If we assume that spaceX is optimizing for speed to Mars colony. Then we must assume they'll optimize for speed to develop starship and booster.
On such a scenario they'll be buidlng a LOT of Starship's and boosters to iterate on the design, to develop flight operations and to develop the industrial capacity to build them at scale.
On that scenario there's no scarcity of starship flights to launch starlink, there'll be excess capacity. What will they do to use that capacity?
1. Supply side capacity assumptions seem off to me. If costs to build and operate each Starship are relatively low, it could very easily be the case that it is profitable for SpaceX to operate far more launch capacity than is needed to launch Starlink satellites. The assumption would be that the marginal value of an additional Starlink satellites drops as Starlink coverage and bandwidth reach and exceed the required levels to provide competitive Internet services in relevant markets. In other words, eventually Starlink's market will be saturated, and more Starlink satellites will no longer add value, so the launch cadence for Starlink will slow to only the replacement rate for the satellites that deorbit each year. In this case, the opportunity cost tradeoff you describe above somewhat falls apart; SpaceX gains nothing by reserving excess slots for Starlink, and it starts to makes sense to allocate additional capacity to other profitable launch customers at lower and lower profit margins as the marginal value of a Starlink satellite declines.
2. Latent demand for low cost launch is likely so high that even profit maximizing monopolist behavior will result in much lower prices than we currently see. In other words, I think we do have enough data to make a guess about price elasticity of demand, and we can guess with high confidence that demand is VERY elastic depending on price. Our data points are straightforward: at $10k/kg or more, nation states were basically the only clients for launch. At the current $2500/kg, startups have blossomed wherever moderately well-heeled investors have had an interest, and there has been a ton of such interest. Even academic institutions have had undergrad classes building cubesats and actually getting them to space. And at, say, $500/kg, laypeople with a bit of disposable income will start being able to get involved. Speaking personally, I run a small propulsion tech startup, and as a pre-revenue company $500/kg would allow me to get my project into space on nothing but personal funds and an SBIR Phase 1, whereas $2500/kg is going to require me to win an SBIR Phase 2, or raise some venture capital. That difference is huge for my company, as it will preserve our cap table and give us better prospects in future rounds; the capability makes entering the cislunar economy much more attractive for businesses of all kinds. So, my expectation is that the profit-maximizing price really will be a lot lower than current prices, and though it will take a few years to get there (the whole Starlink constellation will probably launch first) I will bet that it does get there eventually. Probably not $5M/launch, but maybe $25M or $30M. For 150,000 kg to orbit, even those prices would make the market explode.
I don't have one! As I sketched out, I think it depends on factors I can at best guesstimate to an order of magnitude. Demand elasticity in particular I have absolutely no clue about (but then I don't know that anyone else does either, someone should really estimate that). I just don't find $5M/flight particularly realistic given the projections about Starlink.
If Michael's conjecture about dynamic trade-offs is correct, though, I would expect the initial flights to be below the static integrated monopolist's price. So, maybe cheaper than I would think based on Starlink revenues, but above $5M/flight (assuming Casey Handmer is right about the average total cost).
I somewhat agree with Michael, but I think you are being too fixated on Starlink numbers here. Linking Starship launch cost directly to Starlink marginal revenue could only work if Starlink launch demand is greater or equal than Starship launch rate.
SpaceX's plan for Starlink is for 42000 satellites with lifespan of 5 years. Current Starship design could only launch 54 Starlink V2 (volume limited). So, 42000÷5÷12÷54=13.
That's at most 13 launches per month.
SpaceX plans to built five launch pads for Starship (one next to LC-39A, two at Boca, a new launch site at LC-49), and there's a plan for sea launch platforms. Currently, SpaceX has launched nine times per month from three launch pads.
SpaceX's fully and rapidly reusable Starship is clearly planned to launch at least double the current rate of nine launch per month of partially reusable F9, that is way more than the demand for Starlink, an excess capacity.
So, with such an excess capacity, does the numbers changed?
> Linking Starship launch cost directly to Starlink marginal revenue could only work if Starlink launch demand is greater or equal than Starship launch rate.
