Starship is meant to imitate (to a degree) the affordability of passenger jets. This applies to Earth-to-Earth travel as well as space travel.

Jetliners started with 4 engines, but have since migrated to using 2, regardless of size. Where the number of engines might have been increased for an earlier generation of wide-body aircraft, now, as with the 777X, companies simply make larger engines for every new design. This twinjet configuration is chosen for saving costs, as each engine requires separate service, paperwork, and certificates.

Why does the same not apply to Starship?

  • 6
    $\begingroup$ I suspect you have answered your own question here - available tech forces us closer to the 'six turning four burning' of the B-36 en.wikipedia.org/wiki/Convair_B-36_Peacemaker than 777X $\endgroup$ Commented Aug 20, 2021 at 13:41
  • 7
    $\begingroup$ The simple answer is that rocket flight has extremely little in common with airflight. For one, the amortized cost of rocket per flight vs fuel costs are not at all comparable to the amortized cost of a 777 vs fuel (and it's much less analogous when looking at F9). But also the flight profiles and stresses introduced are entirely different as well. $\endgroup$
    – eps
    Commented Aug 20, 2021 at 22:59
  • 9
    $\begingroup$ Simiarly, a F1 racecar has a much different engine and shape and has little in common to a highly efficient 4 cylinder honda. $\endgroup$
    – eps
    Commented Aug 20, 2021 at 23:11
  • 4
    $\begingroup$ Aircraft companies don't really make significantly larger engines for each new generation, they generally build smaller airplanes. For instance, the 777X you reference is considerably smaller than the 4-engined 747, while the larger than 747 Airbus A380 uses 4 engines. $\endgroup$
    – jamesqf
    Commented Aug 21, 2021 at 0:12
  • 2
    $\begingroup$ Scaling up rocket engines doesn't work very well. The bigger the engine the more instability issues you have. An unstable burn tends to be extremely dramatic. $\endgroup$ Commented Aug 22, 2021 at 1:29

6 Answers 6


Economy of scale, strictly. SpaceX focuses heavily on streamlining and automation of production of these engines. High up-front cost, but low unit cost per engine once the process is perfected.

The larger the engines, the higher the up-front cost would be, as problems of combustion stability, cooling, material durability and so on crop up; this is a well-known problem of massive engines, one that buried Soviets' plan of manned moon landing too.

Then there is the problem of production errors - e.g. faulty 3D prints, problems that become apparent in tests after print completion. Let's say the 3D printer has 1 in 10 chance it will glitch once on given day, producing a fault that ruins the currently printed engine. If the engine is smaller, and takes 1 day to complete the print, one in 10 engines will be faulty, 10% work time of the printer wasted, 10% of production lost. Now increase the engine size, so it takes 2 days to print. Same chance of a glitch ruining it, but now two days are wasted, one engine in five is a reject, and twice as much of materials is wasted with each rejection. That automatically means per-unit cost of the engines is increased, as the loss due to the faulty one is spread between four good ones, instead of nine.

Then there's the matter of redundancy. If you have, say, 15 engines, 2 or 3 flaming out won't mean loss of mission. If you have two and lose one, you won't be going to space today.

The aviation industry is way more mature now - in the beginning they did use many smaller engines because scaling that much up simply wasn't technologically viable - and also for reasons similar as with SpaceX currently, economy of scale, reliability, redundancy. With constant, steady stream of revenue from existing production of smaller engines, innovation, improvements in safety and reliability, and active competition, they were able to develop incrementally larger and more powerful engines without running at "infant mortality" problems of new production, where before it becomes profitable, there are a lot of problems to solve and costs to bear.

