There have been several times in recent weeks that people either in person or on the internet have speculated about SpaceX’s finances and business model. In some cases the speculations have been that SpaceX is pricing their Falcon 9 below cost to try and drive ULA and other competitors out of business. I’ve seen other analyses on the pro-SpaceX side that seem to think that $6M F9R flights to orbit are right around the corner. I finally got curious enough that I wanted to run some of my own numbers using publicly available information to see what I could learn. I was a bit worried at first that if my analysis didn’t come out looking anything short of amazing for SpaceX that I’d get burned at the stake as a heretic, but I decided to publish this analysis anyway, including my spreadsheet I came up with so others can play with it and draw their own conclusions. Also, to save space I’m going to put all my disclaimers into this footnote1.
The four main questions I wanted to try to answer were:
- How realistic/sustainable is the current $61.2M price for an expendable Falcon 9?
- How realistic is their goal of $40M for a partially-reusable Falcon 9?
- How realistic is their goal of eventual $6-7M Falcon 9 flights if they could recover/reuse both stages?
- How much has SpaceX had to rely on spending pre-payments from future launches?
This was a fairly brief, 3hr exercise, so consider at best a really, really crude SWAG, but I gained some interesting insights I had never noticed before. Before I get into answering my two questions, and discussing those interesting insights, I’d like to first discuss the methodology I used, and provide a copy of the spreadsheet in case you want to start poking around.
The point of this exercise wasn’t to create a precise cashflow history of SpaceX, but to create a simplified model that could give some insights into some of the questions I previously mentioned. I wanted to stick with publicly available sources, and wanted to keep things simple and high-level. So I gathered key pieces of data including headcount at various times, list prices for their vehicles, timing of launches, and data on the value of government development contracts they’ve received 2. I then made several core cost assumptions:
- I assumed that each employee cost SpaceX an average of $150k/yr burdened. This is lower than the traditional $200-250k/yr number you hear for bigger industry player, but is consistent with numbers I’ve seen for aerospace startup companies. This number includes not just direct salary, but also employment taxes, unemployment insurance/worker’s comp, fringe benefits (health insurance and other benefits), and other direct costs of employing people (computers, software seats3, etc). I had originally gone for $120k/yr for this number, but found that the numbers jibed better with historical data points we had if I used the higher $150k/person/yr number.
- I assumed that “overhead” cost 50% of the direct labor costs. Overhead normally includes the salaries of non-engineering labor (executive, marketing, business, etc), but since I’ve just used a lump rate based on the headcount, this overhead number just includes facilities, infrastructure, tools and machinery, and R&D costs that weren’t for producing flight hardware. So Merlin test engines that didn’t fly would fit under this. As would the pads, the factories, the drone ships, Grasshopper, any Raptor development, etc. This is ridiculously high-level, but it would take a lot of work to get anything more precise. But this seems pretty reasonable since the labor cost included fringe benefits, and all of the indirect labor.
- I assumed that on average for Falcon 1, Falcon 9, and Dragon, that the non-labor cost of goods sold was 25% of the item’s list price. Ie, if SpaceX says an expendable Falcon 9 costs $60M, I assumed that non-labor COGS was ~$15M. This covers raw materials, consumables, range fees, shipping costs, non-labor marginal testing costs, and all the components and sub-assemblies SpaceX still purchases for the launch4. This number was based on my assumption that SpaceX was probably targeting list prices that they expected to be able to make at least a 10% profit margin on, and that ~2/3 of the cost was labor and fixed costs. It was also based on my limited experience at Altius on space hardware projects. The analysis didn’t seem wildly sensitive to this number, but it’s also one of the ones I’m least confident in.
I don’t think any of those are wildly controversial, though you can try fiddling with the numbers as you see fit using the spreadsheet I provided.
I then made some more or less dubious simplifying assumptions (many documented in comment boxes in the spreadsheet):
- Instead of trying to figure out exactly when SpaceX got paid for what milestones for each contract, I typically took the contract value and duration, and evenly spread the value of the contract over the years in question. In reality payments may have been more lumpy, more front-loaded or more back-loaded. If someone really wants to take the time to dig through public records to try and time the payments more accurate, be my guest (and send me a copy of the updated spreadsheet and I’ll put it in an update).
