Insurance rates for different launch providers

[john smith 19]

Indeed. People seem to forget that Ariane 5 suffered two (2) complete failures (Ariane 501 and 517) as well as two (2) partial failures (Ariane 502 and 510). The two partial failures left several payloads in unusable orbits and resulted in insurance companies paying for the "lost" payloads (A payload does not have to be destroyed in a catastrophic launch mishap to be a total loss. Ending up in a useless orbit is every bit as much a total loss).


Additionally, for Proton:

Past 20 years of service: 13/162 = 8,0%
Past 10 years of service: 10/85 = 11,8%

So, the failure rate of Proton shot up in the last 10 years. A failure rate of nearly 12% in the last decade fits nicely with current insurance rate (12%) for Proton.

Some interesting numbers, but there are also some unmentioned assumptions in them.

I think the key one is the extent to which Ariane 5 and F9 change over time.

AIUI Arianespace is basically a sales and services operation. It does no development work. So the A5 design will have been changed on the results of the MIB investigations for those 4 events but is otherwise pretty stable.

OTOH F9 has been continually changed since first launch. Merlin thrust levels have gone through what 4 levels? Engine mounts shifted to Octaweb and IIRC at least one tank stretch.

These are only the visible changes.  I can't imagine how many versions the flight control software has gone through.

The point is every change is a possibility for failure and every so often SX has gotten it wrong as AMOS 6 demonstrated. 

On that basis Ariane 5 has clocked up 78 successful flights with (AFAIK) the same configuration, for a failure rate < 1 in 78 IE < 1.28%

F9 has clocked up 23 launches of F9/FT or 2 of the F9/FT/Blk4 version (according to Wiki).
Or 15 since AMOS 6 (whatever version). So is that  a failure rate <4.34%, <6.66% or 50%  all flown ?
And this disregards which of those boosters has already been flight proven.

I'm not saying either approach is good or bad. I'm saying the insurance rates don't seem to match the actual risk profiles of the companies involved.

IOW are insurers discounting the effect of reusability?  Are they under weighting the frequent changes? Or are Arianespace being over charged (until they have another flight failure)?

[john smith 19]

Unsure how the program is managed, but if I were in charge, a 2nd complete telescope would be ordered as spare. Or at least the critical parts, one-offs, that are hard to remake, like custom sensors, optics, electronics, etc.

There are precedents. The book on NASA computers said the 2nd Voyager was quite a bit cheaper than the first.

That said some stuff is so eyewateringly expensive you really want to think hard about wheather you need a spare (Beryllium is much lighter and stiffer than Aluminum.  IIRC anything Beryllium is about 200x the cost of something in Aluminum). 

[woods170]

Indeed. People seem to forget that Ariane 5 suffered two (2) complete failures (Ariane 501 and 517) as well as two (2) partial failures (Ariane 502 and 510). The two partial failures left several payloads in unusable orbits and resulted in insurance companies paying for the "lost" payloads (A payload does not have to be destroyed in a catastrophic launch mishap to be a total loss. Ending up in a useless orbit is every bit as much a total loss).


Additionally, for Proton:

Past 20 years of service: 13/162 = 8,0%
Past 10 years of service: 10/85 = 11,8%

So, the failure rate of Proton shot up in the last 10 years. A failure rate of nearly 12% in the last decade fits nicely with current insurance rate (12%) for Proton.

Some interesting numbers, but there are also some unmentioned assumptions in them.

I think the key one is the extent to which Ariane 5 and F9 change over time.

AIUI Arianespace is basically a sales and services operation. It does no development work. So the A5 design will have been changed on the results of the MIB investigations for those 4 events but is otherwise pretty stable.

OTOH F9 has been continually changed since first launch. Merlin thrust levels have gone through what 4 levels? Engine mounts shifted to Octaweb and IIRC at least one tank stretch.

These are only the visible changes.  I can't imagine how many versions the flight control software has gone through.

The point is every change is a possibility for failure and every so often SX has gotten it wrong as AMOS 6 demonstrated. 

On that basis Ariane 5 has clocked up 78 successful flights with (AFAIK) the same configuration, for a failure rate < 1 in 78 IE < 1.28%

F9 has clocked up 23 launches of F9/FT or 2 of the F9/FT/Blk4 version (according to Wiki).
Or 15 since AMOS 6 (whatever version). So is that  a failure rate <4.34%, <6.66% or 50%  all flown ?
And this disregards which of those boosters has already been flight proven.

