What are the top five innovations that SpaceX was able to accomplish that allows the first stage to land vertically and be reused?

I am sure the Apollo program would have liked to reuse their first stage. What changed between Apollo and Spacex that allowed Spacex to land and reuse first stage assemblies?

May I guess:

  • Better/faster software and computing power
  • Better engine modulation

Thanks, looking for your input.


2 Answers 2


I don't believe there is any technical innovation that is responsible. After all, all the other competitors have access to the same technologies, but they aren't doing it.

I would argue the technology was there before:

  • Blue Origin also autonomously lands their New Shepard booster, and even did it successfully before SpaceX did (although, of course, New Shepard is not an orbital system, it just goes straight up and down again).
  • The DC-X demonstrated autonomous vertical landing in 1993 (it was intended to land from orbit, but all demonstrations were just hops of at most ~3km; on its penultimate flight, it even demonstrated the Starship-style reorientation from belly-down to landing feets-down).
  • Autonomous landing from orbit (but not vertical) was demonstrated with Buran in 1988.
  • Autonomous landing of an aircraft was demonstrated in 1945.

The innovation is a commercial one. It is the simple realization that it could actually make sense to do this.

Before SpaceX, nobody was doing it, because nobody saw the need to do it: why reuse a rocket when you can simply charge the government 10 times as much money to build a new one every time? It was the commercial realization that there are organizations without an infinite supply of money that also want to launch stuff into space, and the commercial realization that there is enough demand for launching stuff into space that building rockets fast enough would become a problem, which led SpaceX to focus on reusability.

Note that 37% of all SpaceX Falcon 9 launches were for a single customer. And that customer is SpaceX itself! If we exclude the first 70 launches and only focus on the launches since Starlink actually exists, that number goes up to over 50%. More than half of all Falcon 9 launches since the first operational Starlink launch have been Starlink launches.

Or, looking at it from the payload perspective instead of the launch system perspective: 50% of all satellites launched by all countries in the entire history of spaceflight are Starlink satellites. That means, SpaceX has launched as many satellites just for Starlink in 4 years as all of humanity combined in 70 years.

Now, all of a sudden, cost becomes very important because you don't have a customer paying for the rocket, you yourself are paying for the rocket. And if you are launching thousands of satellites, you have to launch very often, very fast, otherwise it takes decades to launch your constellation, so you simply do not have the time to build rockets fast enough. Now, reuse not only makes sense but becomes a necessity.

But Starlink is a customer no other company has. So, no other company (so far) has seen such a pressing need for reusability.

I don't believe it is a coincidence that among the other companies working on reusability, some of the ones pushing hardest, are Rocket Lab and Blue Origin. Rocket Lab also designs, builds, and runs satellites, and offers a full end-to-end mission service package. And Blue Origin is owned by Jeff Bezos, who also owns Amazon, which is planning to launch their own Internet mega-constellation.

There is one technology that helped especially with the drone ship landings, though: the ubiquitous availability of high-precision, high-accuracy, realtime position information through GNSS systems like GPS, GLONASS, Galileo, and Beidou, especially combined with augmentation systems like dGPS and SBAS (e.g. WADGPS, WAAS, NDGPS).

As mentioned above, the first autonomous landing of an aircraft was achieved in 1945 using technology that later became the internationally standardized Instrument Landing System that is still in use today. While it would be possible to use a similar system for rocket landings, it would require some large antennas and power hungry transmitters on the drone ships. GNSS definitely helps here, but it is not strictly necessary, especially for land landings.

It is quite ironic that most of SpaceX's "old space" competitors like Boeing, Lockheed-Martin, Northrop-Grumman and co. are not only launch system providers but also weapons manufacturers and defense contractors as well as spacecraft designers. All of them have already demonstrated precision guidance in hypersonic, supersonic, transsonic, and subsonic regimes using various aerodynamic control surfaces such as fins, speed brakes, and grid fins. Both missiles and precision bombs, for example, use grid fins. Several of them have built VTOVL or STOVL aircraft that perform jet-powered vertical landing (which is not too dissimilar from rocket-powered vertical landing). All of them have autonomous drones. Several of them have demonstrated autonomous precision landing of spacecraft on other celestial bodies.

Hitting a target with high accuracy is quite literally their business.

Furthermore, Boeing is the current owner of McDonnell-Douglas, the company which already demonstrated a VTOVL orbital vehicle in 1993. Boeing is also the current owner of Rockwell, which built the Space Shuttle Orbiter.

And yet, Boeing did not implement reusability either in SLS or in Vulcan. And why would they? For SLS, their customer seems to be quite happy to pay \$1–\$4 billion (depending on estimate) per launch, so why make the rocket cheaper through reuse if that only means you get less money?

ULA only seriously started to consider partial reuse of Vulcan after they got awarded the 38-launch mega-contract for Kuiper. However, their reuse concept does not involve landing; instead, the engine section and thrust structure which also houses the first stage avionics, detaches from the booster and reenters using an inflatable aeroshell. This concept had been presented for 10 years, but it only existed in PowerPoint and work didn't start until after the Kuiper contract.

