Impact of SpaceX rideshare on small sat launchers market

[M.E.T.]

If you could make $10 profit while leaving some business for your competitors, but instead you choose to make $2 profit while starving your competitors, is that illegal? Nope.

That’s what’s happening here.

Smart.

How sure of that can we be?  Is there documentation or quantitative argument that supports it?

FWIW, and I'm too busy/lazy at the moment to hunt it down, but I seem to recall that Transporter is actually SpaceX's second attempt at dedicated rideshare; the first time their prices were too high, and their offering failed to gain traction until they met with smallsat company reps to find out what they were willing to pay.  This would have been a few months before the current Transporter service was announced IIRC.

Maybe Rocketlab should do the same. Of course, since they are ALREADY losing money on a $6M launch, it’s not going to help them to hear that what most smallsat customers can afford is actually less than half that amount.

[RedLineTrain]

Nah, there are ways to do it. Stoke’s approach is one way.

Stoke hasn’t even built a rocket let alone a launch business. I don’t think it’s possible for any of the small LVs to make it just by being a launch company. It’s not that big of a market and has some significant established players. The amount of launches they would need are unrealistic. Just look at the SPAC decks for Astra and Virgin. You’re talking 40+ launches a year, which just doesn’t exist. Rocket Lab did it right by having adjacent businesses. Also note that we know Rocket Lab loses money on every launch but I don’t see anyone calling them a monopoly trying to stomp out their small LV competition.

It seems obvious that nobody's rocket launch business plan will close.  All of these plans (small, medium, and large, reusable and expendable) were made when Falcon was launching at most at a rate 1/4th the current rate and fairing recovery was still experimental.  So SpaceX's marginal launch costs were maybe 3x what they are now.  And now, Falcon is the most reliable rocket family in the history of spaceflight and can give you guaranteed slots on short notice.

I have reservations about Rocket Lab's move into an adjacent business.  It is a good move to survive, but all of these payloads are surviving at the forbearance of SpaceX.  As soon as the non-communications markets become big enough, I would imagine that SpaceX would wish to move into those markets.

[john smith 19]

It seems obvious that nobody's rocket launch business plan will close.  All of these plans (small, medium, and large, reusable and expendable) were made when Falcon was launching at most at a rate 1/4th the current rate and fairing recovery was still experimental.  So SpaceX's marginal launch costs were maybe 3x what they are now.  And now, Falcon is the most reliable rocket family in the history of spaceflight and can give you guaranteed slots on short notice.

I'd dispute a 3x reduction in launch costs. Refurb costs of both fairings and stages will go up with the number of landings that need to be dealt with.  If that can all be done by the existing workforce all well and good but if you take on new hires then the that's a cost rise.

There would obviously be a sweet spot where the existing workforce is fully occupied but no more. I doubt anyone outside senior SX management knows if they are at, above or below this number of launches right now.

I have reservations about Rocket Lab's move into an adjacent business.  It is a good move to survive, but all of these payloads are surviving at the forbearance of SpaceX.  As soon as the non-communications markets become big enough, I would imagine that SpaceX would wish to move into those markets.

The whole "Everyone is doooomed once starship starts launching" line has served SX well and created enough fear, uncertainty and doubt in the industry to have probably killed at least a few of the more badly run projects before they started.

So let's look at what we know.

Even at F9 size it's quite difficult to fill up a dedicated ride share launch, hence the need for aggregators to round up sufficient payloads together.

The lessons from Ariane 5 are quite instructive. Arianespace had a fair stretch when they could routinely get two primary satellite payloads. As comm sats got bigger it was more difficult to find one "normal" sized and one just-a-bit-smaller that needed a similar orbit (because despite using hypergolic propellants their US could only do a single start, a stunningly dumb decision IMHO).

At some point SS will start flying and it's SX's stated goal (as with F1) to retire F9 ASAP after that.  At which point that problem will become much more difficult.

150t of small sats is a lot of small sats, but SX say they will be able to do an SS launch for less than the cost of an F9 launch. If they can't then they get hit at both ends. Their costs will have gone up at the same time the problem of getting full occupancy has also gone up. Of course they could launch with a part load (which could still be several full F9 flights) to build up their safety case for HSF

Of course this assumes starship is flying. It's now over 6 months since their last attempt. TBH I'd thought 3x over "Musk time" would have covered it but it looks like I was being optimistic. 