I would have agreed with you, but... the optimization problem in the appendix (which drives the analysis) doesn't assume they have to be linked; it only assumes vertical integration, lack of capacity constraint, and profit maximization as a monopolist. From there the linkage to Starlink profits *emerges* as a solution; that piece wasn't assumed initially. The numbers are guesstimates, but without changing the underlying structure of the objective the character of the solution won't change.
As others have noted, it may be the case that SpaceX isn't a profit maximizer in the sense I assumed here. That could absolutely change the result. It may also be the case that past a certain size, the marginal profit from another Starlink satellite is zero; that could also change the result.
One important detail of economic optimization problems is that heuristics to increase profits have a funny way of leading firms to these solutions. I don't imagine anyone at SpaceX is sitting down and doing the same calculus I did (though they do have some folks with econ degrees who would be familiar with this type of thinking). But if they're ever looking at their spreadsheets and fiddling with numbers to eke out some more profit (e.g. to deal with a perhaps-temporary financing crunch), those optimizations will lead them along the path I described here.
Gonna be honest, I skip your appendix because I can't understand it (too many symbols for me). That's why I ask what's for your launch cost prediction, since you rejected the $5M/launch estimate.
>It may also be the case that past a certain size, the marginal profit from another Starlink satellite is zero; that could also change the result.
This, perhaps, since Starlink launches is a fixed number? Granted, I don't run the math, but if you have excess capacity, it surely makes sense to offer it at the price in which it doesn't link to Starlink profit?
Whether with Tesla, or SpaceX, we see the same philosophy: what I’ll call Total Commitment to each “cause”, which in his mind has always required Total Commitment to true mass production…. from. the. Beginning. … which, as he has said, allows him to “cover a multitude of sins”…. One of which is, for example, raising/lowering prices on almost a weekly basis depending on His market reading and where he wants to point things at the moment. IOW, your Assumptions may not be His….. - Dave Huntsman
If we look at what they are doing with Starship and the market demand (including Starlink), it just doesn't make sense economically. Like for example, if their plan is to have a launch vehicle for Starlink, designing it to be fully and rapidly reusable from the get go, just doesn't make sense (Relativity, Blue Origin, and Rocket Lab are a good examples).
They could reach orbit way earlier and way cheaper if they just go with an expendable second stage. But they did not. In fact, it's their very first focus. They are building two massive factories at Boca and Cape, that's way more than enough to build second stages that meet Starlink launches needs.
From launch pads to factories, it's clear this thing is built for a launch rate that's way more than the current Falcon 9 could ever achieve. And I'm sure that's also way more than any "reasonable" estimate of future market demand.
This is very true :) exactly why I stated my assumptions up front
> Relativity, Blue Origin, and Rocket Lab are a good examples
Notice that these companies aren't vertically integrated in the same way SpaceX is; their objective functions look meaningfully different as a result. For example, the second first-order condition (linking marginal launch costs to marginal satellite profits) wouldn't show up in e.g. RL's problem. The analysis here speaks more to the specifics of SpaceX's business than anything general about the sector.
As for demand... really, I have no idea what that will be or what the curve looks like beyond "it slopes downward". All I can say now is that (perceived) demand elasticity is critical to optimal pricing, with the markup being inversely related to elasticity. This is consistent with e.g. DOD customers receiving higher prices than commercial -- something I have heard but have not verified.
I 100% agree reusability is an interesting wrinkle. Again, the economics of reusability are subtler than I expected at first. I'll write a post about it soon!
Thanks for reading and commenting, this has been stimulating :)
I think we can conclude that there is some demand elasticity to launch although the drop in mass of payloads (both in straight mass and utility per kg) makes up the other part.
As far as not knowing what the curve will look like, we have been calling this the Starship Singularity. Its capabilities are such a significant break from current or past capabilities that the market really has no idea what to do with the things. When we gamed out the market opportunities the only economically useful thing Starship enables are things that involve large numbers of human beings. Starlink is a possible exception but you could accomplish the Starlink task with something a lot less ambitious.
And I think that's the key, SpaceX maximizes profit internal to each businesses post-deployment (e.g. they discount development costs to $0) but over the entire family of companies there is a goal that all of that profit and products are in service to and it doesn't appear to require that profit be maximized across the entire enterprise.