- I inflated the value of old contracts and revenues into 2016 dollars using Wolfram Alpha, to make costs and revenues easier to compare with 2016 numbers. I’m not sure what inflation calculator they used or if it’s one that is uncontroversial. I also realized that on the cost side I didn’t inflate the labor costs to 2016 dollars. So that may be an update worth doing down the road to be internally consistent. Non-labor COGS is inflated though because it’s based on the inflated list prices. Most importantly, I’m not 100% sure this was a useful way to do things. But that’s what I did.
- For commercial Falcon 1 and Falcon 9 flights, I assumed the actual revenue for the flight was 10% above the list price on average for “added services.” As I’ve heard from some people who’ve spoken with SpaceX in the past, the list price covers a pretty basic service, and that many items people care about cost extra. One NASA mission (I can’t remember if it was Jason or DISCOVR) for instance was listed as costing $97M even though the list price was only ~$60M. I don’t think there’s anything wrong with this, but I wanted to account for it. Once again, if you disagree with my 10% estimate, feel free to tweak it up or down. I didn’t include this added service 10% for CRS flights, since I assumed those were baked into the price.
- Probably the biggest and most explicitly incorrect assumption I made was that SpaceX only got paid for flights upon completion, and that all of the revenue (and non-labor COGS) for that flight happened in the year that the flight occured. In reality, in order to get a manifest slot you almost always have to pay a non-refundable deposit, and there are many milestones along the way, that typically front-load a lot of the cost of a launch so that by the time you get to the actual launch, you’ve already paid most of the money for the flight, with the actual flight itself only the last of several milestones. This is pretty common in industry as I understand it. The reasons I didn’t include some sort of modeling of prepayments was for a few reasons: a) I don’t think there’s enough public info to accurately time prepayments for commercial flights even if I wanted to, and b) one of the main questions I wanted to answer was how much SpaceX was using spending front-loaded pre-payments to finance cashflow. Fortunately, we should be able to estimate how much of the prepayment money they’ve spent based on the difference between cumulative revenues + investments – costs. Basically if they’ve spent more money than they have received from completed flights, R&D contracts, and investment, it seems like the only real place that money could come from would be spending pre-sales.
- For CRS flights, I assumed that the Falcon 9 + Dragon cost $1.6B/12flts = $133M/flt in dollars of the year that the flight occurred. I assumed that the price of the Dragon and added services was this $133M number (inflated to $2016) minus the inflated Falcon 9 list price for that year.
- I assumed SpaceX got paid list price even for flights that failed. My guess is this isn’t precisely true, but probably closeish.
- For years when I didn’t have an explicit headcount, or one I could remember for the earliest years, I interpolated from years on either side.
- For years leading up to the first Falcon 1 flights, I assumed that Elon invested money to counter any difference between the revenues and costs, since there weren’t many preorders they could take milestone money from to finance cashflow.
- There was ~$100M of non-Elon investment in SpaceX from various groups like Founders Fund and DFJ in the 2008-2012 timeframe. We had timing of the 2008 money, but not for the remaining $80M, so I evenly distributed it (and inflated it to 2016 dollars).
- I only went up through 2015 in the historical data, so I don’t include any revenues or data for Dragon V2 flights, Falcon Heavy flights, or reusable Falcon 9 flights in the historical section. Even in the what-if sections, I explicitly leave out Dragon V2 flights5 and Falcon Heavy flights to try and focus on the specific questions I wanted to answer, and to keep this model from rapidly ballooning into something too complex to get useful information out of.
So all told, this is a flawed, but hopefully useful model. And one you can tweak to your hearts content to see if you come to different conclusions than I.
How Well Does the Model Do?
We don’t have a lot of data points to compare with, but there are two data points worth looking at:
- The model suggests that not counting DARPA and USAF contract R&D money, Elon had to put in ~$103M of his own money to get Falcon 1 to the point where it was flying. That’s around $85M in then-year dollars, so in the right ballpark for the ~$90M Falcon I development budget you hear quoted. It’s not perfect, but close enough to suggest we’re in the general ballpark.