Maybe their next one will be immaculate, once they've closed out all the issues the FAA have identified.

However given they didn't manage a full set of engine ignitions (has the Raptor 2 design been frozen or is it still ongoing?) on stg1 they didn't get to stage sep and their new TPS is completely different to what they've used on F9 and Dragon IMHO their competitors still have some time in hand before being blown away like dust in the wind.

The thing about arguments of "Historical inevitability" (Marx, "the end of history", starship eliminating all competition etc) is this.
They aren't.

So I don't think most of SX's competitors will be laying down to die just yet.

[RedLineTrain]

I'd dispute a 3x reduction in launch costs.

Happy to hear your estimate of the marginal launch costs at a 100/yr launch rate.  Let's say $10 million?  $15 million?  It's got to be much less than two years ago at a 30/yr launch rate and no fairing reuse.

So I don't think most of SX's competitors will be laying down to die just yet.

Falcon is killing SpaceX's competitors pretty well nowadays.  No need to bring Starship into the discussion.  The Falcon of today is a much more formidable competitor than the Falcon of two years ago.

[Zed_Noir]

<snip>
The lessons from Ariane 5 are quite instructive. Arianespace had a fair stretch when they could routinely get two primary satellite payloads. As comm sats got bigger it was more difficult to find one "normal" sized and one just-a-bit-smaller that needed a similar orbit (because despite using hypergolic propellants their US could only do a single start, a stunningly dumb decision IMHO).
<snip>

The Ariane 5's orignal hypergolic upper stage, the EPS have single engine start. However the later EPS-V hypergolc upper stage on the Ariane 5 ES(V) have multiple engine starts. The lower performance EPS and EPS-V stages can loft 6.6 and 7.5 tonnes to GTO respectively, so can only launch one larger payload or two smaller payloads. The Ariane G/G+/GS series with the EPS upper stage retired after 2009. The Ariane ES(V) only launch ATVs to the ISS and three pairs of Galileo navigation satellites before being retired after 2018.

The Ariane 5 ECA with a non-restartable HydroLox upper stage can loft about 10 tonnes of payload to GTO. The follow-on Ariane 5 ECB with a restartable HydroLox upper stage can loft 12 tonnes to GTO was shelved and replaced by the Ariane 5 ME design that was also shelved.

As larger payloads grew beyond 5 tonnes, finding a complementary smaller payload gets harder. Especially after SpaceX starts siphoning off the smaller payloads.

[john smith 19]

The Ariane 5's orignal hypergolic upper stage, the EPS have single engine start. However the later EPS-V hypergolc upper stage on the Ariane 5 ES(V) have multiple engine starts. The lower performance EPS and EPS-V stages can loft 6.6 and 7.5 tonnes to GTO respectively, so can only launch one larger payload or two smaller payloads. The Ariane G/G+/GS series with the EPS upper stage retired after 2009. The Ariane ES(V) only launch ATVs to the ISS and three pairs of Galileo navigation satellites before being retired after 2018.

The Ariane 5 ECA with a non-restartable HydroLox upper stage can loft about 10 tonnes of payload to GTO. The follow-on Ariane 5 ECB with a restartable HydroLox upper stage can loft 12 tonnes to GTO was shelved and replaced by the Ariane 5 ME design that was also shelved.

As larger payloads grew beyond 5 tonnes, finding a complementary smaller payload gets harder. Especially after SpaceX starts siphoning off the smaller payloads.

Thank you. I stand corrected. So we're awaiting A6 with the Vinci engined stg2 at present?

[john smith 19]

Happy to hear your estimate of the marginal launch costs at a 100/yr launch rate.  Let's say $10 million?  $15 million?  It's got to be much less than two years ago at a 30/yr launch rate and no fairing reuse.

It should be fairly obvious that the floor price will be set by the price of a stg2, since that is the biggest non-reusable component.

When I did my little cost modelling game for this SX were saying that the proportion of the launch price of the US came to $17m.

Mass production (by rocket industry standards of "mass production") can lower that a bit, but how much? Perhaps Coastal_Ron could make a suggestion? The standard aircraft industry "learning curve" was a 15% fall in unit price per each doubling of production volume.

