Fuel Costs of Point to Point Starship Travel

In the Starship Project Update presentation at Starbase, Elon Musk responded to a question about the prospect of using Starship for point-to-point long distance transportation on Earth by noting that the capital cost for Starship would be much less than for a conventional subsonic alrliner since the Starship could fly many more missions per day, while the airliner was limited to one or two due to its speed.

In comment #2 on that post, @CTLaw disputed Musk’s calculation, noting that turnaround time, even if the same for the airliner and rocket, would dramatically reduce the benefit claimed by Musk.

Major costs for airline operations include capital cost for aircraft, labour, fuel, maintenance, and fixed costs. I decided to look at one of these, fuel cost. This turns out to be far more difficult than you might expect. Elon Musk said this about Starship fuel costs on 2020-05-08:

I tried to confirm this, and fell down a rabbit hole which consumed several hours of frustrating inquiry, from which I never completely emerged. You’d think it would be easy, knowing the mass fraction of oxygen and methane, which works out to 936/264 tonnes for Starship and 2652/748 tonnes for Super Heavy, or 3588 tonnes LOX and 1012 tonnes liquid methane for the full stack, to estimate fuel cost, but you’d be wrong.

There are many things the Internet knows, but the price of liquid oxygen in bulk doesn’t seem to be one of them, which is kind of odd for an industrial product made and consumed worldwide in a quantity estimated at 16 million tonnes per year. About the best I can find is “around US$ 0.20 per kilogram”, which is US$ 200 per tonne. Multiplying this by the total LOX load of Starship plus Super Heavy yields US$717,600, which is already more than Elon’s estimate for all propellants. Maybe the US$ 200 per tonne is the delivered price for a commercial supplier including their mark-up and SpaceX’s estimate assumes they’re building their own LOX plant near the launch site and producing LOX for cost near that of the electricity to run it (which also varies dramatically from one region to another). Anyway, that’s a substantial discrepancy.

Now we move on to the methane fuel. Forget looking for prices for liquid methane—you’ll find nothing of use. But liquefied natural gas (LNG) is basically methane with some impurities, so I decided to use it as a proxy. LNG is a commodity which is traded worldwide and quoted in real time, so it’s easy to get numbers, but they are in the screwball units of US$ per million British Thermal Units (MBTU), where the British Thermal Unit is a quaint unit of energy defined as the heat input required to raise one pound of water by one degree Fahrenheit, which works out to somewhere between 1054 and 1060 joules in civilised units, depending upon which variety of quaint definition of BTU you use. To convert this to a mass of liquid natural gas, we must choose an estimate of the energy produced by burning a given mass of gas under specified conditions, then take the inverse. Using conventional definitions you end up with 51.9 MBTU per tonne of LNG, which lets you convert the quoted price in MBTU to LNG price per tonne.

But LNG prices are highly volatile and have been all over the map in the last ten years, as fickle policy changes, energy production trends, and production technologies have evolved. In April 2020, LNG in the U.S. sold for US$ 2.85/MBTU, while in October 2021 it was US$ 12.15/MBTU. At this writing, it’s around US$ 4.30/MBTU. Whatever estimate you use is likely to be far off the mark a few months later.

Anyway, I picked US$ 5/MBTU as an eyeball average price over the last few years and assumed that in large enough quantities the cost of refining LNG into sufficiently pure methane for Raptors to inhale is small compared to the cost of the feed stock. Running this through all of the conversions and multiplying by 1012 tonnes of methane in both stages gives a cost of US$ 262,614 for methane, for a total propellant cost of US$ 980,214 for the full vehicle. This is around twice Elon Musk’s estimate, but close enough for a Fermi calculation and highly dependent upon the assumptions made for liquid oxygen production.

The next question is whether a full propellant load would be required for point-to-point transportation on Earth. I assumed it would. For a flight on the order of half way around the world, for example Los Angeles to Singapore, around 14,000 km, the difference in velocity between an orbital trajectory outside the atmosphere and one which re-enters for a landing at such a range is insignificant compared to the other uncertainties in this calculation. Further, while reducing payload would decrease the fuel requirement, I assume the full payload capacity of 100 tonnes, since if the ship were used as a freighter, that would provide the most economical operation and it is conveniently close to the payload capacity of long-haul freighter aircraft. While some have spoken of Starship being used without Super Heavy for point to point service, this makes no sense for long range flights, as it cannot remotely reach sufficient velocity to achieve intercontinental range while carrying a useful payload.