I'm not saying either approach is good or bad. I'm saying the insurance rates don't seem to match the actual risk profiles of the companies involved.

IOW are insurers discounting the effect of reusability?  Are they under weighting the frequent changes? Or are Arianespace being over charged (until they have another flight failure)?

Ariane 5 has changed more over the years than you indicate:
- Two different versions of the core stage main engine (Vulcain 1 and Vulcain 2)
- Continuous improvements made to the core stage main engines.
- Two different versions of the core stage (EPC H158 and EPC H173)
- Two different versions of the solid rocket boosters (EAP 238 and EAP 241).
- Two completely different upper stages. (EPS and ESC-A)
- Two different versions of one of the upper stage types. (EPS L9.7 and EPS L10)
- Multiple changes in core stage tank insulation
- Multiple changes to avionics hardware (replacement of obsolete and out-of-production with newly developed)
- Switch from fairing with acoustics dampening to one without
- Multiple improvements and changes in production of tankage and other structures.
- Smaller changes to hundreds of other details over the Ariane 5 life-span.

Don't assume a current Ariane 5 ECA to be identical to one from 10 years ago. There are lots of (albeit small) changes.

[Rik ISS-fan]

In my opinion the difference in insurance rate between Ariane 5 and Falcon 9 is very small.
Ariane 5 ECA hasn't had a failure or partial failure for about 15 years. [https://csglaunches.eu/] only the first ECA failed.  There have been four improvements to A5 ECA, they are developed under the LEAP program.
They first do ground test (for example ARTA6, last year) and than a test launch at discounted cost (institutional launch).

With SpaceX it is not known what they have changed to the launcher or launch procedure.
Using Super Cryo with composite wrapped GHe tanks is looking for trouble. Let's try filling the He later and faster => Kaboom. A failure mode that was totally to be expected. I hope this has changed at SpaceX.

[john smith 19]

Ariane 5 has changed more over the years than you indicate:
- Two different versions of the core stage main engine (Vulcain 1 and Vulcain 2)
- Continuous improvements made to the core stage main engines.
- Two different versions of the core stage (EPC H158 and EPC H173)
- Two different versions of the solid rocket boosters (EAP 238 and EAP 241).
- Two completely different upper stages. (EPS and ESC-A)
- Two different versions of one of the upper stage types. (EPS L9.7 and EPS L10)
- Multiple changes in core stage tank insulation
- Multiple changes to avionics hardware (replacement of obsolete and out-of-production with newly developed)
- Switch from fairing with acoustics dampening to one without
- Multiple improvements and changes in production of tankage and other structures.
- Smaller changes to hundreds of other details over the Ariane 5 life-span.

Don't assume a current Ariane 5 ECA to be identical to one from 10 years ago. There are lots of (albeit small) changes.

TBH I was aware of the Vulcain upgrade but had forgotten it. 

I was unaware of most of the rest of those changes. Thank you for listing them. 

So the other question would seem to be have these changes been an ongoing thing (with nearly every launch being just a little bit different from every other) or were most of them done following a failure, or partial failure?

I guess what I'm saying is that the scale of changes still seems smaller than that of F9 and their success record is still definitely better, so (in the context of this thread) is SX getting a very good deal on its insurance, or is Arianespace getting rather a bad deal on its launches?

Somehow the facts don't quite add up. 

[john smith 19]

In my opinion the difference in insurance rate between Ariane 5 and Falcon 9 is very small.

Yes it does, which doesn't seem a true reflection of A5 unbroken launch success record (78) versus F9 (15 since AMOS 6).

With SpaceX it is not known what they have changed to the launcher or launch procedure.

True, although presumably the insurance companies who take the contract have somewhat greater insight than we do.

Using Super Cryo with composite wrapped GHe tanks is looking for trouble. Let's try filling the He later and faster => Kaboom. A failure mode that was totally to be expected. I hope this has changed at SpaceX.

That's a bit harsh.  If it really was that predictable they'd have know it was not worth doing in the first place. 

OTOH doing with the payload on the top was an unnecessary risk. I doubt SX will be doing that again when they have any ConOps tweaks planned.

