manufacturing cost of SpaceX first stage

Has SpaceX ever released the estimated manufacturing cost of the Falcon 9 first stage?
From published reports, they talk about the total manufacturing cost of the rocket as about 60 million. Obviously a good part of that will be in the part that contains 9 engines. But does anyone know how much?

SpaceX has not publicly released build costs for the Falcon 9 (other than offhand and often contradictory estimates) but a review of comparable class vehicles shows that the actual materials and fabrication costs represent about 10% of the total launch costs, and in no case exceed 20%.[SUP]1[/SUP] For common liquid propellants such as RP-1/LO[SUB]X[/SUB] the propellant cost is somewhat less than 1%.[SUP]2[/SUP] So, if fabrication isn’t the driving cost of a launch, where does all of the money go?

Costs for a launch vehicle can be broken down into three general categories:
[li]By mission/launch fabrication and operational costs[/li][li]Logistics systems and maintenance costs[/li][li]Non-recurring engineering (NRE) involved in design, improvement, new facilities and upgrades anomaly investigations, et cetera[/li][/ul]

The first is an operational cost per mission; the second is a fixed cost per annum that has to be spread across the flight rate; and the third is a design & development cost that gets spread over the life of the vehicle. How these are accounted for in the per mission costs depends on the costing model used and how cost and profit is accounted. It is reasonable (from an initial standpoint) to assume that the logistics systems, maintenance and NRE costs are essentially fixed, and therefore the only way to affect a reduction is to spread them across more launches per annum.[SUP]3[/SUP]

The by-mission costs can be broken down in to materials & fabrication, vehicle integration, payload integration, mission analysis and mission assurance, launch operations, and range costs. These will all vary depending on the type of mission and requirements.[SUP]4[/SUP] For instance, a low precision “get it into orbit” smallsat delivery may not require much in the way of GN&C studies, special provisions for monitoring or protection of the payload, multiple coupled loads analysis (CLA) cycles, mission assurance data deliveries, et cetera. A high value payload requiring delivery to precise orbital parameters may require substantially more analysis or payload facilities, as well as more verifications in the integration process to assure mission success. But the biggest cost driver in virtually all launches is the integration and verification test cost, which requires substantial labor. Efforts have been made to attempt to automate or reduce labor costs of integration but not with significant cost reduction while maintaining the same standard of verification; the costs either get shifted over to development and operation of the so-called “expert systems” which do the automated checkout, or so many corners are cut that potential problems are missed, resulting in significant anomalies and mission failures. Often, so much additional complexity is created in the attempt at automation that it simultaneously increases cost and risk.

In other areas, such as mission analysis and launch operations, reductions in effort are theoretically possible but developing a suite of integrated analysis tools that can automagically perform an end-to-end mission simulation for all possible variations of payload and mission parameters or hardware that can automate launch operations is more than a program can afford to invest for the marginal reduction in cost over a reasonable number of missions. A reduction in any particular area only represents a fraction of the by-mission cost. Even assuming a stage can be reused without significant rebuild or refurbishment (unlikely) it only means a fractional reduction in cost as other steps of vehicle integration still have to be performed.

NASA performed a study of the costs of developing and operating a multi-stage reusable vehicle in 1970 as part of the Phase B proposals for what became the Space Transportation System (Shuttle) which concluded that a flight rate of 50-60 flights per annum would be required to be favorable to an expendable system of the same class.[SUP]5[/SUP] In the early 'Nineties, Orbital Science Corporation performed a similar study and came to the exact same conclusion, which is unsurprising because despite developments of the intervening years launch vehicles were still using chemical propulsion systems of the same essential capability operating near the same basic material limits. Several other studies of reusability have been performed by various sources (industry contractors, Aerospace Corporation, Air Force Research Laboratory, European Space Agency, et cetera) which have come to basically the same conclusion. In a nutshell, reusability, even if achievable at flight rates well beyond conventional experience, is at best marginal in reducing overall launch costs.