  • $\begingroup$ The chance of a glitch ruining the engine is higher for the larger engine, 20 % instead of 10 % for the smaller one. $\endgroup$
    – Uwe
    Commented Aug 20, 2021 at 20:29
  • 1
    $\begingroup$ @Uwe Generally speaking, if you increase your clearances linearly with size, you are not using scale to your advantage. If you'd increase the gap of an oil lubricated gliding bearing, you'd need to pump much more oil into it and do so at a higher pressure to keep the parts from touching. The oil film does not care how big the parts are, as long as there is a thin film of oil all around. As such, it would be much more prudent to increase the number of oil channels feeding the bearing, and keep the gap size the same if possible. Size changes due to temperature may force bigger gaps, though. $\endgroup$ Commented Aug 20, 2021 at 22:01
  • $\begingroup$ I think with more trustable drives the larger ones would be not only more economic, but also more effective. A drive capable to create double thrust would likely weight lesser than twice, and also producing it would cost lesser than twice. $\endgroup$
    – peterh
    Commented Aug 21, 2021 at 11:24
  • 3
    $\begingroup$ @peterh "Better is the enemy of good". First SpaceX needs to get a working and revenue-generating Starship, then they can work on optimization. $\endgroup$
    – SF.
    Commented Aug 21, 2021 at 12:07
  • 2
    $\begingroup$ Also, "thousands of 2MN vs hundreds of 20MN" - "rocket engines do not necessarily scale in a simple way. " - once again, yes, it doesn't - in particular with problems of combustion stability cropping up exponentially, requiring extremely sophisticated injector systems, possibly multi-chamber construction, new problems related to thermal expansion, and all sorts of problems absent in smaller engines. As the size goes up, up to a to certain point \$/N is dropping gradually, but then it ramps up. Raptor is around the sweet spot. $\endgroup$
    – SF.
    Commented Aug 23, 2021 at 12:05

Because with current technology, the greatest part of expense in building a rocket motor is not the individual construction, but the research needed in the design of it. And it is simpler, easier and cheaper to design a rocket engine of moderate size, than a colossal monster of an engine (like the F1 that Saturn V used)

Even with airliners, the HUGE turbofan engines are not selected because they are cheap to manufacture, the very opposite is true. A single General Electric GE9X as used by the Boeing 777 costs 44.5 million dollars. Each! Whereas each engine on a 747 only costs about 13 million, yet produces 60% as much thrust each.
The huge engines on a 777 are selected because they offer better fuel economy than multiple smaller engines, and slightly less maintenance cost.

Unfortunately, rocket engines are nowhere near the maturity of development that airliner turbofan engines have.

  • 3
    $\begingroup$ I wonder how that relates to Musk's claims, that design is easy, production is difficult. $\endgroup$
    – SF.
    Commented Aug 20, 2021 at 17:25
  • 11
    $\begingroup$ @SF. mass production is difficult, producing an individual engine isn't so hard. The individual engines are actually pretty simple and easy to build in comparison to jet engines. SpaceX's Raptor is estimated to have oxygen preburner temperatures of ~800 K, compared to ~2300 K for a turbojet's combustor. Pressures are high and fluids are dense, so the turbines and pumps are much smaller, and there's a huge flow of propellant for cooling. You see some carefully chosen metallurgy, but not the single-crystal superalloy turbine blades with integrated cooling channels you see in jet engines. $\endgroup$ Commented Aug 20, 2021 at 18:25
  • 2
    $\begingroup$ I was specifically referring to turbojets as being more similar to a rocket engine's turbopumps (particularly in a full-flow engine like Raptor) and to avoid confusion with the large but low-temperature fans of turbofans. $\endgroup$ Commented Aug 21, 2021 at 13:06
  • 3
    $\begingroup$ Turbojet engines also operate at Thrust-to-weight ratios that are way, way WAY WAY less than those for rocket engines. A GE90 engine has a TWR of 6, a Merlin 1Dvac has a TWR that is 30 times better. (mostly because the turbojet's focus is not TWR, but Fuel economy, while the rocket is all about Thrust and ISP) $\endgroup$ Commented Aug 21, 2021 at 17:37
  • 2
    $\begingroup$ Musk and OP are using "design" in two different ways. Musk means actual design of the actual rocket. OP is including manufacturing design in the term "design". Production is difficult--that's why it's better to spread that cost of figuring out how to actually build the engine over 500 engines instead of 50. Why spent all that work designing tooling just to only fart out a handful of engines? $\endgroup$ Commented Aug 22, 2021 at 1:05

It's not clear to me from your question if you're asking about Starship, or about Super Heavy.

Starship is the upper stage of the Starship/Super Heavy launcher, and has relatively moderate 6 engines on it; 3 "sea level" Raptors that can gimbal to point their thrust plus three more optimized for performance in vacuum with large fixed nozzles. One working sea level engine out of the set of three is required for landing on Earth; failure to light at least one engine guarantees destruction of the vehicle. The vacuum engines are not usable here.

