TERRAN — USA, 2023

Relativity Space was a venture I was ignoring, until they signed a launchpad lease at Cape Canaveral, and let it be known that they had already logged a lot of engine firing time on the test stand. They are taking an approach similar to a number of other New Space rockets, in that they have a bottom stage with multiple “Aeon 1” engines (nine) and a second stage with a single vacuum-bell version of the same engine. Unlike any of the others, they’re using methane as a fuel right from the beginning — something that other companies have considered only after they’d gotten themselves well established. To save a bit of money, they use liquid natural gas rather than purified methane. If they succeed they might still have the first methalox rocket to reach orbit, though the LandSpace-2 was the first to attempt it, and their own attempt a few months later failed similarly — in both cases the first stage worked but the second did not. It appears that their novel ignition system, which is apparently “passive” (some kind of caralyst?) struggles in vacuum, or the trouble may have been a gas bubble in one of the pumps. But they counted it as a success just to get their first try through max Q (the point of greatest stress from air resistance), which is the right attitude to have.

visionary plans

But that’s not what makes them really stand out. The unique thing about the Terran 1 is that they set out to 3D-print the entire rocket. Lots of other companies use 3D printing to make complex engine parts, but nobody else is considering printing the fuselage, fuel tanks, and so on. At first blush, the idea seems ridiculous. A printed material is never going to be competitive on, for instance, optimizing the strength to weight ratio for tank walls. But they say that by printing everything, they can make an Aeon engine with fewer than 100 parts, and a whole rocket with fewer than 1000. (Most traditional rockets have tens or even hundreds of thousands of parts.) In fact, they have said that some versions of their engine have just three parts.

That’s only the beginning of their ambition. The reason they wanted to 3D print the entire thing, even if it didn’t result in optimal properties, was so the process could be automated from beginning to end, so they could make a rocket by pushing one button, with no skilled labor required. It would still take weeks for the build to be completed, but that should certainly achieve some reduction in costs.

We're not done. It can’t actually save money to require no human workers at all — there are diminishing returns there, for sure — so why did they want to do it? Well, so the entire factory can be dropped on Mars, and build rockets there. That was the end goal of this odd approach. It’s also why they picked methane as the fuel. Maybe when they start building it there, they’ll call it the Martian 1 instead of the Terran 1.

This begs the question of what raw materials the printer would use. They mentioned nickel, so it sounds like they might plan to make a lot of parts out of the same inconel alloy which is widely used for the high temperature parts of engines. Inconel mixes are generally at least half nickel, sometimes three quarters, and nickel should be reasonably easy to find on Mars. The crucial second ingredient of inconel is chromium... and it looks like this may also be fairly abundant. Aluminum and titanium are not rare either, and their tanks and frame are mostly aluminum.

Their giant superduper 3D printer is named Stargate. For the large parts such as tanks, it rotates them on a turntable as it builds up the metal on the top edge. (The results look rather rough and grainy in the video they’ve released so far.) Apparently it not only incorporates some proprietary metallurgy, but also some kind of artificial inteligence features, to help make it autonomous. They claim it will learn to build faster as it gains experience. They are now happy enough with the prototype that they are building several more printers.

Their first rocket, which they claimed would in the future be scaleable to different sizes, is fairly small: bigger than an Electron but not bigger than much else. They claimed they would sell them for $10 million each. I figured that the engine itself would probably have to scale in numbers, rather than size, as it uses an expander cycle — a rare choice for a booster engine, but a good fit for 3D printing, because it uses complex plumbing but does not make severe demands on the materials. From pictures, it looks like it’s a semi-closed design which dumps the unburned turbine propellant into the bell... and does it with a single fat pipe, rather than a ring of little holes such as most people would use for the purpose. Besides trying to be the first methalox rocket to orbit, they were also in a race to be the first to launch with an expander-bleed booster. They lost the methalox race to China’s LandSpace 2, and then the expander-cycle booster race to Japan’s H3, even though both of those rockets needed two attempts to succeed.

But then they said they plan to make a much larger first stage engine once they get to orbit with the little Aeon 1 — a gas generator. This would allow their small rocket to have just one booster engine, thereby reducing the part count further. This engine would be called Aeon R. It would also allow them to build a large rocket, at least as big as a Falcon or Vulcan, called the Terran R. And yes, the R stands for reusable... and in order to compete on cost, they also said they wanted to go for second stage reuse. The thing might end up looking like a mini Starship... but not all that mini, as in their latest designs the thing would be at least as big as a Vulcan, able to lift over twenty tons with reuse, or thirty without.

So why go for nine little engines on the first rocket, when they say it was “definitely not the optimum choice in hindsight”? Well, they say it’s because this makes it a better pathfinder toward a big rocket, but other companies have tried to make two engines in different sizes and then found that getting just the little one to work was the limit of their capabilities.

This vision has somehow attracted tons of investor capital — a lot more than many of their competitors. It’s also attracted customers to contract for flights on rockets they haven’t even started any visible work on yet — OneWeb, for instance, wants to hire the promised Terran R. But as with Virgin Orbit, it’s questionable how it will be possible to make back all that money they’re spending.

walking it back

In 2023, they announced that they were going all in on the Terran R, and that they would not even try another shot with the Terran 1. I consider this a bad sign, but an approach like this seems to be working out okay for LandSpace, so maybe they’ll avoid ending up like Astra.

Another thing they announced was essentially an admission that it was stupid to make fuel tanks with 3D printing. For the Terran R they will make the cylindrical tank parts out of conventional sheet aluminum... which begged the question of how they plan to join the printed end caps to the sheet walls. By 2025 they had given up on that and started buying conventional tank domes from an outside supplier. So all the visionary guff about 3D printing entire rockets has now come to nothing, with their unique fabrication technology not being used for much more than how other companies use 3D printing.