Note that there are other reasons to strive for reusability, at least in a limited sense; for one, assuming that a vehicle is sufficiently reusable that refurbishment can be done at a depot site rather than return to factory it can allow for more throughput in production, so if you are aspiring to build up a large fleet of vehicles it makes sense. A partially reusable system may make sense in making use of expensive or difficult to manufacture components. Reusability in crewed systems is desirable from the standpoint of all the costly integration that has to be performed on a system which, by definition, experiences limited environments compared to the propulsion system. But there are significant development and testing costs that goes along with making reusability viable without significant refurbishment costs. The Shuttle is an object lesson in how seemingly viable reusability in concept can translate into an enormous cost sink in trying to make it work in reality, especially when wedded to an inherently flawed design. An effort to develop a genuinely reliable vehicle needs to have sufficient flexibility and budget to make significant design changes and even modify or eliminate requirements that are contrary to achieving reliability and reusability.

The real opportunity for reduction in launch cost lays in scaling (the larger the payload, the less cost per mass), simplifying both the vehicle systems and launch operations (allowing for both less effective risk and reduced labor), reduction of logistical systems and maintenance (being able to integrate and launch with a minimum of support systems), and getting the payloaders to build more robust space vehicles. Minimum cost design concepts such as the Bob Truax Sea Dragon and the various Big Dumb Booster concepts offer achievable reductions in operating costs using conventional technologies without additional complexity and resulting mass and reliability drivers.


[SUP]1[/SUP] The high end of the variance is a difference between the way costs are broken down between solid propellant rocket motors and liquid propellant engine systems; the propellant is a significant part of the fab cost of a solid motor and costs more per unit impulse than liquid propellants, but historically the overall cost of the vehicle is significantly lower, so fab ends up being a larger portion of a smaller overall bill; this becomes especially complicated when looking at vehicles which use both liquid engines and solid motors.

[SUP]2[/SUP]All cryogenics, such as LH[SUB]2[/SUB]/LO[SUB]X[/SUB] are somewhat more expensive, depending on volume and delays requiring detanking, and hypergolic propellants like the Aerozine 50/N[SUB]2[/SUB]O[SUB]4[/SUB] used on Titan increased significantly in the past fifteen years, reaching or exceeding 5% of the cost of launch.

[SUP]3[/SUP]In reality, significantly more launches probably entails more facilities, logistics systems, and NRE, and so it’s not a simple division; regardless of how many launches are conducted there will be some fixed minimum of cost per mission.

[SUP]4[/SUP]The advertised costs that SpaceX publishes for its launches are a bare-bones manifest cost, which gives the user a slot on the manifest, an initial and final CLA cycle, basic mission analysis with loose tolerances on orbital precision, and basic payload integration in a 100K facility. If a payloader requires more stringent accuracy or cleanliness requirements, health & status monitoring, data for independent mission assurance, et cetera that are typical for higher value commercial or government payloads, there is additional significant cost.

[SUP]5[/SUP]This was the source of the 50 flights per year requirement for the STS even though all of the contractors were well aware that even with the original planned fleet of five Orbiter Vehicles a rate of 24 flights per year split between Cape Canaveral and Vandenberg would be challenging.

Interesting. Where does that 10% number come from? Is that based on historical data? I could believe it if it was based on NASA’s costs, as it’s pretty inefficient operationally.

But unless I’m missing something, it doesn’t seem right for SpaceX. A Falcon 9 launch has a commercial price of $61.2 million dollars, plus payload prep costs and such. Assuming a reasonable margin for risk and profit, that would mean SpaceX’s cost is probably more like $50 million, and maybe less. If the rocket itself only makes up 10% of the launch cost, then the entire Falcon 9 rocket only costs $5 million, including all stages? That sounds completely wrong to me. If so, and the first stage was half of that cost, and there would still be a cost to refurbishing the booster between flights, we’re talking about a cost savings from reusability of only a couple of percent? SpaceX wouldn’t be going to this effort if that was the case.

