Are modular space stations cost effective?

Part of my series on countering misconceptions in space journalism.

In my previous blog, Unpopular Opinions In Space, I wrote that while expensive, developing a bigger rocket is often a cheaper and easier solution to any given problem in space. In this post, I elaborate on this theme while trying to understand design constraints in space station construction.


Humans have launched 16 space stations since the 1960s, of which 10 were functional enough, for long enough, to be occupied at least once. Following the cancellation of the Soviet N1 Moon rocket, the Soviet Union built a series of space stations for civilian and military purposes, exploiting their Proton medium lift launcher. These stations weighed about 20 tons and had a pressurized volume of about 100 cubic meters, which is similar to the average living room.

Starting in the late 1970s, Salyut 6 and 7 pioneered modular station design, wherein a space station would be built from multiple parts, launched separately and integrated in space. Extra modules brought more supplies and could extend missions. Salyut 7 was followed by Mir, which was built with four launches over four years starting in 1986. About 10 years later, during the Shuttle-Mir program, three additional smaller modules were added, and by the time Mir was splashed in 2001 it had hosted 125 astronauts and cosmonauts in its 350 cubic meters of pressurized volume.

Of comparable size and cost (~$2b) was Skylab, launched on the last of the Saturn V rockets in 1973. Because the Saturn V was substantially more capable than the Proton, Skylab was launched in a single shot, and subsequently hosted 3 crews for a total of 171 days.

The 1980s saw the first flights of the Shuttle. With its enormous cargo bay and reusable design, Shuttle was originally intended to enable the construction of Space Station Freedom. By the time its moral descendant, the International Space Station (ISS), was funded the Shuttle was nearing the end of its serviceable life. Also, by 1998, it was painfully obvious that Shuttle was never going to achieve anything like its design flight rate and costs, which were initially targeted at $100/kg to orbit, with about 50 launches a year. Actual numbers were more like $60k/kg with up to four launches a year.

And so the ISS was built with 33 main launches over 13 years, costing more than $150b and providing just over 900 cubic meters of pressurized volume, of which about 350 is not consumed with equipment. That is, for nearly 100x the cost and 20x the time, the ISS provides 2.5x more volume than Skylab, which was designed in the late 1960s. This is not progress.

Why is this the case? Modular space stations are a neat solution to a difficult problem. How to build a bigger station than can fit in an available rocket? Simply break the components up, build a set of common interfaces and plug it all together like LEGO.

And yet it is clear from the ISS’s example that modular efficiencies do not scale perfectly forever, or indeed at all. For example, the Saturn V could have been rebuilt from scratch for 1/20 the money in 1/4 the time, and 4 Skylabs docked together would be much larger. To take a more modern example, the SpaceX Starship (currently in development) has more than 2400 cubic meters of internal volume, including tanks, and is targeting a mere $5m launch cost.

Image result for starship and ISS
(Image: SpaceX)

Of course, there’s nothing magical about 900 cubic meters of volume. It just happens that this is the capacity of a Shuttle-launched station pushed to the most grotesque limit. If the ISS designers had had a mature Starship to work with we might have ended up with a huge Stanford torus. It is possible, though I think unlikely, that at some critical size space station utility undergoes a positive inflection, where operations magically become generically profitable.

To take the other extreme, consider Rocketlab. The Electron is a plucky rocket but if New Zealand wanted their own space station, I very much doubt they’d attempt to assemble one from thousands of pieces all weighing less than 200kg!

Indeed, if a new station was needed of a certain mass, then the designers could do worse than examine which launchers or payload masses had a high mass capacity to orbit on a per-year basis! For instance, while more launches of smaller payloads (~5T) occur each year, the peak shifts upwards (~20T) when normalized by total mass. Station designers may benchmark module mass at the limit of the most capable launcher available, which today is Falcon Heavy. But if a station requires more than, say, 4 modules with the launcher of choice, then it will probably save time and money to develop a new launcher!

Modules have certain advantages. They can be specified and farmed out to multiple contractors, allowing parallel construction with decoupled schedules. They are easier to transport and may be able to take advantage of existing launchers. Building huge new launch systems is fun but if they aren’t routinely flown then the skills to run them may be degraded over time.

Modules also have a range of disadvantages. Mechanically, modules are linked together with docking adapters – heavy double doors that are used only once and serve as stress concentrations as the station moves. Indeed, metal fatigue within ISS modules places a definite limit on station longevity. Even astronauts working out shakes the station enough to damage the structure and eventually cause cracks!

Schedule benefits are clearly not what they would seem. Every modular station ever built has not achieved more than one major component per year (on average) during construction. If the purpose of the station program is endless construction, then that is no big deal. But if the purpose is to execute and commission some well-defined design, then a smaller number of modules will enable the program to finish more quickly. It’s not impossible to imagine the mass production of identical modules, but no stations are built simply to maximize internal volume. Each module has a unique purpose so requires a lot of custom equipment. None of that equipment can break anything else, even on other modules that aren’t built yet, and none of the equipment can break since the launch is so expensive. All these factors add costs and slow production rate.