I suspect that SpaceX may not care all that much about being a monopolist in the launch business (low margins, high risk, bla bla) but it does care about monopolizing LEO comms and EO using the Starlink platform. Imagine the same constellation that is generating $30 B in comms revenue is also taking market share away from national security and civilian EO, providing mobile cellular backhaul and direct to cell service, etc. That revenue per Starlink may be too low.
Perhaps they don't! But notice that the theoretical analysis in the appendix doesn't assume any particular quantitative relationship between the cashflows from launches vs from satellites. That general optimization problem produces the second fundamental result for Starship slot pricing: the more profitable Starlink satellites are, the pricier Starship slots will be via the marginal cost channel.
Competition in the launch sector will drive down the markups (first result), but won't affect the second first-order condition -- that's entirely about how vertical integration affects the profit function. A violation of assumption 3 ("SpaceX is a profit maximizer") or a different form of the profit function (something other than "TR - TC" as I've written it) would certainly change the analysis.
I think SpaceX is a profit maximizer but over an atypical timeline. A profit maximizer can decide to forgo current profit in order to increase the size of a market that can later realize a far larger profit even assuming the discount rate on those future cash flows. E.g. would you call Palantir a profit maximizer when the company was unprofitable until only a few months ago? It could have been profitable sooner by increasing prices or decreasing costs but the total market opportunity may not have been realized.
Or, as one investor in our SpaceX SPV said, "I hope he lets us split our investment between SpaceX and Starlink when it IPOs because I want to see the short term gain from Starlink and I want my great-grandchildren to see the gain from owning a large chunk of the Solar System."
That's very possible. The problem I solved above is static -- while it can accoun for capitalized long-run costs and benefits, it can't account for dynamic trade-offs like you describe. Off the top of my head I'd expect the solution to such a dynamic problem to involve lower prices initially, with the internal discount rate used by decision-makers driving the balance between short-run and long-run profits. Prices would probably rise over time as the market is realized. The initial price and rate of increase would likely be constrained by the cost of funds (market discount rate) and expected rate of new competitor emergence (in both launch and constellation). It's an interesting problem for sure.
One factor you didn't account for.
If we assume that spaceX is optimizing for speed to Mars colony. Then we must assume they'll optimize for speed to develop starship and booster.
On such a scenario they'll be buidlng a LOT of Starship's and boosters to iterate on the design, to develop flight operations and to develop the industrial capacity to build them at scale.
On that scenario there's no scarcity of starship flights to launch starlink, there'll be excess capacity. What will they do to use that capacity?
So what is the opportunity cost for a non starlink launch on Starship? $500m?
Are you assuming that starlink demand never saturates?
What's your Twitter again?
A couple thoughts:
1. Supply side capacity assumptions seem off to me. If costs to build and operate each Starship are relatively low, it could very easily be the case that it is profitable for SpaceX to operate far more launch capacity than is needed to launch Starlink satellites. The assumption would be that the marginal value of an additional Starlink satellites drops as Starlink coverage and bandwidth reach and exceed the required levels to provide competitive Internet services in relevant markets. In other words, eventually Starlink's market will be saturated, and more Starlink satellites will no longer add value, so the launch cadence for Starlink will slow to only the replacement rate for the satellites that deorbit each year. In this case, the opportunity cost tradeoff you describe above somewhat falls apart; SpaceX gains nothing by reserving excess slots for Starlink, and it starts to makes sense to allocate additional capacity to other profitable launch customers at lower and lower profit margins as the marginal value of a Starlink satellite declines.
2. Latent demand for low cost launch is likely so high that even profit maximizing monopolist behavior will result in much lower prices than we currently see. In other words, I think we do have enough data to make a guess about price elasticity of demand, and we can guess with high confidence that demand is VERY elastic depending on price. Our data points are straightforward: at $10k/kg or more, nation states were basically the only clients for launch. At the current $2500/kg, startups have blossomed wherever moderately well-heeled investors have had an interest, and there has been a ton of such interest. Even academic institutions have had undergrad classes building cubesats and actually getting them to space. And at, say, $500/kg, laypeople with a bit of disposable income will start being able to get involved. Speaking personally, I run a small propulsion tech startup, and as a pre-revenue company $500/kg would allow me to get my project into space on nothing but personal funds and an SBIR Phase 1, whereas $2500/kg is going to require me to win an SBIR Phase 2, or raise some venture capital. That difference is huge for my company, as it will preserve our cap table and give us better prospects in future rounds; the capability makes entering the cislunar economy much more attractive for businesses of all kinds. So, my expectation is that the profit-maximizing price really will be a lot lower than current prices, and though it will take a few years to get there (the whole Starlink constellation will probably launch first) I will bet that it does get there eventually. Probably not $5M/launch, but maybe $25M or $30M. For 150,000 kg to orbit, even those prices would make the market explode.