Now, let’s look at a conventional air freighter. I’ll use the Boeing 777F, which has a maximum payload of 103.7 tonnes, essentially the same as Starship, fuel capacity of 145.5 tonnes, and a range of 9200 km at its maximum payload, but up to 18,057 km with reduced payload. The 777F burns Jet A fuel, whose cost is also volatile (although not to the degree of LNG), and is currently quoted as US$ 882.3 per tonne. Thus the fuel cost for the 777 would be US$ 128,408 for the flight. (The full payload might require a refueling stop, but given the liquid global market for Jet-A, this would not make a substantial difference in total fuel cost.)

So, the bottom line is that the Starship rocket freight mission would cost US$ 980,214 (or around a million bucks) by my estimate, or half that by Elon Musk’s, while the Boeing 777 would burn US$ 128,408 in fuel for the same trip. This makes Starship between 3.9 and 7.6 times as expensive in fuel costs as the subsonic airliner. The end-to-end flight time for the Starship is estimated as around 45 minutes, while United Airlines Flight 37 from Los Angeles to Singapore (on a Boeing 787-9) takes 17 hours and 50 minutes, or 23.8 times as long, which will increase the capital and crew costs for the airliner, but these are at this time so uncertain for Starship as to be impossible to estimate for comparison.

My take away from this is that from the standpoint of fuel cost alone, the viability of point to point suborbital transport cannot be ruled out. Fuel may cost between four and eight times as much, but the freight or passengers arrive almost 24 times faster, and given airline-type turnaround times, a rocket transport could certainly complete more revenue flights per day than the airliner.

Will there be a sufficient market to develop and sustain such a service? That’s hard to say. The first generation of passenger jet transports (Boeing 707, Douglas DC-8, etc.) were fuel-thirsty and expensive compared to the piston engine planes they replaced, but they flew almost twice as fast and before long almost completely displaced them on long haul routes. It could be that for passenger service, rocket flight is just too fast—given time zone differences, flight halfway around the world will necessarily either depart or arrive at an inconvenient time on one end and passengers will probably be half-awake zombies for a day or two after arrival, negating some of the advantage of getting there so quickly. It’s easy to see the advantage for time-critical freight, but not clear if there’s enough to fill up a 100 tonne freighter for frequent trips to the antipodes.

Some have argued that passengers won’t accept the inconveniences of acceleration and weightlessness that suborbital flight entail: “Half of the time you can’t get to the toilet and the other half it doesn’t work”. Hey, the flight is only 45 minutes instead of 16+ hours and, as to the toilet, have these critics ever flown on EasyJet?


Also, some future carbon tax on fuel will add to the difference.

1 Like

If it is a rational carbon tax, it should close the gap between LOX/methane and air/Jet A because combustion of Jet A, which contains lots of long-chain hydrocarbons, produces substantially more CO₂ than burning methane, which has only one carbon for four hydrogen atoms. There is no hydrocarbon fuel with less hydrogen than methane. This is why natural gas produces so much less CO₂ when burned for electrical generation than coal or petroleum.

Because hydrogen weighs so much less than carbon, the CO2 per kilo of methane would be only slightly less than the CO2 per kilo of jet fuel. A rounding error when the difference in fuel loads is greater than a factor of five.


Musk’s Starship promises to be more revolutionary than Fulton’s Steamboat.

Fulton’s Steamboat voayage in 1807 on the Hudson River from New York City to Albany was a commercial success (i.e. profitable with flesh and blood customers), and a bit quicker and more comfortable than riding horseback!

My prediction is SpaceX Starship passenger voyages will be profitable – commercial success via BOTH government contract – and private PASSENGER travel.

If SpaceX tech goes well, would multi-national corporation airlines or aviation industry put up barriers to try and stop it?

Will airlines buy SpaceX Starships, instead of Boeing 787 Dreamliners?

Has 787 or Starliner signaled Boeing’s peak in the aviation industry?

How long before China steals the tech if they haven’t already?

We truly live in fascinating times.

1 Like

It sounded like Musk’s point was more on capital efficiency of point-to-point suborbital flights than on fuel costs. Of course, “capital efficiency” is a quaint concept in a world of negative real (and, in Europe, negative nominal) interest rates.

Obviously, a sub-orbital transport would be designed very differently from a Mars-bound Starship. Multiple G acceleration would be a non-starter. The bigger question is – What kind of passenger or cargo would require such rapid transport? The only one I can think of would be military – the capacity to deliver soldiers & military supplies very rapidly to crisis situations. But that would impose a whole lot of additional requirements even further from a Starship.