Perhaps your accounting looks at the entire cost of a launch platform from inception through retirement to come up with that 10% number? If so, that would be irrelevant to calculating the marginal cost of each new launch of a production rocket.

Thinking about it more, the 10% number for the actual vehicle build cost cannot be correct for SpaceX. The Falcon 9 has ten Merlin engines in it. Now, the cost of those engines is not public, but speculation in some of the engineering forums I read is that they are roughly $1 million each, give or take. Certainly that’s the ballpark. In that case, the engine cost alone for a Falcon launch would be on the order of $10 million dollars, which is in the neighborhood of 20% of the launch price.

Even if we assume that SpaceX could build those engines for $500,000 each (which would be remarkable), there’s your 10% of the launch cost right there, just for the engines.

You did say that the analysis was based on ‘comparable vehicles’. I assume that’s the Delta IV, Atlas V? The versions of the rockets that match the Falcon 9 V1.1 performance cost anywhere from $150 million to $350 million with an average mission cost of $220 million according to ULA. I could believe the 10%-20% cost for the vehicle build for those missions, as they seem to have a hell of a lot of operational overhead compared to SpaceX. However, even that seems iffy as the RD-180 engine the Atlas V uses costs $10 million alone.

SpaceX is launching cargo missions to ISS for $100 million, and simple satellite missions for $75 million or so.

Any analysis of the build-out cost of the Falcon 9 has to take into account that SpaceX is already undercutting ULA by 50% or more, so the price of the rocket must be a higher percentage of their overall costs unless they’ve managed to cut the cost of rocket building by a similar percentage already.

Musk has said that 75% of the cost of F9 is the first stage.

Right. So if the build cost of the rocket was 20% of the launch price, the entire rocket would cost about 12 million, and the first stage would cost about 8.5 million. I don’t think that passes the smell test. The nine Merlin 1D engines alone probably cost that much. If the vehicle build only accounted for 10% of the launch cost, then the first stage would only cost 4.25 million to build. Sure doesn’t sound right to me.

The data I’ve looked at is for medium lift class vehicles developed for the US Air Force, looking at the actual per vehicle build and integration costs not including the amortized development costs that are factored into the “block buy” contracts. I wouldn’t use vehicles designed for NASA for comparison because those are without exception either designed specifically as crewed vehicles to higher reliability standards and/or develop cutting edge technologies with operational modes that are different from standard multistage rockets. There are some significant differences in costing due to the different acquisition and reporting requirements (EELV vehicles were developed under FAR Part 15 versus the commercial acquisition of the Falcon 9 under FAR Part 12)[SUP]1[/SUP] but I’ve tried to estimate actual build and integration costs.

You are correct in noting that SpaceX has a different cost structure from ULA, and also that the costs of ULA missions appear to be overstated even given the greater FAR requirements; one study estimated that the real price of ULA launches should be about 60% of the costs the government was paying (and they should be able to build and fly twice the volume using existing facilities), and interestingly, ULA CEO Salvatore Bruno admitted that they could get costs down to about half of the current prices in response to competition with SpaceX being certified for EELV contracts. Given that, it is reasonable that the proportion of hardware manufacturing costs for the Falcon 9 should be higher. Without having any data on SpaceX manufacturing costs or scrap rates I can only guestimate but $8M to $12M for a Falcon 9 is probably a reasonable estimate, which would make the fab cost (of the whole vehicle, including lower stage, upper stage, interstage, payload adapter, and fairing) about 13% to 20% of the manifest launch cost of $61.5M (as of 2015 Q4)[SUP]2[/SUP].