- The cumulative cost through the first flights of Falcon 9/Dragon in 2010 are estimated at ~$800M inflated. The canonical number was $390M for Falcon 1 + Falcon 9v1.0 development, which would imply that Dragon was around $350M or so once you separated out operational costs and such. This also seems close-ish.
I’m sure that with more time and a more granular approach you could probably get the numbers closer, but this suggests we have a model that’s at least in the right ballpark.
Answers to The Four Questions
So, if you assume that my model isn’t entirely useless, we can now take a look at my three questions from earlier. This is what the “what if” columns on the right are for. The short version is that I think:
- While I think you can make the case that SpaceX isn’t yet to a flightrate where they are making profit at the current $61.2M list price, my model suggests they’d only need ~13 Falcon 9 flights with 3 of those being CRS flights in 2016 in order to breakeven at that price point. With the amount of people and infrastructure they have, 13 flights per year (with 3 being CRS flights) doesn’t seem unreasonable, even if they don’t make it all the way there this year. So this confirms my intuition that their $61.2M number for Falcon 9 isn’t so much them trying to sell at a loss to push out their competitors, but more them not having reached the flight rate that they’re theoretically capable of with their current team and infrastructure.
- Based on this model, I also don’t think getting down to $40M/yr for a semi-reusable Falcon 9 is totally unrealistic. There’s a lot of squishiness in my model about how I account for reusability, but it seems like we’re probably only talking about ~15-18 flights with ~3 of those being Dragon flights in order to make that at least somewhat realistic, assuming the downsize to ~4000 people after the commercial crew development is over. Which seems doable. If they keep their full 6000 people, they’d need nearly 30 flights per year to break even at $40M/flt, which seems optimistically high, but I don’t think they need the full 6000 people once the commercial crew development and certification is completed. This more or less confirms my intuition that a modest price decrease with reuse seems realistic.
- Dramatic drops in price seem pretty optimistic though. Even if you assume that the non-labor COGS drops by 90% per flight with reuse, and that they can get back down to 2500 people to service everything, it still seems like you’d need >50 flts per year to make those prices work, and I don’t consider that remotely realistic yet with the current market. If they kept their current team size, they’d need over 100-150 flights per year to make the $7M/flt number work… I don’t think that’s likely to happen. That said, even a 30% drop from their current prices is pretty amazing.
- There does seem to be some merit to the belief that SpaceX has been living off of prepayments. If you ignore the $1B fidelity investment last year (ie assume that it was set aside explicitly for the satellite business, and not used to finance cashflow), SpaceX has currently spent around ~$1.2B of prepayments (down a little from a high of around ~$1.3B in 2014). If you assume that they priced Falcon 9 with only a modest 10% profit margin6, that means that around $1B of that prepayment money that they’ve already spent is money they’ll need to carry out the missions on their manifest. With ~40ish flights on their manifest, they have a backlog worth ~$3B, so that represents a lot of their backlog that they’ve already spent. They’ve probably done some of the work for those flights, but does anyone really think they’ve done ~1/3 of the work needed for those 40 flights? Probably not. That said, is this some fatal problem? Probably not. The Google/Fidelity investment is about the same size as this amount, so even if something were to happen they’re probably safe now. And with the commercial crew contracts, they actually had more completed revenue than costs, and that’s likely going to get better this year if they can get their flight rate up. Lastly, so long as their manifest continues to either grow or at least stay steady, that will also help with cashflow. Unless they have another launch failure and 6 month standdown within the next year or two, I think they’re probably safe. Or at least as safe as any other commercial operator in this industry.
