfunction. It isn’t. It’s the unavoidable geometry of combining very high exhaust speed with very low thrust. You don’t get one without the other at these power levels. The mass savings are completely real. They just get paid for in a currency the launch-mass table doesn’t track – time.
The number I won’t pretend is real
To put a figure on the time penalty, this comparison runs the NEP case at 20 kilowatts of electric power. Twenty kWe.
Stop here, because this is the number that needs a warning label, and I’m going to give it that label three times before this article is done. The first is now: 20 kWe is an illustrative placeholder. It is not a confirmed NASA specification. It’s a working assumption I’m using to make the math run, nothing more.
With that caveat planted, here’s what the assumption produces. Chemical and NTP cargo stages reach Mars in roughly 7 to 9 months on conventional high-thrust trajectories. The 20 kWe NEP stage needs roughly 28 to 36 months to haul the same cargo to the same place.
Propulsion System Comparison
| Launch Mass (tonnes) | Transit Time (months) | |
| Chemical | 15.4 | 8 |
| NTP | 10.6 | 8 |
| NEP (20 kWe) | 7.5 | 32 |
Comparison of Chemical, NTP, and NEP propulsion systems based on required launch mass and estimated transit time to Mars.
Three years. Most of it spent not crossing to Mars but grinding up out of Earth orbit in that endless spiral.
Read that gap the right way. It is not a flaw in electric propulsion as a technology. It’s the price of running electric propulsion at low power. Pump the power up toward hundreds of kilowatts and the spiral tightens, the trip shortens, and NEP starts to look like a real Mars freighter instead of a tortoise. At 20 kWe, it’s a patient, frugal tug. Brilliant for cargo nobody’s waiting on. Useless for anything with a clock.
Mars Cargo Propulsion Trade-offs

A comparison of three propulsion architectures showing the trade-off between launch mass, propellant fraction, and transit duration.
And that 28-to-36-month figure doesn’t stand on its own. It hangs entirely off the 20 kWe assumption. That’s the second time I’m flagging it. The transit time inherits every bit of uncertainty baked into the power level underneath it. If the power number is soft – and it is – then the transit number is exactly as soft. You can’t quote one as a guess and the other as a fact. They rise and fall together.
This is the part of the discipline I’ve come to care about most, and it has a name worth keeping: uncertainty inheritance. Once a shaky assumption feeds a calculation, everything downstream is at least as shaky. The danger is that a placeholder hardens into a fact as it travels – the 20 kWe gets cited, then the 28 months gets cited, and three citations later nobody remembers the whole tower was built on an assumption somebody picked to make a model run.
Sorting the facts by how much they can betray you
Working through all this, I kept reaching for a habit that turned out to be the most useful thing I took away from the whole effort. I started tagging every claim by what kind of thing it is – and more to the point, by how it might let me down later.
Four kinds.
Physical law. The rocket equation. The fact that thrust and exhaust velocity are different quantities. These don’t change. They’re the bedrock. If physics shifts under you, the article is the least of your problems.
Engineering limit. The launch masses, the Isp values, the propellant fractions. Solid today, revisable tomorrow. Better engines, lighter tanks, and the numbers move. Trustworthy, but not eternal.
Program choice. The 2.60-tonne payload definition is one – somebody decided that, it isn’t handed down by nature. So is NASA’s current strategic leaning toward one architecture over another. Real, but a decision, and decisions get unmade.
Live claim. Mission names, dates, stop-work memos, hardware-reuse rumors. The stuff that’s true this week and gone next week. The material that needs a date stamp and a re-check, because it rots fastest.
Evidence-Label Tier Comparison
| Physical Law | Engineering Limit | Program Choice | Live Claim | |
| Robustness | 10 | 7 | 4 | 2 |
| Evergreen Status | 10 | 10 | 10 | 1 |
| Source Stability | 10 | 8 | 6 | 1 |
A comparison of the four evidence tiers based on their robustness, source requirements, and nature of the claim.
The whole point of sorting this way is that a reader – or a future version of me – can tell at a glance which claims to lean on and which to hold loosely. A physical law and a leaked mission codename should never carry the same weight in your head, and the trouble starts the moment they get mixed into the same confident sentence.
What NASA has actually said, versus what’s just going around
Here’s where I have to draw a hard line through the middle of the topic, and it’s the line that buys everything else its credibility.
There’s a confirmed version of NASA’s direction, and there’s a rumored version, and they do not get to share a paragraph.
The confirmed version, the part grounded in published strategy documents and budget submissions: advanced nuclear propulsion sits inside the agency’s Moon-to-Mars planning as an enabling technology. Within that, NEP gets pointed at uncrewed cargo and logistics – the slow, mass-efficient runs where nobody’s waiting on the dock. NTP stays in the mix for crewed missions, precisely because its high thrust keeps trip times short. And short trip times for a crew aren’t a convenience. Stretch a human mission to 28 months and you’ve turned a scheduling question into a radiation-dose question, which is a different and far more serious kind of problem.