Unlike an airliner, the engines are turned off during most of the descent and only start a few seconds before landing, which adds a large risk factor (as the development flights of Starship have shown painfully). Furthermore, an airliner even has a good chance to land with zero working engines. So the landing modes just aren't directly comparable, and the actual 3:1 redundancy isn't excessive.

Super Heavy, on the other hand, mounts 29 31 engines. Here, the driver is engine commonality with the upper stage and economy of scale. SpaceX only has to develop one really good methane engine, and then concentrate on how to reduce production costs over a large number of them.

  • 1
    $\begingroup$ Starship specifically. 6 is still rather more than 2 $\endgroup$ Commented Aug 20, 2021 at 15:26
  • 2
    $\begingroup$ But only 3 can be used for landing. $\endgroup$ Commented Aug 20, 2021 at 15:26
  • 1
    $\begingroup$ Well, they better learn to make do with 2, cuz the rest of us below the karman line have to :) $\endgroup$ Commented Aug 20, 2021 at 15:29
  • 12
    $\begingroup$ Airliners know well before descent if their engines are working correctly, and even have a fighting chance at landing with zero working engines. The landing modes just aren't directly comparable. $\endgroup$ Commented Aug 20, 2021 at 15:34
  • 3
    $\begingroup$ @Russell Fighting chance? When a gliding airliner gets near an airport, they tend to succeed, and many off-airport landings result in flyaways also (TACA 110 @ NASA Michoud; Alrosa 514). The only way dual engine fail crashes an airliner is "too soon after takeoff" (Sully), using an approach guide slope below the airplane's unpowered glide slope (British Airways 38) or somewhere no airport is within glide range (Air Transat 236; Pinnacle 3701; Transair 810). Note every one of the latter group includes blunders. $\endgroup$ Commented Aug 21, 2021 at 0:59

Others have already mentioned the advantages in redundancy and manufacturing scale. Other advantages:

  • It's structurally more efficient to place engines near the skin of the vehicle. The Superheavy booster in particular takes advantage of this, with its outer ring of engines actually protruding somewhat beyond the diameter of the vehicle.
  • Using a large number of engines makes it much easier to achieve a wide effective throttle range by shutting engines down, which is important for recovery of the vehicle.
  • Smaller, lighter engines are easier to transport and handle. Raptors are small enough to be moved around with forklifts, which makes them easier to install or swap out.
  • Smaller engines are easier to test. Engine test stands are smaller, and vehicles can test fire single engines or subsets of their engines.
  • Smaller engines are less likely to inflict severe damage on the vehicle in the event of a major failure.
  • A large number of smaller engines is actually quieter, due to how incoherent noise sources add together. In short, with the same total sound power, the multiple incoherent sources will partially cancel each other out, resulting in lower average pressure levels.
  • The same goes for vibrations internal to the vehicle. People who've flown on both Crew Dragon and the Shuttle have remarked about how smooth the portion of the flight powered by the 9-engine booster was in comparison, and how the 1-engine upper stage actually felt rougher than the Shuttle.
  • 1
    $\begingroup$ "Raptors are small enough to be moved around with forklifts, which makes them easier to install or swap out." – Indeed. SpaceX has improved installation times tremendously. Just one year ago, it took roughly a day to install an engine, a couple of weeks ago, they installed all 29 engines on B4 in one night. (Although that was only a fit check, so it might be the case that the plumbing was not hooked.) The engine installation before the static fire test of B3 took only about an hour per engine. $\endgroup$ Commented Aug 24, 2021 at 12:34

As airline safety improved, they focused more on cost. Without safety, cost didn't matter. SpaceX engineers asked "what's the largest engine we can make and still have room for vacuum engines and multi-engine-out capability for landing?" The answer was the current engine size of raptor. There were more factors, but that's the gist.

Next, since engines are so expensive, they decided to mass produce that size. Mass production leads to innovations in cost and quality, which should increase safety.

Lastly, you want max thrust on liftoff to reduce fuel costs. So you pack the booster full of them, which ends up being 29-32 engines. That plus 3 on the ship and 6 more with vacuum nozzles means 38-41 total, which is a lot.