The big surprise for me in Musk’s presentation was the comment about designing the launch/landing system for 1-hour turnround of the booster after its 6-minute flight. That is quite an aspiration!

In the initial presentation that mentioned point to point service (and included the video in the original post), it was said that ticket prices were expected to be no more than present-day business class once the service matured. If this is the case, it may be compelling to airlines, who might pay a high price for missing out. Airlines make most of their money on business class and use economy to fill up the plane and extract a little more money from the asset. Today, they can only offer some convenience and creature comforts for the large price differential. But if a business class ticket meant you got there in less than an hour rather than 12 to 16 hours, that’s something entirely different. To professionals who bill hundreds of dollars an hour, being able to make a day trip to see client across the Atlantic is worth a premium.

This is why United Airlines has signed an (easy to get out of) contract with Boom, the supersonic airliner start-up. The CEO of United has explicitly said that if he can offer his current business and (few) first class customers getting there twice as fast, that’s something that’s a lot easier to sell than wider seats and better food.

Give the fuel price difference (which, based in physics, there isn’t much you can do about), suborbital express will probably always remain a premium service, with economy remaining in wide-body jumbos for the foreseeable future and probably moving further down-market from even its present state.

What intrigues me, and may be the subject of a follow-up post at some time, is the prospect that Starship may be a much less expensive technology to produce in quantity than airliners. A present-day jet engine is, by itself, fantastically complicated, and airliners are chock full of hydraulics, actuators, moving parts, fuel management systems, control surface actuators, etc. etc. Raptor engines are demanding to manufacture but in terms of moving parts vastly simpler than fanjets, and they don’t have to worry about icing conditions or sucking in birds. Yes, there may be 42 on the whole ship, but that means you just get to slide down the learning curve faster. The rest of the ship is largely propellant tanks which are easily fabricated. There is no complicated landing gear to worry about. So, once they’re making these things in the hundreds or thousands, they may be a lot cheaper than airliners, which would throw another monkey wrench into the conventional wisdom predictions.


That is an interesting point - brings us back to capital efficiency. I recall long ago reading an article about the only rocket which has ever been produced in large numbers – the WWII German V2 rocket. Something like 3,000+ were manufactured under difficult wartime conditions, largely by unreliable low-trained impressed labor – and most of them worked!

In other talks, Musk has made a point of talking about manufacturing at a very large scale. At a large enough scale, the cost per item asymptotically approaches the cost of materials plus IP – if I recall his statement correctly.

The Mars-bound Starship program to move a million tons of supplies to another planet clearly would require vast numbers of rockets. Whether an upgrade on First Class/Business Class air travel would ever generate the same demand for rockets is another question. Supersonic Concorde was not an economic success – at least in part because of the relatively small market for high-priced faster travel.

1 Like

As it happens, I think that may have been an article I wrote in 1993, “A Rocket a Day Keeps the High Costs Away”. I have recently been told that a number of the people now in the smallsat business say that article contributed to their thinking along those lines.

Supersonic Concorde was not an economic success – at least in part because of the relatively small market for high-priced faster travel.

Concorde had a lot of problems. Its range meant it couldn’t fly many potentially profitable routes. It only carried 100 passengers and the very expensive and costly to operate and maintain plane meant ticket prices had to be 40% higher than first class, which was out of reach except for plutocrats, celebrities, or those who wanted a once-in-a-lifetime treat (including me—I flew Washington to London once in 1991), the prohibition on flight over land made it non-viable for all potential U.S. airline customers for transcontinental flights, and it was designed for cheap fuel and was un-affordable after the first oil shock which happened before it went into revenue service.

If a supersonic transport had been able to deliver Mach 2 service at business class prices, it might have been viable. We’ll see—that’s what United is betting on with Boom (well, Mach 1.7, which is still twice as fast as as steerage).


No, that 1993 article is not the one I remember reading. Your 1993 article was much more complete and thoughtful – and wittier into the bargain! I suspect whoever wrote the article I remember had read yours.

That was a highly innovative proposal for a contract to jump-start the space industry, which as you point out was well within the means of any developed nation (and several less developed nations). The fact that none of the leaders of any of those nations ever acted on your proposal makes one even more grateful for Elon Musk – and for the guys who pushed him out of Paypal with money he was prepared to risk.

As an aside, my attention was caught by the 1993-era assumption: “policy makers worry about how to convert defence industries without harming readiness by eroding the industrial base.” Those were the days! When Western policy makers did not think that “Industrial Base” was a city in China.

1 Like