Notice that this split – NEP for cargo, NTP for crew – isn’t a press release talking. It’s the physics from the first half of this article, showing up in the org chart. Save mass where time is cheap. Save time where lives are exposed. The strategy is the trade-off, wearing a suit.
Now the rumored version, and this is the third and final time I’m flagging the soft number. There’s a cluster of mission-level specifics that have circulated in reporting – a mission name, “SR-1 Freedom,” a payload called “SkyFall,” a reported April 2, 2025 stop-work date on the DRACO nuclear-thermal demonstrator, plans to reuse Gateway’s Power and Propulsion Element hardware. As far as I can confirm against primary sources, none of it is verified. It’s live-claim material. It might be true. It might be rebranded or canceled by the time you read this.
And here’s the connection I want to be honest about: that 20 kWe figure I’ve been leaning on traces back to this same cluster of unverified reporting. Which is the real reason it’s a placeholder and not a spec. I’m not using it because NASA confirmed it. I’m using it because it’s a reasonable working number, and I’m telling you plainly that its pedigree is rumor, not record.
That’s the discipline. The mission name and the power level get quarantined – walled off from the part of the argument meant to last. Not because rumors are worthless, but because an article built on them expires the day the next budget announcement lands. The physics doesn’t expire. The verdict shouldn’t either.
The objection I keep turning over
The strongest case against everything I’ve said isn’t that the numbers are wrong. It’s that the whole framing is rigged – and rigged by me, in the very first move.
Fix the payload at exactly 2.60 tonnes, hold everything constant, and of course the higher-Isp systems come out lighter. I built a track where nuclear was always going to win on mass. A fair fight, a critic could argue, but a fight I designed to have a known winner.
I sat with that, because it’s a good objection and it deserves more than a brush-off.
Here’s why I think the framing holds anyway. The fixed-payload rule isn’t there to help nuclear win. It’s there to stop nuclear from cheating. Without it, the genuinely dishonest comparison becomes possible – the one where a reactor counts as cargo and the mass numbers turn to mush. The constraint doesn’t tilt the table. It levels it, and it levels it against the side I’m ultimately favoring, by forcing nuclear to carry its reactor as overhead instead of hiding it as payload.
And the launch-mass win was never the real conclusion anyway. If I’d stopped at “nuclear is lighter, build it,” the critic would be right that I’d proven only what I set up to prove. But the actual finding is the trade, not the win. NEP is lightest and slowest. NTP is the one that takes the nuclear mass savings without surrendering the schedule. That result fights the simple “lighter is better” story rather than confirming it. A rigged track doesn’t usually produce a winner that complicates the case for racing in the first place.
The deeper objection survives, though, and I’ll name it straight: every engineering number here comes from one model. Different reactor masses, different radiator assumptions, different trajectory choices, and the digits shift. I’m not claiming the model is the territory. I’m claiming the pattern under the model – nuclear beats chemical on mass, low-power electric pays for that mass in years – is sturdier than any single figure inside it. That’s a more modest claim than the table makes it look. It’s also the one I’m actually willing to defend.
What I’ll say flat out, and what I won’t
So let me separate what I’m asserting from what I’m only entertaining, because the gap between those two is where honest writing lives.
What I’ll say flat out: for a fixed cargo to Mars, both nuclear options beat chemical on launch mass, and that’s driven by physics that won’t reverse. NTP keeps a fast 7-to-9-month trip. Low-power NEP is the lightest on the pad and the slowest by a wide margin, and that slowness is structural – it’s the spiral, not a tuning problem.
What I won’t say: that NEP definitively beats NTP on mass. That depends on reactor and radiator weights I can’t pin down. I won’t say the 28-to-36-month figure is a fact – it’s downstream of a placeholder power level borrowed from unverified reporting, and it’s exactly as soft as that placeholder. And I won’t repeat any of the mission names or program dates as though they’re confirmed, because they aren’t, and dressing a rumor in confident prose is how good arguments rot from the inside.
The verdict that’s left after all that subtraction is smaller than the one I walked in with. It’s also one I can stand behind without flinching.
Neither nuclear option is simply “the winner.” That was the answer I expected to find, and the work took it away from me. What decides between NTP and NEP isn’t a column of launch masses. It’s a question the spreadsheet can’t answer: how much time can this particular cargo afford to spend in flight?
Send the patient cargo – the supplies that just need to be there before the crew arrives, with no clock running – on the slow efficient tug, and let it spiral. Send anything with a deadline, and certainly anything with people, on the fast one, and pay the propellant.
The lightest ship and the right ship are not the same ship. That’s the whole thing, and it’s the part the confident takes leave out.