Let’s say it is 20% of the advertised launch cost for for the first stage alone; that means even if the stage is totally reusable with minimal refurbishment, you can look at a reduction to 80% of the manifest costs for repeat flights. That’s not insignificant, but it isn’t the half-order-of-magnitude reductions Musk and Shotwell have been promoting. And it doesn’t consider the costs involved in adding features for recovery and additional robustness for repeated use. Realistically, a returned stage will have to be disassembled and inspected; engines cleaned and retested; sensitive components such as valves and seals replaced; avionics tested, COPVs removed and NDId, et cetera. This may still be a reduction in cost over new manufacture, but it will likely be on the order of a few percent rather than offering a dramatic slashing of costs, hence why the historical estimate of needing to achieve flight rates of 50-60 flight per annum to achieve cost parity still holds.

Now, SpaceX may be doing some kind of magic and get more reusability with less refurbishment, but at the end of the day, the cost savings from reusing propulsion hardware are marginal. There just isn’t enough trade space or fungible costs reduce the overall costs by more than a modest amount. The bulk of the costs in the end price to the user come from integration, processing, launch operations, schedule overruns, et cetera. Most of the viable solutions for reducing those costs have little to do with the launch vehicle per se; they have to do with streamlining the acquisitions process (for government launches), committing to realistic rather than “success oriented” schedules, reducing the integration effort through realistic automation and testing; developing standard payload interfaces and tools to automate mission analysis; and simplifying launch and range operations. The biggest impact one can make to reduce costs from a vehicle design standpoint is to reduce complexity and make the vehicle robust against failure, which is a challenging set of requirements from an engineering standpoint. The benefit of reusability, if it exists, is in allowing you to have a higher throughput of launches from a given production capacity and thus amortize fixed operating costs over a larger number of launches and/or ensure higher volume of cash-flow (and thus, higher EBIT and realized net profit). There is a real benefit here in making more money without additional investment in facilities, but one that would require a large volume of flights with minimal refurbishment costs.


[SUP]1[/SUP]FAR is the Federal Acquisition Regulations which specify how a program must be contracted and governed; FAR Part 12 is an acquisition of ostensibly commercial-off-the-shelf capability with no development or quality data provided and the program overseen only at a contract management level; FAR Part 15 is intended for programs that receive development funding and are overseen at a technical level by government representatives or independent contractors with detail data provided.

[SUP]2[/SUP]It is interesting to look at the history of advertised launch costs of the Falcon 9 vehicle:
2005: Falcon 9 (v1.0) US$ 35.0M for 10,450 kg to LEO
2010: Falcon 9 (v1.0) US$ 44.0M
2011: Falcon 9 (v1.0) US$ 54.0M
2014: Falcon 9 (v1.1) US$ 56.5M for 13,120 kg to LEO
2015: Falcon 9 (v1.1) US$ 61.2M

This kind of cost growth is typical from an early development program to maturity, but per kg to orbit but it does represent a 40% in cost per unit mass to LEO from inception, notwithstanding the actual costs of launching; costs for the self-insured government payloads have been around US$ 90M to 100M.

Perhaps this is the reason Musk talks about eventually turning around the first stage in a day or two. By drastically reducing the time, he per force reduces the cost of reuse even without considering the increased number of launches. The reusability advantage isn’t in the initial manufacture, but being able to fly an already tested and proven component. Instead of testing six components for six new launches, test once and fly that component six times. Whether it works for SpaceX only time and determination will tell.

According to this report, the materials cost of an entire Falcon 9 booster is 1.2 million. I guess that is the cost of the raw aluminum etc. that goes in to the manufacture. Doesn’t seem right, but then I have no experience in manufacturing.

That’s not necessarily wrong. Materials cost is drastically different from manufactured cost. 2,000 pounds of steel costs about $170. 2,000 pounds of steel in the form of a car starts at about $5,000 and goes up from there.

I think it is safe to assume that non reusable vehicles have a less lopsided cost ratio between stages.
With that in mind the savings from re-usability are even lower when compared with one-shot rockets

I may be misunderstanding you, but the 75% figure quoted above refers to a non-reusable F9. The quote is from 2013.

Oh, never mind then.