Modules have a major systems-level drawback too. In theory, modules are like LEGO and can be reconfigured in an infinite number of ways. But in practice, they are all exquisitely engineered machines with overlapping sets of requirements. The effect is that every additional module increases the complexity of systems interactions and degrades the versatility of the structure as a whole.

This aspect of tightly coupled, complex systems is well known to designers of nuclear power stations or large jet planes. In short, in tightly coupled complex systems, failures can occur and propagate too quickly for human operators to understand or mitigate, greatly increasing the risk of catastrophe. The ISS, due to large numbers of modules, is compromised at the architectural level in a way that no amount of careful design and redundancy can fully address.

There is a further design challenge. Conway’s Law describes how systems always mirror the communications structure of the organization which designs them. The effect of splitting a space station into many different parts is that the requisite interfaces create new and otherwise unnecessary barriers to communication, negatively affecting overall integration and increasing complexity. Relatively few subsystems or behaviors will fit exactly within one module, so in practice already complex systems such as life support or power will either have to be duplicated or split between multiple modules, exponentially increasing failure modes and complicating analysis.

Individually, small design compromises here and there are not a big deal. In the aggregate, however, they subject an already marginal system to gratuitous levels of functional degradation. Getting stuff to orbit and keeping it working is hard enough!

This is why, I think, that the ISS produces fewer human-hours (per month) of research than Skylab or Salyut, despite weighing 20x more and having nearly 10x as much internal volume. Of the 6 astronauts flying on the ISS at any one time, about 5.5 of them are busy with station maintenance.

Space station construction is just one area where building a bigger rocket is the cheapest, quickest solution. Indeed, even SLS development is proving to be cheaper and faster than ISS construction.

Regarding NASA’s goals with human exploration of the Moon or Mars, it is not clear to me that the space station is able to answer any of the important questions. Would a station built on the Moon or Mars be better placed for research? Definitely! Would another station built in LEO at vast expense help to get to the Moon? NASA didn’t need one the first time around!

This blog isn’t the place to relitigate design choices that were made for the ISS well before I was born. The concept of modular design was evidently compelling enough to give it a try! The real question, in 2019, is whether we will learn from our mistakes. Will our future space stations be more focused, discrete vehicles designed to execute on a well-defined mission?

Space station design has to reflect purpose. We don’t need to rebuild the ISS from scratch to relearn the harm of excessive modularity!

14 thoughts on “Are modular space stations cost effective?

  1. Great read.

    I’m not sure I’m convinced that ISS proves that modularity is a poor choice, or if this particular implementation of modularity is a demo of the possible pitfalls of modularity. Be careful not to reason from the particular to the general.

    Ditto with the Space Shuttle. Have we proved that reusability is highly expensive? That winged vehicles are inferior to capsules? Probably not, just that this particular design of a reusable spacecraft + launcher is expensive.

    Liked by 2 people

  2. The only current proposal for a US space station seems to be the LOPG/[Lunar] Gateway near the moon. Are you suggesting an alternative design to NASA’s proposals ?

    The quickest and cheapest way to have a less modular design seems to be to use Starship. Design now and build once a Starship payload user guide is published by SpaceX ?

    China seem to be following the Soviet MIR model : self-docking modules rather than ones that need EVAs to assemble.

    Probably the same arguments could apply to a base on the Lunar surface, but if you want a multinational collaboration, or even a multi-supplier build, a modular approach seems most practical. Well defined interfaces should isolate module internals from each other (just as in software). ?

    Liked by 1 person

      1. Thanks. I like your May 2019 blog on the Lunar Gateway. [but] There you say “An enormous fraction of the ISS’s total mass is dedicated to mating adapters…” – I’d love to know what that fraction is.

        You mention again that the astronauts exercising is fatiguing the ISS joints but the exercise equipment [now at least] seems isolated from the ISS structure.

        The problems with the modularity of the ISS may be more due to how NASA has defined the interfaces (the Common Berthing Mechanism?), constrained by the limitations of STS, and the desire to involve other nations.

        – but I agree, Skylab might have been a better foundation and approach.


  3. The hypermodular ISS was an effect of NASA’s decision to build the Space Shuttle with a payload bay 15 ft diameter x 60 ft long. There were several ideas floated in the 1970s and 80s for a “Shuttle C”, a cargo version of the Shuttle that replaced the Orbiter with large cargo pod carrying 50 tons of payload. Those pods could have been the basis of a large space station consisting of several units instead of the two dozen ISS modules.
    And, of course, there were many ideas floated for using placing the External Tank into LEO, none of which made it past the PowerPoint level.
    Interestingly, in Skylab NASA had launched a potentially much larger space station. Both the Skylab payload and the attach S-II second stage of the Saturn V were placed into LEO that day. The S-II was jettisoned within minutes after reaching LEO. Von Braun had considered a “wet/dry workshop” idea in which Skylab and the S-II LH2 tank were connected by a tunnel. That LH2 tank would be scarred for equipment that would be moved from the Skylab Workshop into the S-II tank. The combined pressurized volume of that space station would have been 1435 m^3. ISS has 916 m^3 of pressurized volume. So near and yet so far.