So, what's your prediction on how much SpaceX would charge for Starship?
I don't have one! As I sketched out, I think it depends on factors I can at best guesstimate to an order of magnitude. Demand elasticity in particular I have absolutely no clue about (but then I don't know that anyone else does either, someone should really estimate that). I just don't find $5M/flight particularly realistic given the projections about Starlink.
If Michael's conjecture about dynamic trade-offs is correct, though, I would expect the initial flights to be below the static integrated monopolist's price. So, maybe cheaper than I would think based on Starlink revenues, but above $5M/flight (assuming Casey Handmer is right about the average total cost).
I somewhat agree with Michael, but I think you are being too fixated on Starlink numbers here. Linking Starship launch cost directly to Starlink marginal revenue could only work if Starlink launch demand is greater or equal than Starship launch rate.
SpaceX's plan for Starlink is for 42000 satellites with lifespan of 5 years. Current Starship design could only launch 54 Starlink V2 (volume limited). So, 42000÷5÷12÷54=13.
That's at most 13 launches per month.
SpaceX plans to built five launch pads for Starship (one next to LC-39A, two at Boca, a new launch site at LC-49), and there's a plan for sea launch platforms. Currently, SpaceX has launched nine times per month from three launch pads.
SpaceX's fully and rapidly reusable Starship is clearly planned to launch at least double the current rate of nine launch per month of partially reusable F9, that is way more than the demand for Starlink, an excess capacity.
So, with such an excess capacity, does the numbers changed?
> Linking Starship launch cost directly to Starlink marginal revenue could only work if Starlink launch demand is greater or equal than Starship launch rate.
I would have agreed with you, but... the optimization problem in the appendix (which drives the analysis) doesn't assume they have to be linked; it only assumes vertical integration, lack of capacity constraint, and profit maximization as a monopolist. From there the linkage to Starlink profits *emerges* as a solution; that piece wasn't assumed initially. The numbers are guesstimates, but without changing the underlying structure of the objective the character of the solution won't change.
As others have noted, it may be the case that SpaceX isn't a profit maximizer in the sense I assumed here. That could absolutely change the result. It may also be the case that past a certain size, the marginal profit from another Starlink satellite is zero; that could also change the result.
One important detail of economic optimization problems is that heuristics to increase profits have a funny way of leading firms to these solutions. I don't imagine anyone at SpaceX is sitting down and doing the same calculus I did (though they do have some folks with econ degrees who would be familiar with this type of thinking). But if they're ever looking at their spreadsheets and fiddling with numbers to eke out some more profit (e.g. to deal with a perhaps-temporary financing crunch), those optimizations will lead them along the path I described here.
Gonna be honest, I skip your appendix because I can't understand it (too many symbols for me). That's why I ask what's for your launch cost prediction, since you rejected the $5M/launch estimate.
>It may also be the case that past a certain size, the marginal profit from another Starlink satellite is zero; that could also change the result.
This, perhaps, since Starlink launches is a fixed number? Granted, I don't run the math, but if you have excess capacity, it surely makes sense to offer it at the price in which it doesn't link to Starlink profit?
Whether with Tesla, or SpaceX, we see the same philosophy: what I’ll call Total Commitment to each “cause”, which in his mind has always required Total Commitment to true mass production…. from. the. Beginning. … which, as he has said, allows him to “cover a multitude of sins”…. One of which is, for example, raising/lowering prices on almost a weekly basis depending on His market reading and where he wants to point things at the moment. IOW, your Assumptions may not be His….. - Dave Huntsman
If we look at what they are doing with Starship and the market demand (including Starlink), it just doesn't make sense economically. Like for example, if their plan is to have a launch vehicle for Starlink, designing it to be fully and rapidly reusable from the get go, just doesn't make sense (Relativity, Blue Origin, and Rocket Lab are a good examples).