A quick Google says they've launched F9 274x, that's 2^7 roughly, which would suggest they've cut costs by 105%.

IOW every new F9 US is making them money  . As you can see deciding what the "initial production run" was makes quite a difference if you are trying to figure out a learning curve effect.

And something I've never quite been sure about it is how much (or rather how little) what's being made is allowed to change before the learning curve has to be reset. Obviously how it's made can change quite a lot (that's where the benefits come from after all), but how much can what is being made change?

Because I'm pretty sure that the stg2 design being built now is a looong way from the one that launched in 2010, and I'm not sure how many of these launches have been made with that design.

So I don't think most of SX's competitors will be laying down to die just yet.

Falcon is killing SpaceX's competitors pretty well nowadays.  No need to bring Starship into the discussion.  The Falcon of today is a much more formidable competitor than the Falcon of two years ago.

True but a few of them will still survive.  As others have pointed out you have to deal direct with SX if you want the rock bottom launch price, otherwise you're dealing with an aggregator, and they will have a substantial markup.

SX's stated goal has always been full reusability (remember Musk saying he'd feel he failed if he hadn't built a fully reusable vehicle?). That's their stated end goal and is necessary for mars settlement. But it's also a damm sight bigger than anything apart from SLS.

SX want SS flying to build out starlink quickly and cost effectively. At which point the competition landscape shifts quite substantially.

Time will tell how much longer it will be before SS flies and more importantly returns a stg2 for reuse. Perhaps by the end of the year?

[meekGee]

Nah, there are ways to do it. Stoke’s approach is one way.

Such as?

Instead of starting small and expendable, start small and reusable, starting with the upper stage and doing gradual envelop expansion tests all the way up to simulating a full orbital flight under tether or with shallow hops. Do it for the first stage, too. This is would be a more extreme version of what Stoke has done. I have seen it suggested for first stages mostly.

Note that stoke can probably go a lot faster than spaceX due to the small size of the vehicle.

Late to the party, sorry

Isn't this exactly BO's path?  Vertically integrated too!

Didn't realize it before but it'll be an interesting comparison to make as things progress.

I don't know how small-medium launchers can survive in a future where small-medium payloads have a hard time surviving. It is already tough enough to eek a living as it is  and Starlink/Starshield is going to eat up a lot of things in orbit too.

[Zed_Noir]

.....So we're awaiting A6 with the Vinci engined stg2 at present?

The solid boosters and the Vulcain 2.1 main engine are not quite the same as the ones on the Ariane 5 ECA. So might have teething issues as well with the first few launches.

But, yeah. Arianespace is waiting for the Ariane 6 to get through the integrated system test (aka wet dress rehearsal).

[john smith 19]

.....So we're awaiting A6 with the Vinci engined stg2 at present?

The solid boosters and the Vulcain 2.1 main engine are not quite the same as the ones on the Ariane 5 ECA. So might have teething issues as well with the first few launches.

But, yeah. Arianespace is waiting for the Ariane 6 to get through the integrated system test (aka wet dress rehearsal).

I didn't know about the Vulcain upgrade but I thought the SRB's were aligned with the Vega upgrade?
I'm not a big fan of solids but I am impressed how Vega has really proven the case for all electric TVC. I definitely think it should be in the toolkit of every rocket designer.

However this is OT for the thread so I'll leave it there.

[tbellman]

Refurb costs of both fairings and stages will go up with the number of landings that need to be dealt with.  If that can all be done by the existing workforce all well and good but if you take on new hires then the that's a cost rise.

You are assuming that the personnel doing the refurbishments are just idling when they are not refurbishing.  That is not a safe assumption.  Not necessarily false, but not necessarily true either.  They may be assigned other productive tasks when they have no stages or fairings to refurbish.  Or not.  As far as I know, SpaceX has not said anything about that.

And even if they are idling when not refurbishing, there is no universal standard for how to report that in their accounting.  (That applies not just for personnel costs; there is no universal standard for what should be counted as fixed costs, and what should be counted as variable costs.)

And that is why you can't blindly compare marginal costs; they may have been calculated using different standards.

When I did my little cost modelling game for this SX were saying that the proportion of the launch price of the US came to $17m.