    Liked by 2 people

  4. Interesting read.

    On astronaut time occupation: do we have evidence that a non-modular station would require less mantainance? I tend to agree with the other comment that mentioned too few data point to decide against modularity.

    Another point that must be considered is political stability. One could see modularity as an insurance against program cancellation by assigning the construction to different countries. This happened when the USA wanted to finance the Constellation program with ISS money by deorbiting it in late 2010s. Political pressure of international partners prevented that, and this allowed doing science in it for a longer time.

    Liked by 1 person

  5. Interesting points on the difficulties/drawbacks of modular design, but there are several purposes for having work spaces in orbit. Should they all be physically separate? Presumably each would still need habitation areas, work areas, labs, etc. I would think there could be added expenses in having duplicated power supplies, having to take smaller shipments of supplies to more destinations, etc. Could it be that, as others have commented, that ISS is just a bad implementation of modularity? Or, at a more basic level, assuming that modularity were implemented in its best form, is there a non-modular form that is even better?

    This blog is several months old now and NASA has awarded a contract to the private company that will create a successor to the ISS, Axiom Space. This successor will be modular, and the first couple modules will be attached to ISS in about 2024/5 before separating in about 2028 and thereafter growing on its own. I don’t know what Axiom is planning to use to get its modules to space – Falcon Heavy? – but would they be better off planning on having separate facilities, separate space stations, one for tourists, one for a national lab, one for manufacturing?


  6. It´s a bit unfair to just compare internal volume and cost. A large part of the cost of ISS is the equipment inside, which would be expensive no matter the launcher used. ISS is not just a collection of metal boxes. You also suggested a station made of 4 skylab sized parts – so still modular, just from larger modules. So I guess the real question is how large the modules should be? Building an even bigger rocket just for a single use to launch a space station makes no sense, you use what you have. So obviously modules as large as possible, but saying 20t was the wrong choice is easy in 2019, not so much in 1970s when Saturn 5 was canceled. Shuttle was supposed to be a cheap way to LEO and if it worked as intended then ISS would be cheaper. You are blaming ISS for the economy of the shuttle.

    Liked by 1 person

    1. I think that a few modules is possible but like anything, it can be taken too far.

      Building a bigger rocket for one launch makes no sense, you say, but is unlikely to have cost $400b and taken 30 years.


      1. What’s the $400b ? Your blog says $150b. The $150b I guess includes module design and build costs, and the cost of 33 STS missions to launch and assemble them. Its arguable if an STS mission should be costed at $1.5b (including share of dev costs), or just the oft quoted $450m marginal cost.

        Looking at Constellation (Ares V) and SLS, there seems to be no upper limit to how much time and money NASA can spend if asked to build a bigger rocket.


  7. Interesting. Just out of mental boredom, I´ve been designing two different 1G circular stations. One is based on a 50 segment dodecagon and the other an 84 segment hexadecagon. Both use appoximately 20m x 6m modules. The smaller one is about 25,000 cubic m. The larger one about 40,000. The idea would be to use electron welding combined with outer end mounting points to ensure there is no pressure on the seals once welded. The outer rim in both would be a single circle with a 2m x 2m opening and walkway between them. There would be perpendicular science modules between each living/sleeping segment. So the stations could support over 100 people each and the larger as many as 400 if pushed. A docking ring would support up to 7 Dragons at each end of the 0G axle. The axle would also have 6 or 8 water/air storage modules at .1G. The spokes would be comprised of hydroponic modules at varying Gs. There would be 6 or 8 aquaponic modules connecting them for growing fish and filtering water. The goal would be to produce almost all of the necessary food, water and oxygen on board. It seems far fetched compared to existing 0G station proposals, but dramatically more realistic than the Von Braun and Clarke Stations proposed by one organization as an apparent money making scheme. Those are off the chart impossible in our lifetimes. These are large enough to allow “mass production” of the 6 or 7 module types (6 to 16 of each needed) and yet small enough to be launched and assembled with under 100 launches each. I actually think the modules could be partially constructed through extrusion processes or 3D printing, which will be critical for future in space construction. The goal is also to use basically a corrugated (triangular substructure) multi layer exterior hull which allows structural water channels that heat and boil water to purify while also helping cool and transfer and disippate heat around the metal hull. This would also act as radiation shielding and help absorb micro meteorite impacts.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s