There were also several other interesting observations that stuck out to me:
- SpaceX has only recently reached the point where their revenue from actual flights has surpassed their revenue from DARPA and NASA R&D contracts. They’re currently at $1.7B from flights vs $1.5B for R&D contracts. And most of their flight revenue to-date has come from CRS missions.But while SpaceX has benefited a lot from their public-private partnership with NASA, it looks like over the next several years more and more of their business will be coming from commercial and non-NASA customers.A lot of their current team-size is likely driven not by Falcon 9 and Dragon fabrication and flight operations, but development work for Commercial Crew. With how NASA chose to run the Commercial Crew program, more as a traditional contract development with deep NASA oversight, maybe this isn’t that surprising.If SpaceX suffers another launch failure in the next 1-2 years, I think they could probably survive as a company, but expect that would significantly delay their ability to field their LEO commsat constellation–ie they’d have to spend a lot of that Fidelity/Google investment to cover cashflow while they work through the return to flight.I wouldn’t be surprised if SpaceX downsized after commercial crew certification is over. It would make achieving their cost targets more realistic, and they probably won’t need as big of a production staff if reuse really pans out in the way they expect.Based on my model’s prediction of their cost structure, and how much of their prepayments they’ve already eaten through, I’m skeptical that they’re moving anywhere near as fast with Raptor and MCT as most of their fans seem to think they are.
All told, I think this was an interesting exercise, even if it turns out that some of my assumptions were off by a bit. My big takeaways are that SpaceX’s current price numbers seem realistic, and their $40M price target with reuse is also probably also achievable eventually. Their financial situation seems less precarious now than it has been at any point in their history, though even one more launch failure anytime soon would hurt quite a bit. I also really don’t think they have a clear path forward to the more optimistic numbers they’ve thrown out, even with full stage reuse, but $40M for a Falcon 9 is still pretty amazing. I genuinely hope ULA and/or Blue Origin can continue to step up their game enough to stay competitive with SpaceX–it would be awesome to have two or three US providers able to launch rockets reliably at those kind of prices–that would go a long way towards enabling the kind of space future we’d all like to see.
Anyhow, go nuts with the spreadsheet, and if anyone has a ton of time on their hands and wants to try and time the revenue and estimate prepayments and all that better than I have, I’d be interested in seeing what you come up with, and may even post the results if they’re interesting enough.
[Update 1: A commenter pointed out that this site (https://www.sec.gov/cgi-bin/browse-edgar?company=space+exploration+technologies&owner=exclude&action=getcompany) provides data on the timing and size of previous SpaceX investment rounds. Between this and government data on contract payment timing, we could probably increase the fidelity of the spreadsheet by quite a bit. If anyone wants to do that, let me know. If not I’ll see if I can find the time to do that in the coming days.
Also, while this impacts revenue going forward, not historical revenue, SpaceX did win a contract for matching funds from the USAF for upper stage engine development. This will help lower the number of flights they would need to break even by at least one or two. Continuing to win big government development contracts like this will help SpaceX going forward.]
- Disclaimers: #1: I’ve done paid work for both SpaceX and their main competitor ULA, though while I’m not doing any work for either of them at the moment, I’ve done significantly more work for ULA over the past 5yrs. I also have friends at both companies, so I’m not the most unbiased person in the world. #2: I’m not doing this with any input from anyone at either SpaceX or ULA. This is purely an amateur exercise that in no ways represents the official position of SpaceX, ULA, or my own company. #3: I’m not a finance/green-eyeshades guy, and am not claiming this is an accurate financial assessment. In fact, I’m pretty sure this model is at best a SWAG. I’m going to make all sorts of simplifying assumptions, and this was based on a couple of hours of scouring public sources on the internet, so take my model with the appropriate sized grain of salt. That said, I’ll provide the spreadsheet so you can play with it and make it as fancy as you want.
- primarily for the DARPA FALCON SLV program, and the NASA COTS and Commercial Crew programs
- Don’t underestimate how expensive this can be–good engineering software is often crazy expensive even for lower-end software than what SpaceX uses
- They may do the assembly and manufacturing of many big ticket items in-house, but they’re definitely still ordering a huge number of components such as valves, actuators, structural elements, fasteners, plumbing fittings, sensors, electronics, connectors, etc, etc. And that stuff can add up really quick.
- Though I expect they’ll likely cost and be priced around the same range as the cargo Dragon flights, so theoretically you could treat a Dragon V2 flight as another cargo Dragon flight for sake of spreadsheet tweaking.
- Probably a safe bet