[kenny008]

Ariane 5 has changed more over the years than you indicate:
- Two different versions of the core stage main engine (Vulcain 1 and Vulcain 2)
- Continuous improvements made to the core stage main engines.
- Two different versions of the core stage (EPC H158 and EPC H173)
- Two different versions of the solid rocket boosters (EAP 238 and EAP 241).
- Two completely different upper stages. (EPS and ESC-A)
- Two different versions of one of the upper stage types. (EPS L9.7 and EPS L10)
- Multiple changes in core stage tank insulation
- Multiple changes to avionics hardware (replacement of obsolete and out-of-production with newly developed)
- Switch from fairing with acoustics dampening to one without
- Multiple improvements and changes in production of tankage and other structures.
- Smaller changes to hundreds of other details over the Ariane 5 life-span.

Don't assume a current Ariane 5 ECA to be identical to one from 10 years ago. There are lots of (albeit small) changes.

TBH I was aware of the Vulcain upgrade but had forgotten it. 

I was unaware of most of the rest of those changes.

So the other question would seem to be have these changes been an ongoing thing (with nearly every launch being just a little bit different from every other) or were most of them done following a failure, or partial failure?

I guess what I'm saying is that the scale of changes still seems smaller than that of F9 and their success record is still definitely better, so (in the context of this thread) is SX getting a very good deal on its insurance, or is Arianespace getting rather a bad deal on its launches?

Somehow the facts don't quite add up. 

I would imagine that the insurance premiums at some point asymptotically reach some number, and I would also assume that this number is not zero.  As you get more and more flights, I would not expect that the rates will continue to go down to zero.  Therefore, as both launchers continue to fly, both of their rates should become pretty much the same.  When does this happen?  Have Ariane rates pretty much bottomed out, and F9 are just about there?

I'm also not sure what the real question is.  Are we looking for evidence that insurance companies have made some backroom decision to cut SX a deal and potentially lose money on future launch failures?  I don't think I can see the motivation for insurance companies to keep the rates for any particular launcher artificially low.

[abaddon]

I'm not saying either approach is good or bad. I'm saying the insurance rates don't seem to match the actual risk profiles of the companies involved.

The simple answer is they are professionals who have a better understanding of the actual risk profiles than you (or I) do.

It's easy to forget, but SES has a team embedded with SpaceX sitting right there.  NASA and the USAF have very deep insight as well.  They all know way more than we do, to base their assessments on.

[Lars-J]

I was unaware of most of the rest of those changes.

So the other question would seem to be have these changes been an ongoing thing (with nearly every launch being just a little bit different from every other) or were most of them done following a failure, or partial failure?

JS19, All rockets go through a lot of changes. ULA does the same thing with the Atlas V. I doubt there has been many Atlas V cores that have been identical in every aspect - if any.

SpaceX is more open about its changes, and they are sometimes more visible. And there is also certain group of actors that do their best to highlight SpaceX changes and downplay changes to their own vehicles.

[john smith 19]

It's easy to forget, but SES has a team embedded with SpaceX sitting right there.  NASA and the USAF have very deep insight as well.  They all know way more than we do, to base their assessments on.

How is that relevant exactly? It's the insurance companies that set the rates on payloads. Do they have a team "embedded" with SX, or any LV supplier? that would be more relevant to this discussion.

JS19, All rockets go through a lot of changes. ULA does the same thing with the Atlas V. I doubt there has been many Atlas V cores that have been identical in every aspect - if any.

I never doubted it.   

Except some do seem to go a lot longer than others before one of their vehicles goes bang. Russians, for example seemed to manage 100s of launches without a RUD.

SpaceX is more open about its changes, and they are sometimes more visible. And there is also certain group of actors that do their best to highlight SpaceX changes and downplay changes to their own vehicles.

No doubt. However the simple numbers suggest something is wrong in the market. 

I would imagine that the insurance premiums at some point asymptotically reach some number, and I would also assume that this number is not zero.  As you get more and more flights, I would not expect that the rates will continue to go down to zero.  Therefore, as both launchers continue to fly, both of their rates should become pretty much the same.  When does this happen?  Have Ariane rates pretty much bottomed out, and F9 are just about there?

I'm also not sure what the real question is.  Are we looking for evidence that insurance companies have made some backroom decision to cut SX a deal and potentially lose money on future launch failures?  I don't think I can see the motivation for insurance companies to keep the rates for any particular launcher artificially low.

So let's see an ELV with 92 flights, the last 78 successful, has about the same insurance rate as one which has flown 43 times, the last 15 successful.