A couple of years ago, Elon tweeted that a Starlink launch cost about $15M.  I don't have the time to locate the tweet, but I remember that it was phrased as that included recover and refurbishment costs for the first stage and for the fairings, but assumed that building the first stage and the fairings were already paid for, and fixed costs were not included either.  But I did get the impression that personnel costs for recovery and refurbishment were included.

In about the same timeframe, he also said that the Falcon 9 upper stage cost about $10M, a new first stage about $30M, and a new fairing pair cost about $6M.

Mass production (by rocket industry standards of "mass production") can lower that a bit, but how much? Perhaps Coastal_Ron could make a suggestion? The standard aircraft industry "learning curve" was a 15% fall in unit price per each doubling of production volume.

A quick Google says they've launched F9 274x, that's 2^7 roughly, which would suggest they've cut costs by 105%.

IOW every new F9 US is making them money  .

That's not how cummulative percentages work.  You should calculate (1-0.15)⁷ ≈ 0.32, which would mean a 68% savings.

But that 15% per doubling sounds like a very rough rule of thumb, that is only true within some specific conditions.  Yes, the longer you produce a product, the cheaper you can typically make it (when correcting for general inflation).  But how quickly you can lower your costs depends on a lot of factors, including where in the learning curve you are.  And in the real world there are often stair-step changes; it is not a continous function.

And something I've never quite been sure about it is how much (or rather how little) what's being made is allowed to change before the learning curve has to be reset. Obviously how it's made can change quite a lot (that's where the benefits come from after all), but how much can what is being made change?

There is no generic answer to that question.  It Depends.

[FutureSpaceTourist]

https://twitter.com/andrewparsonson/status/1717429367121342644

In case you missed it, I published issue 76 of the newsletter yesterday. In this issue, I looked at how European small launch companies view @SpaceX  rideshare mission pricing.

https://europeanspaceflight.substack.com/p/european-small-launch-companies-respond

Lots of quotes from different European small launch companies.

[sebk]

When I did my little cost modelling game for this SX were saying that the proportion of the launch price of the US came to $17m.

Mass production (by rocket industry standards of "mass production") can lower that a bit, but how much? Perhaps Coastal_Ron could make a suggestion? The standard aircraft industry "learning curve" was a 15% fall in unit price per each doubling of production volume.

A quick Google says they've launched F9 274x, that's 2^7 roughly, which would suggest they've cut costs by 105%.

IOW every new F9 US is making them money  . As you can see deciding what the "initial production run" was makes quite a difference if you are trying to figure out a learning curve effect.

Of course those percents do compound, so it's ~0.32 or about 6M 2010 dollars. With inflation this is ~$8.5M today which actually sounds about right

It seems obvious that nobody's rocket launch business plan will close.  All of these plans (small, medium, and large, reusable and expendable) were made when Falcon was launching at most at a rate 1/4th the current rate and fairing recovery was still experimental.  So SpaceX's marginal launch costs were maybe 3x what they are now. 

Same as John Smith, I'd dispute that 3x value. ~30% since Falcon launch rate was 4x less -- that I would agree.

Remember that marginal costs by definition exclude stuff like paying for facilities, and to a large percent salaries (unless you're contracting / hiring just to handle that yet another launch).

We actually got numbers from the times when the launch rate was 1/4 the current one. Musk tweeted $15M for yet another launch (likely marginal cost) while in the similar timeframe SpaceX accidetally released some video from investors meeting (they pulled it in a few hours, but it was too late) and there the total cost of commercial launch was ~$27M. So applying 15% learning curve per doubling combine with inflation we should get ~$12M of marginal cost. And the non-marginal part of the fully burdened cost at the 1st order would be ~1/4 t was back then (they didn't increase Falcon related employment, nor did they add new facilities) so from $12M (since 27 - 15 = 12) to $4M.

So $16M cost of F9 launch now vs $27M few years ago. This is very very rough estimate, of course.

[RedLineTrain]

When I did my little cost modelling game for this SX were saying that the proportion of the launch price of the US came to $17m.

Mass production (by rocket industry standards of "mass production") can lower that a bit, but how much? Perhaps Coastal_Ron could make a suggestion? The standard aircraft industry "learning curve" was a 15% fall in unit price per each doubling of production volume.

A quick Google says they've launched F9 274x, that's 2^7 roughly, which would suggest they've cut costs by 105%.