They have a design for the Terran R which looks rather Falconlike, with four legs and four grid fins, but with thirteen engines. Like the Falcon they plan to land at sea, but unlike it they are designing the booster to come in sideways through the upper atmosphere, with strakes running up the booster’s flanks to give it lift and braking.

They also admitted that going for second stage reuse was pie in the sky.

In 2025 they changed CEOs. They got Eric Schmidt, who was CEO of Google from 2001 to 2011, and executive chairman after that — a guy who could probably do rocket-company CEOing as a weekend hobby. Correction: they didn’t get him, he got them. He’d been the guy funding them for the previous six months, so his availability as CEO was an offer they couldn’t refuse. I figured the company would probably become a lot more boringly sensible and realistic now, though probably no more likely than it was before to start doing fifty-plus launches a year, as they still say they are planning to. But nope — looks like they may be going in a direction that’s more hare-brained than ever. If the stories are true... well, I have to go on a bit of a rant here.

Reportedly, Schmidt’s business plan is to put data centers in orbit. If your initial reaction is to wonder aloud how the fuck anyone could think that idea makes any goddamn sense, technically or financially, you’re not alone. Apparently Schmidt is floating a plan to build data centers so huge that putting them in inhabited areas is a real problem — centers whose intake of power and outgive of heat, which result in shortages of water and longages of angry neighbors, make them untenable to locate in any spot where building them would be affordable. But in space these limits don’t apply, so by making them many times more unaffordable, their capacity can expand without restraint. Brilliant!

So what would these hyperhuge data centers do? AI, that’s what. So yeah, I don’t think this is a serious attempt to solve a technical problem... it’s an instance of the crap that Silicon Valley always pulls nowadays, where some buzzword catches on and becomes the inescapable topic of the year, and countless hopeful investors and VC players all start hoping it’s the next big breakthrough that will disrupt something or other and create new billionaires, so anyone doing adjacent work gets funded, and funding produces progress which increases hype which attracts more funding, and smooth talkers see the money and start lying about what they’re going to be able to accomplish with it, and what might have started as just a vague science-fictional concept inflates itself into a crazed market bubble. AI, having broken through to a level where it started to be genuinely useful for people who value convenience more than correctness, made itself one of the biggest such bubbles in decades, and those trying to make the hype come true are getting used to expecting infinite money. That expectation is the only reason I can think of for anyone to believe that this vision of giant flying robot brains could be built.

The AI mogul Sam Altman also wants to put up orbital data centers, and made some kind of offer to Stoke Space to fund or buy them in pursuit of that goal, but to their credit they didn’t take the deal.

And the thing is, a smaller version won’t work at all. Pilot or demonstration versions can only lose money; only at an absurdly enormous scale that no one else can match would it have a chance of some kind of market victory. And by then, because its solar panels and heat radiators make it an enormous target, it may well have pushed its luck too far and run afoul of a stray orbiting object, leading to a chain reaction of satellite destruction — a possibility that becomes more likely every time another me-too “swarm” starts getting launched.

So yeah, to me it sounds like the alleged plan is just Ponzi bait, and it gives them so many technical challenges that getting the rocket to work might not be their first priority. If they stick to this plan and don’t clarify in some way that they intend to do normal rocket work first, I’ll want to move this page straight into the “Failed Plans” section while they’re still actively spending money on developing it.

Oh great, now Elon (who has been a major hyper of the AI bubble before it even got going) wants SpaceX to also make flying robot brains. And Google and Amazon, who are probably the two biggest data center owners on the ground, may want to as well. And the math is starting to look like it’s not all that unreasonable. If something like Starship significantly lowers the cost per ton to launch them, the number of megawatt-years of computing power you can get for a billion dollars in orbit might actually be competitive with costs on the ground... and if ground centers make themselves unpopular enough, some might want to go there just because they’d rather pay a premium than deal with political hassles. The distance to go to make the costs converge is apparently not near as wide as I originally thought, and getting the economic balance into the black might actually be doable.

But that’s all based on pretty optimistic calculations. For instance, they want to put the sats into sun-synchronous orbits where they never need batteries to get through power interruptions, but such orbits will fill up if all these big players compete to put constellations into them, and these perpetually sunlit orbits also can make it tricky to give the computing hardware adequate cooling. Also, the cooling issue is one of several reasons why giant kilometer-plus satellites will probably not be very viable, so the way they will probably have to build these is as tight clusters of smaller sats flying in close formations, filling nearby space with a lot of noise from all the data zipping back and forth within the clusters. This vision of a single band of orbits around Earth’s terminator packed with competing constellations, in which each dot is actually a swarm of smaller projectiles weaving around each other, turning that part of the sky into a continual blast of buckshot where each pellet is in turn composed of birdshot, just doesn’t seem likely to end well. The huge growth of communication swarms was already putting us at risk of a Kessler Syndrome catastrophe where low orbit fills up with shrapnel as any one accident causes more than one additional accident... and now they want to escalate to more and heavier sats packed into fewer and more crowded orbits, so an outcome where everyone loses becomes ever more likely. And because they already perceive these orbits as scarce and valuable, their attitude is to rush in as hastily and carelessly as they can in order to stake a claim, even if the sats all lose money.

I’m sure I’m not the only one hoping for the AI investment bubble to pop as quickly as possible, because this is just one of many many areas where the harm it does will only increase the longer it goes on. Sure, a crash would cause a big economic setback for the general population, but even that will only be more severe if it’s postponed.

Terran 1: mass unknown, diam 2.3 m, thrust 900 kN, imp ~3.5 km/s, semi-closed expander (methane), payload 1.2 t, cost $9M/t, record 0/1/0 (final).