My point is either SX has had a remarkably fast drop in rates (to 4% after 15 launches since RTF), given the last claim by a payload operator was only 15 launches ago, or the insurance industry has been very slow to recognize any  benefit to SX in having a flight proven booster (which is known to work and whose resistance to flight wear has actually been measured after flight, and which incorporates 9 of the 10 engines and 1/2 the stages of the whole vehicle)

As SOP for SX launches is to move to "flight proven" boosters shouldn't those rates go down, as every launch (from the same set of hardware) cease to be "unique" IE expendable,  but one of a series, whose performance can be judged against its performance during previous launches with the same hardware?

It's an anomaly, but I cannot say which direction the anomaly is.

It's not a question, it's an observation.

The "question" your looking for is "why" ?

[LouScheffer]

But why should premiums for a year cover a launch failure?  Ariane 5 has not failed for about 15 years.   Therefore a break-even premiums for year  would only cover 1/15 of a launch failure.  As of now, the insurance companies have collected 80 straight premiums for Ariane, without a single payout.  At 4%, that's about 3.2 payloads worth in the bank, not including what they've earned by investing the money over that time.  They could afford to lose one or two payloads and still be solidly in the black.  If anything, insurance for Ariane seems way more expensive than the statistics dictate.

I believe the insurance talked about here is for launch and first year on-orbit.

No idea if that changes the numbers.

This appears to be the case, and changes the numbers by a lot, and then they make much more sense.  From Falling Satellite Insurance Premiums Put Market at Risk of Major Upheaval, at least in 2012, the payouts from satellites failing in the first year were more than the risk of launch failure.

So if the insurer needs 2% to cover satellite failures, then the total for Ariane might be something like 3.5%, so 1.5% for launch, representing a 1/66 odds of failure.   For SpaceX, perhaps 5%, so 3% for launch, for a perceived failure rate of 1 in 33.   For Proton, perhaps 2% satellite + 8% launch (1 in 12 fails) for 10% total.

So adding a few percent for satellite failure makes the premiums line up with the observed failure rates much better.

[Mike Jones]

Insurance rate do not directly follow statistical success rates but they reflect the level of trust from insurance and reinsurance companies (mainly western entities whose employees read the same enthusiastic press articles about the future of space enabled by reusability). SpaceX benefits from the great hype (well deserved overall) around their company and new space in general to have similar insurance rate than Arianespace or ULA despite 2 major failures in 2 years.

[Rik ISS-fan]

Ariane 5 has changed more over the years than you indicate:
- Two different versions of the core stage main engine (Vulcain 1 and Vulcain 2)
- Two different versions of the core stage (EPC H158 and EPC H173)
- Two different versions of the solid rocket boosters (EAP 238 and EAP 241).

These were changes applied at P2

- Two different versions of one of the upper stage types. (EPS L9.7 and EPS L10) Difference G and GS/ES

- Continuous improvements made to the core stage main engines.
- Multiple changes in core stage tank insulation
- Multiple changes to avionics hardware (replacement of obsolete and out-of-production with newly developed)
- Switch from fairing with acoustics dampening to one without
- Multiple improvements and changes in production of tankage and other structures.
- Smaller changes to hundreds of other details over the Ariane 5 life-span.

These are implemented at block upgrades. And as I write in my reply this is tested before implementation in scope of LEAP. Most are applied because the materials are no longer allowed or are no longer available.

Using Super Cryo with composite wrapped GHe tanks is looking for trouble. Let's try filling the He later and faster => Kaboom. A failure mode that was totally to be expected. I hope this has changed at SpaceX.

That's a bit harsh.  If it really was that predictable they'd have know it was not worth doing in the first place. 

OTOH doing with the payload on the top was an unnecessary risk. I doubt SX will be doing that again when they have any ConOps tweaks planned.

The first thing teachers told me about epoxy / polymers is that they are permeable, so LOx gets into it.
So my oppinion is that this was totally to be expected. Very bad risk judgment from SpaceX, or ...

[john smith 19]

Ariane 5 has changed more over the years than you indicate:
- Two different versions of the core stage main engine (Vulcain 1 and Vulcain 2)
- Two different versions of the core stage (EPC H158 and EPC H173)
- Two different versions of the solid rocket boosters (EAP 238 and EAP 241).