IOW every new F9 US is making them money  . As you can see deciding what the "initial production run" was makes quite a difference if you are trying to figure out a learning curve effect.

Of course those percents do compound, so it's ~0.32 or about 6M 2010 dollars. With inflation this is ~$8.5M today which actually sounds about right

It seems obvious that nobody's rocket launch business plan will close.  All of these plans (small, medium, and large, reusable and expendable) were made when Falcon was launching at most at a rate 1/4th the current rate and fairing recovery was still experimental.  So SpaceX's marginal launch costs were maybe 3x what they are now. 

Same as John Smith, I'd dispute that 3x value. ~30% since Falcon launch rate was 4x less -- that I would agree.

Remember that marginal costs by definition exclude stuff like paying for facilities, and to a large percent salaries (unless you're contracting / hiring just to handle that yet another launch).

We actually got numbers from the times when the launch rate was 1/4 the current one. Musk tweeted $15M for yet another launch (likely marginal cost) while in the similar timeframe SpaceX accidetally released some video from investors meeting (they pulled it in a few hours, but it was too late) and there the total cost of commercial launch was ~$27M. So applying 15% learning curve per doubling combine with inflation we should get ~$12M of marginal cost. And the non-marginal part of the fully burdened cost at the 1st order would be ~1/4 t was back then (they didn't increase Falcon related employment, nor did they add new facilities) so from $12M (since 27 - 15 = 12) to $4M.

So $16M cost of F9 launch now vs $27M few years ago. This is very very rough estimate, of course.

Yes, I was basing this off of roughly $27 million a couple of years ago (assumed that this was a marginal cost).  Then I have no qualms with your experience curve calculations other than to say that reliable and repeatable fairing reuse is a very recent phenomenon.  You should take $6 million off the top and then run your experience curve.  Or you can do a bottom-up calculation and you would probably arrive at roughly the same amount.  I think it's something less than $10 million, but if you say it's $15 million, I would not argue.

The marginal cost is the primary cost figure of merit related to competition because if push comes to shove, SpaceX can readily price their launches at that amount and happily take on an additional launch.  Falcon 9's marginal cost per kg appears unbeatable in the foreseeable future (medium term) by any of the existing or planned launchers, except maybe Starship.  The money chasing this market is dead money.  Rocket philanthropy at best.

[john smith 19]

You are assuming that the personnel doing the refurbishments are just idling when they are not refurbishing.  That is not a safe assumption.  Not necessarily false, but not necessarily true either.  They may be assigned other productive tasks when they have no stages or fairings to refurbish.  Or not.  As far as I know, SpaceX has not said anything about that.

My actual assumption was there is a refurbishment team. It has a finite capacity of between 1 and X stages (and now fairings). Above that (including overtime) they are maxed out and more staff will need to be hired. 

Like a lot of stuff in this industry it's another step-change  situation where it makes sense to push usage up to the limit, like using the whole capacity of the FSW machines to make F9 and skipping any smaller designs as they simply weren't worth it.

And even if they are idling when not refurbishing, there is no universal standard for how to report that in their accounting.  (That applies not just for personnel costs; there is no universal standard for what should be counted as fixed costs, and what should be counted as variable costs.)

And that is why you can't blindly compare marginal costs; they may have been calculated using different standards.

True, certainly on a global scale.

And also why you should be exceptionally wary when a company changes its accounting report standards.

A couple of years ago, Elon tweeted that a Starlink launch cost about $15M.  I don't have the time to locate the tweet, but I remember that it was phrased as that included recover and refurbishment costs for the first stage and for the fairings, but assumed that building the first stage and the fairings were already paid for, and fixed costs were not included either.  But I did get the impression that personnel costs for recovery and refurbishment were included.

In about the same timeframe, he also said that the Falcon 9 upper stage cost about $10M, a new first stage about $30M, and a new fairing pair cost about $6M.

Which was why these were included as variables that  you could change them to see how that affected profit margin and full recovery of development costs (I think Musk said that was about $1Bn. IIRC with my original numbers the profit was around $20m a launch. With those numbers it's at least $25, and possibly more like $30m

That's not how cummulative percentages work.  You should calculate (1-0.15)⁷ ≈ 0.32, which would mean a 68% savings.

Which sounds impressive. But...