They could reach orbit way earlier and way cheaper if they just go with an expendable second stage. But they did not. In fact, it's their very first focus. They are building two massive factories at Boca and Cape, that's way more than enough to build second stages that meet Starlink launches needs.
From launch pads to factories, it's clear this thing is built for a launch rate that's way more than the current Falcon 9 could ever achieve. And I'm sure that's also way more than any "reasonable" estimate of future market demand.
> your Assumptions may not be His
This is very true :) exactly why I stated my assumptions up front
> Relativity, Blue Origin, and Rocket Lab are a good examples
Notice that these companies aren't vertically integrated in the same way SpaceX is; their objective functions look meaningfully different as a result. For example, the second first-order condition (linking marginal launch costs to marginal satellite profits) wouldn't show up in e.g. RL's problem. The analysis here speaks more to the specifics of SpaceX's business than anything general about the sector.
As for demand... really, I have no idea what that will be or what the curve looks like beyond "it slopes downward". All I can say now is that (perceived) demand elasticity is critical to optimal pricing, with the markup being inversely related to elasticity. This is consistent with e.g. DOD customers receiving higher prices than commercial -- something I have heard but have not verified.
I 100% agree reusability is an interesting wrinkle. Again, the economics of reusability are subtler than I expected at first. I'll write a post about it soon!
Thanks for reading and commenting, this has been stimulating :)
I think we can conclude that there is some demand elasticity to launch although the drop in mass of payloads (both in straight mass and utility per kg) makes up the other part.
As far as not knowing what the curve will look like, we have been calling this the Starship Singularity. Its capabilities are such a significant break from current or past capabilities that the market really has no idea what to do with the things. When we gamed out the market opportunities the only economically useful thing Starship enables are things that involve large numbers of human beings. Starlink is a possible exception but you could accomplish the Starlink task with something a lot less ambitious.
And I think that's the key, SpaceX maximizes profit internal to each businesses post-deployment (e.g. they discount development costs to $0) but over the entire family of companies there is a goal that all of that profit and products are in service to and it doesn't appear to require that profit be maximized across the entire enterprise.
Demand. Chicken-and-egg problem. Either way, we eating good :)
I suspect that SpaceX may not care all that much about being a monopolist in the launch business (low margins, high risk, bla bla) but it does care about monopolizing LEO comms and EO using the Starlink platform. Imagine the same constellation that is generating $30 B in comms revenue is also taking market share away from national security and civilian EO, providing mobile cellular backhaul and direct to cell service, etc. That revenue per Starlink may be too low.
Perhaps they don't! But notice that the theoretical analysis in the appendix doesn't assume any particular quantitative relationship between the cashflows from launches vs from satellites. That general optimization problem produces the second fundamental result for Starship slot pricing: the more profitable Starlink satellites are, the pricier Starship slots will be via the marginal cost channel.
Competition in the launch sector will drive down the markups (first result), but won't affect the second first-order condition -- that's entirely about how vertical integration affects the profit function. A violation of assumption 3 ("SpaceX is a profit maximizer") or a different form of the profit function (something other than "TR - TC" as I've written it) would certainly change the analysis.
I think SpaceX is a profit maximizer but over an atypical timeline. A profit maximizer can decide to forgo current profit in order to increase the size of a market that can later realize a far larger profit even assuming the discount rate on those future cash flows. E.g. would you call Palantir a profit maximizer when the company was unprofitable until only a few months ago? It could have been profitable sooner by increasing prices or decreasing costs but the total market opportunity may not have been realized.
Or, as one investor in our SpaceX SPV said, "I hope he lets us split our investment between SpaceX and Starlink when it IPOs because I want to see the short term gain from Starlink and I want my great-grandchildren to see the gain from owning a large chunk of the Solar System."
That's very possible. The problem I solved above is static -- while it can accoun for capitalized long-run costs and benefits, it can't account for dynamic trade-offs like you describe. Off the top of my head I'd expect the solution to such a dynamic problem to involve lower prices initially, with the internal discount rate used by decision-makers driving the balance between short-run and long-run profits. Prices would probably rise over time as the market is realized. The initial price and rate of increase would likely be constrained by the cost of funds (market discount rate) and expected rate of new competitor emergence (in both launch and constellation). It's an interesting problem for sure.
Thanks for reading and commenting!