These were changes applied at P2

- Two different versions of one of the upper stage types. (EPS L9.7 and EPS L10) Difference G and GS/ES

- Continuous improvements made to the core stage main engines.
- Multiple changes in core stage tank insulation
- Multiple changes to avionics hardware (replacement of obsolete and out-of-production with newly developed)
- Switch from fairing with acoustics dampening to one without
- Multiple improvements and changes in production of tankage and other structures.
- Smaller changes to hundreds of other details over the Ariane 5 life-span.

These are implemented at block upgrades. And as I write in my reply this is tested before implementation in scope of LEAP. Most are applied because the materials are no longer allowed or are no longer available.

So you're saying that while A5 has had fairly extensive changes they were done en-mass as a block upgrade, and the design cut over to the new version after that? IOW although there were many individual changes because they were done all at once there is actually only 1 or 2 "change events" ?

Using Super Cryo with composite wrapped GHe tanks is looking for trouble. Let's try filling the He later and faster => Kaboom. A failure mode that was totally to be expected. I hope this has changed at SpaceX.

That's a bit harsh.  If it really was that predictable they'd have know it was not worth doing in the first place. 

OTOH doing with the payload on the top was an unnecessary risk. I doubt SX will be doing that again when they have any ConOps tweaks planned.

The first thing teachers told me about epoxy / polymers is that they are permeable, so LOx gets into it.
So my oppinion is that this was totally to be expected. Very bad risk judgment from SpaceX, or ...

One option would be they didn't have  your teachers.   

More seriously one thing I've noticed about aerospace engineering is there is a lot of "lore" about what you can and can't do.

A lot of this evolved during the crash R&D programmes of the lat 50's post Sputnik, when the US was desperate to get something off the ground. If something proved (or in fact even looked) problematical then it was a case of "here be monsters."

Hence the pervasive belief that it is impossible to cool a combustion chamber with LOX (despite nearly all rockets that don't use LOX as an oxidizer doing so) and in fact NASA testing such a 40 000lb engine (including a deliberate leak inside the chamber itself) in the early 90's.

So any "rules of thumb" should be treated with extreme caution and tested.  BTW HMX (who's company also tested LOX cooling) pointed out that composite LOX tanks have also been constructed decades ago.

Obviously SX thought the reward outweighed the risk. In the short term it did not. In the long run they may have been able to find something that has improved their ConOps and/or their costs.

[ZachF]

But why should premiums for a year cover a launch failure?  Ariane 5 has not failed for about 15 years.   Therefore a break-even premiums for year  would only cover 1/15 of a launch failure.  As of now, the insurance companies have collected 80 straight premiums for Ariane, without a single payout.  At 4%, that's about 3.2 payloads worth in the bank, not including what they've earned by investing the money over that time.  They could afford to lose one or two payloads and still be solidly in the black.  If anything, insurance for Ariane seems way more expensive than the statistics dictate.

I believe the insurance talked about here is for launch and first year on-orbit.

No idea if that changes the numbers.

This appears to be the case, and changes the numbers by a lot, and then they make much more sense.  From Falling Satellite Insurance Premiums Put Market at Risk of Major Upheaval, at least in 2012, the payouts from satellites failing in the first year were more than the risk of launch failure.

So if the insurer needs 2% to cover satellite failures, then the total for Ariane might be something like 3.5%, so 1.5% for launch, representing a 1/66 odds of failure.   For SpaceX, perhaps 5%, so 3% for launch, for a perceived failure rate of 1 in 33.   For Proton, perhaps 2% satellite + 8% launch (1 in 12 fails) for 10% total.

So adding a few percent for satellite failure makes the premiums line up with the observed failure rates much better.

This seems the most likely scenario.

Would also be interesting to see if there is any difference in insurance rates between satellite manufacturers as well.

[Rik ISS-fan]

These are implemented at block upgrades. And as I wrote in my reply, this is tested before implementation in scope of LEAP. Most are applied because the materials are no longer allowed or are no longer available.

So you're saying that while A5 has had fairly extensive changes they were done en-mass as a block upgrade, and the design cut over to the new version after that? IOW although there were many individual changes because they were done all at once there is actually only 1 or 2 "change events" ?