But that 15% per doubling sounds like a very rough rule of thumb, that is only true within some specific conditions.  Yes, the longer you produce a product, the cheaper you can typically make it (when correcting for general inflation).  But how quickly you can lower your costs depends on a lot of factors, including where in the learning curve you are.  And in the real world there are often stair-step changes; it is not a continous function.

It is. That was sort of my point.

The "learning curve" and 15% figure came out of after-the-fact studies of cost for warplanes built by US industry in WWII. I would agree that is a very specific set of circumstances. Essentially it represents all the little mfg and process hacks that were found by the mfgs as they gradually grew production of a particular aircraft.

It would be interesting if a similar study was done for say Liberty Ships (production run of 2000 ships across 18 mfgs) to see how their production methods improved over the run. what that "learning curve" would be.

AFAIK it's still the goto figure that production engineers use when Management asks "So once you get used to making <new product X> what do you think you can lower the production costs to?"

One factor that has changed is the amount of routine automation. I recall reading somewhere that the learning curve for the F15 was flatter than 15% because so much of the processes had already been automated.

There is little room to improve the amount of re-work if parts are mfg with high accuracy and process conditions are maintained in tight tolerances so scrappage is low to begin with. Serious cost reduction now takes serious re-structuring of the whole mfg sequence.

There is no generic answer to that question.  It Depends.

Exactly my point.

We know the stg2 on an F9 launching in 2023 is different from that in 2010, although the materials are (probably) mostly the same.

The joker is to what extent have those design changes offset the mfg cost reductions produced by lessons learned? How far along the learning curve is the current version?

I suspect only a handful of people in SX have the overview necessary to answer that question.

[TrevorMonty]

Good interview with Dan from ABL. Listen about halfway on what he said on rideshare.

This is first time I listen to anybody from ABL, always been bit of dark horse. I was impressed by him.

[Blackjax]

Good interview with Dan from ABL. Listen about halfway on what he said on rideshare..

I agree, the video is worth watching just for the nuanced discussion of rideshare alone.  I feel like I gained more perspective rather than just hearing a rehash of stuff I already knew.

[deltaV]

The 70+ flights that ABL has apparently sold so far demonstrates the existence of a market for non-ride-share small launch much better than ABL's CFO's opinion can. (Unless those launches don't actually involve any commitment or money by the buyers, which I can't rule out.) Rocket Lab Electron and Firefly Alpha also have significant manifests. I'd guess that a few small launchers will survive competition from each other and from 20+ tonnes to LEO launchers including Starship but small launchers may not survive competition from 2-20 tonnes to LEO fully reusable launchers such as Stoke's Nova, especially if the small launchers don't go reusable themselves.

[M.E.T.]

Happy to hear your estimate of the marginal launch costs at a 100/yr launch rate.  Let's say $10 million?  $15 million?  It's got to be much less than two years ago at a 30/yr launch rate and no fairing reuse.

It should be fairly obvious that the floor price will be set by the price of a stg2, since that is the biggest non-reusable component.

When I did my little cost modelling game for this SX were saying that the proportion of the launch price of the US came to $17m.

Mass production (by rocket industry standards of "mass production") can lower that a bit, but how much? Perhaps Coastal_Ron could make a suggestion? The standard aircraft industry "learning curve" was a 15% fall in unit price per each doubling of production volume.

SpaceX have never said that an upper stage costs $17M. They have said it costs about $10M, with the fairings a further $6M if not reused. Fairing reuse has now been mastered. So upper stage plus reused fairings would likely be less than $12M, assuming ~$1M per refurbished fairing half, which I think is quite generous.

[trimeta]

The 70+ flights that ABL has apparently sold so far demonstrates the existence of a market for non-ride-share small launch much better than ABL's CFO's opinion can. (Unless those launches don't actually involve any commitment or money by the buyers, which I can't rule out.) Rocket Lab Electron and Firefly Alpha also have significant manifests. I'd guess that a few small launchers will survive competition from each other and from 20+ tonnes to LEO launchers including Starship but small launchers may not survive competition from 2-20 tonnes to LEO fully reusable launchers such as Stoke's Nova, especially if the small launchers don't go reusable themselves.

The 70+ flights in ABL's manifest includes 58 flights booked by Lockheed Martin...which has a "strategic investment" in ABL. I'll let you decide how real those 58 launches are.