As safety is proven and technology improves, we may see fewer, larger engines on Starship V2. Bigger has historically meant more efficient, and fewer might lead to lower cost. But by then 3D printing or some other innovation might make fewer/bigger engines a bad strategy. Time will tell.

  • $\begingroup$ I might not be very bright, but I don't see what additional facts (apart from speculations) this answer has brought, to what the OP question has already hinted at, and to what others have already answered. $\endgroup$
    – Ng Ph
    Commented Aug 23, 2021 at 13:43
  • $\begingroup$ The last paragraph is roughly what happened with aircraft jet engines. The 747-100 needed four engines partly due to the tech limit on how powerful an engine could be, and partly due to the tech limit on how reliable an engine could be. $\endgroup$ Commented Aug 23, 2021 at 15:55

Let's say you want to build a big rocket (or airplane). You're going to need a lot of thrust. You can get a lot of thrust from a bunch of small engines or just a few big ones. Now, we know that more engines are more expensive than fewer engines, right?

But wait! How many rockets are we going to build? Probably not a lot. Of the heavy lift rockets, there have only been 13 Saturn Vs, 5 Shuttles, 2 Energias, and probably will only ever be 10 SLS(es?). Let's consider Starship--it's going to be reusable, so even though Musk wants a crazy launch rate, there's probably not ever going to be more than 15-30 built.

So what if we put only a few massive engines on each of our vehicles? There's probably going to ever be 50 (at most!) built. While that's not quite low enough that basically every engine will be hand-crafted, it's going to be pretty close. Building the tooling to make a rocket engine will always be very expensive, and with only 50 engines to divide that cost over, each engine is going to be wildly pricey.

Let's compare this to aircraft engines. The PW4000 and the GE90, two of the most common engines you'll find in a Boeing 777, have each had over 2,500 manufactured--and recall, this is on a century's worth of incremental work into jet engine manufacture by P&W and GE. Much of their tooling is probably shared with previous engines. These are far, far cheaper a peace than the 50 engines you're planning on building.

So when building a rocket, a market in which volume is very low and development costs are very high, it's actually more expensive to use fewer engines, because so few of them will ever be built. It makes more sense to develop a smaller engine--something which (we haven't mentioned this yet) is MUCH easier to design & validate than a very large engine--and just make a bunch of them.

Note: we haven't even mentioned the redundancy issues yet--if one of 20 engines dies, it's not that big a deal. If one of 4 engines die, you're not going to space today.

Afterthoughts: if we ever get to the point where there's 100s or 1000s of rockets being built a year, and engine design has matured to be very reliable, I expect you'll start to see only a few engines per rocket. But we're building 10s (at ABSOLUTE best!) of rockets per year.

  • 2
    $\begingroup$ "there's probably not ever going to be more than 15-30 built" – Elon Musk estimated they need about 1000 Starships. Estimating about one Booster for every 5 Ships (a totally uninformed guess on my side), that is a total of ~1200 rockets with a total of ~15000 engines. They have already built 17 vehicles, and have started construction on at least 2 more; they have also built around 70 engines. 8 of those vehicles and about 18 of those engines have already flown. $\endgroup$ Commented Aug 22, 2021 at 9:54
  • 3
    $\begingroup$ "But we're building 10s (at ABSOLUTE best!) of rockets per year" – SpaceX is already building 10 Ships per year by hand: they don't even have the factory yet! And they intend to have multiple factories near multiple launch sites (very likely at least one at Starbase and one at Cape Canaveral), including on Mars. $\endgroup$ Commented Aug 22, 2021 at 9:55
  • $\begingroup$ You should have consulted some public launch manifest, such as this $\endgroup$
    – Ng Ph
    Commented Aug 23, 2021 at 15:15
  • 1
    $\begingroup$ clearly I should have clarified: 10s of rockets per design per year. It's not like a 777 & an a321 share engines & ought to be counted in the same mass-production tally. I'm not a moron; I know more than 10 rockets are being launched between the Altas Vs and the Long Marches & the Falcons & the Antares & the Soyuzes & H-2s (?) & the whatever the ISRO launches & the Delta IV when they manage to not scrub. Are they all sharing engine tooling? absolutely not. $\endgroup$ Commented Aug 24, 2021 at 7:47
  • $\begingroup$ AH! This makes sense. Thx. $\endgroup$
    – Ng Ph
    Commented Aug 24, 2021 at 20:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.