Yes.
I think they are now at version 10 or 11. (G, G+, GS, ES [3x], ECA{4|5x}. It's unfortunate that ESA removed the pages about the G, G+ and GS versions.
They build the launchers in blocks of the same configuration, although there are several performance options.
Arianespace announced that they are going to move away from block orders. But I think that they will implement the changes late. They are really conservative. If I'm not mistaken the Vulcain 2.1 has been performing bench tests from 2014. Possibly they will implement it in the PC batch, but Ariane 6 could also be the first use.
The difference between Vulcain 2 and Vulcain 2.1 is the production method for the nozzle.
(and a lot of other tweeks most likely)
In my oppinion the move form A5 ECA to A6 is roughly the same as the move from A5 G to ECA.

If I'm not mistaken SpaceX is at least at version 7 of Falcon 9. (F9; F9 v1.1; F9 FT (five versions)).

[ChrisWilson68]

If I'm not mistaken SpaceX is at least at version 7 of Falcon 9. (F9; F9 v1.1; F9 FT (five versions)).

I believe you are mistaken.  When they say "Falcon 9 Block 5", the reason it's a 5 is that the first Falcon 9 corresponds to block 1, v1.1 corresponds to block 2, and full thrust corresponds to block 3.

Of course, there are more minor changes than the block numbering implies, but by that count the number of versions is far more than seven.

[whatever11235]

If I'm not mistaken SpaceX is at least at version 7 of Falcon 9. (F9; F9 v1.1; F9 FT (five versions)).

I believe you are mistaken.  When they say "Falcon 9 Block 5", the reason it's a 5 is that the first Falcon 9 corresponds to block 1, v1.1 corresponds to block 2, and full thrust corresponds to block 3.

Of course, there are more minor changes than the block numbering implies, but by that count the number of versions is far more than seven.

I wish spacex would adopt semantic versioning (http://semver.org/) for its vehicles. This "1.1" mixed with "block 3/4/5 system" is very confusing.

[ChrisWilson68]

The first thing teachers told me about epoxy / polymers is that they are permeable, so LOx gets into it.
So my oppinion is that this was totally to be expected. Very bad risk judgment from SpaceX, or ...

SpaceX has been using COPV helium tanks inside LOX tanks since day one, so this kind of simplistic "epoxy and LOX don't mix" argument is just silly.  SpaceX certainly knew about the issues, and thought they had ways to safely handle all of them.

For the most part, SpaceX has been very successful doing what you deride.  Even with sub-cooled LOX, after they discovered the specific circumstances that led to that one failure, they were able to change their procedures and make it work.  All the launches since then have successfully used sub-cooled LOX with those same COPV helium tanks.

[woods170]

These are implemented at block upgrades. And as I wrote in my reply, this is tested before implementation in scope of LEAP. Most are applied because the materials are no longer allowed or are no longer available.

So you're saying that while A5 has had fairly extensive changes they were done en-mass as a block upgrade, and the design cut over to the new version after that? IOW although there were many individual changes because they were done all at once there is actually only 1 or 2 "change events" ?

Yes.
I think they are now at version 10 or 11. (G, G+, GS, ES [3x], ECA{4|5x}. It's unfortunate that ESA removed the pages about the G, G+ and GS versions.
They build the launchers in blocks of the same configuration, although there are several performance options.
Arianespace announced that they are going to move away from block orders. But I think that they will implement the changes late. They are really conservative. If I'm not mistaken the Vulcain 2.1 has been performing bench tests from 2014. Possibly they will implement it in the PC batch, but Ariane 6 could also be the first use.
The difference between Vulcain 2 and Vulcain 2.1 is the production method for the nozzle.
(and a lot of other tweeks most likely)
In my oppinion the move form A5 ECA to A6 is roughly the same as the move from A5 G to ECA.

It is in fact much more substantial than you assume:
- New tankage manufacturing methods to substantially lower production costs.
- Different length of tankage compared to A5.
- New fairing manufacturing method to substantially lower production costs.
- Increased fairing length (optional).
- Completely new thrust-frame for core stage engine.
- Completely new forward thrust-take-out skirt.
- Completely new solid rocket boosters, based on Vega-C first stage. Composite casing in stead of steel casing.
- Compared to the ESC-A upper stage of A5 the upper stage of A6 is completely new, with a new engine, new thrust-frame, new tankage, new avionics and new insulation.  Compared to that the ESC-A upper stage of A5 was really just a mash-up of the Ariane 4 upper stage using the original A4 thrust-frame and LOX tank combined with a new LH2 tank.
- Vulcain 2.1 engine features radically re-designed nozzle extension as well as re-designed turbines and increased chamber pressure.
- Completely new launchpad and associated infrastructure.