  • 1
    $\begingroup$ I think denying technological advances that did make vertical landing of the first stage precise, practical and efficient - is simply wrong. It is the same as saying that AI does not change our way of life. $\endgroup$ May 13 at 17:00
  • 5
    $\begingroup$ @TheMatrixEquation-balance: All of those technological advances are available to ULA, Boeing, Northrop Grumman, Lockheed Martin, Arianespace, Mitsubishi, the Russians, the Chinese, the Indians, the Koreans, etc. as well. All of them could have done it. Heck, Boeing, which is part of ULA, bought McDonnell-Douglas, the makers of the DC-X, so in some sense, ULA did do it 30 years ago. But none of them have done it, not because the technology didn't exist, but simply because they thought it makes no sense. $\endgroup$ May 13 at 17:50
  • 6
    $\begingroup$ @TheMatrixEquation-balance no, it's not a philosophical issue, you just don't have any expertise in these topics but feel compelled to comment anyway. I wish you'd spend more time listening instead. $\endgroup$
    – Erin Anne
    May 13 at 23:52
  • 2
    $\begingroup$ @TheMatrixEquation-balance - be careful how you use the word 'AI' since it has a specific meaning for 'Artificial Intelligence', and you generally do not want them in flight control systems, see space.stackexchange.com/a/61895/26356 and the need to get a reliably real time solution. You would certainly be right in saying the techniques they do use to get landing are post Apollo though, being a result of the early fly by wire work. $\endgroup$ May 14 at 0:49
  • 3
    $\begingroup$ grand edit. Wish I could upvote it again. $\endgroup$
    – Erin Anne
    May 14 at 23:00

Apollo mind set In case of Apollo the focus had been placed on getting the absolute maximum performance out of the largest rockets to put the greatest tonnage into LEO for the Moon landings as quickly as possible. Recovery was probably not considered at all or was rapidly ruled out on the basis that it would decrease the tonnage capable of being delivered to LEO, that it would delay the Moon landings and that it was not a practical proposition to fit legs onto a Saturn V booster without redesigning it.

SpaceX mind set Elon Musk went through the basic underpinnings of rocketry from first principles with a different mind set. His objective was not to beat the Russians to the Moon using what could be developed quickly in the 1960’s without consideration of costs. Musk’s focus was on making as much money as quickly as possible for his Mars ambitions and minimizing the $/tonne cost to LEO using twenty first century technology.

Computer technology changed beyond recognition in the 50 years between 1960 and 2010 and could be used to simulate in far greater detail what previously had to be roughly calculated by hand.

Given these changes in goals and circumstances it was possible for Elon Musk to realise that a relatively modest reduction in absolute payload to LEO was a price worth paying for a greatly reduced overall $/tonne to LEO. And that with modern materials, technology and simulation it could be developed into a practical proposition.

  • 1
    $\begingroup$ "Computer technology changed beyond recognition in the 50 years between 1960 and 2010 and could be used to simulate in far greater detail what previously had to be roughly calculated by hand." The Apollo Guidance Computer in the LM did the real-time calculations needed to point and throttle the LM descent engine for landing. The basic control loop is not that complicated. Doing it under Earth's higher gravity and with the complications of aerodynamic drag, is (in computing terms) only a little bit harder, done repeatedly in the 1990s by DC-X. $\endgroup$ May 13 at 20:51
  • 2
    $\begingroup$ "Recovery was probably not considered at all or was rapidly ruled out" -- there were several recovery plans considered for Saturn V stages, including but not limited to a gigantic catch helicopter. $\endgroup$
    – Erin Anne
    May 13 at 21:40
  • $\begingroup$ @RussellBorogove But how many Crays would you need to match the finite element fluid dynamic simulations that can now be on a high end graphics card? Vs using a less aerodynamically optimized nose, and less efficient engine turbines, and just substituting bigger parts instead. Its not so much about the control on vehicle, as the software used to optimize the design of the vehicle. $\endgroup$ May 14 at 13:52
  • 1
    $\begingroup$ So the vehicle is 5% more efficient. That’s great, but it’s not a breakthrough that suddenly made recoverable first stages possible. $\endgroup$ May 14 at 19:04
  • 1
    $\begingroup$ @Russell Borogove the calculation is not at all straightforward. In fact it is very complex. The behaviour of the aerodynamic surfaces changes between supersonic, transonic and subsonic giving different effects on the vehicle. Variable wind direction and speed must be taken into account. On top of that unlike the LEM, Falcon 9 cannot hover as the Merlin engine cannot be throttled down sufficiently to balance the mass of the rocket. So the calculation has to be to achieve zero velocity at zero altitude when the engine is cut. That's why they had so much problem with the early landing attempts $\endgroup$
    – Slarty
    May 15 at 11:45

Your Answer

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

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