STAGE 01
Red rock, white powder
Alumina is half aluminium and half oxygen by weight, bonded so firmly that neither chemicals nor heat alone can separate them. Bayer gets you a white powder that is already half metal — and then stops dead. Everything expensive about aluminium happens because that last bond won't break with heat or chemistry. It takes electricity.
Aluminium starts as a reddish rock called bauxite, dug out of the ground. The rock is cooked in a hot bath of chemicals until a fine white powder drops out — and that powder is aluminium still glued tight to oxygen.
For engineers: A selective-dissolution separation exploiting the amphoteric solubility of gibbsite/böehmite in caustic soda above ~100 °C and ~200 °C respectively — digestion temperature and pressure are set by which mineral phase dominates the orebody. Economics are dominated by ore grade: residue runs 0.3 kg per kg of alumina for high-grade ore versus 2.5 kg for low-grade — an eightfold swing in waste handling for the same product.
↓ Which is why: Bayer stops dead at a powder that is already half metal. The last bond — aluminium to oxygen — will not break for heat, and it will not break for chemistry. There is exactly one thing left to throw at it.
STAGE 02
The metal made of electricity
Aluminium isn't rare — it's the most common metal in the Earth's crust. It was historically precious purely because nobody could afford the energy to free it. Hall and Héroult's first commercial cells needed more than 40 kWh per kilogram. That's the whole story of aluminium: not a scarce element, a scarce power bill. Smelters today are built where the electricity is, not where the ore or the customers are.
To pull the oxygen off, you melt the white powder in a bath hotter than lava and push a huge electric current through it. The aluminium sinks to the bottom as liquid metal. A rocket is basically electricity you can hold.
For engineers: The consumed carbon anode is not incidental — it is thermodynamic leverage. Reducing alumina with carbon (2Al₂O₃ + 3C → 4Al + 3CO₂) has a theoretical minimum of 5.99 kWh/kg, versus 9.03 kWh/kg for inert-anode reduction (2Al₂O₃ → 4Al + 3O₂). The carbon supplies part of the reduction energy — which is precisely why the process emits CO₂ by design and why inert-anode retrofits carry a large energy penalty. Practical anode consumption (~0.45 kg C/kg Al) runs ~35% above the theoretical 0.33 kg.
↓ Which is why: You now hold pure aluminium — the most common metal in the crust, finally let out of its rock. It is also nowhere near strong enough to be a rocket. It is soft.
STAGE 03
The lightest trick in the book
On a rocket, every kilogram of tank is a kilogram of satellite you don't fly. But the reason aluminium-lithium took a century to reach a launch vehicle isn't that nobody wanted lighter tanks — it's that you couldn't weld the stuff. The alloy was waiting on the welding.
Pure aluminium is too soft for a rocket. So you stir in a pinch of lithium — the lightest metal there is. The mix comes out both stronger and lighter, which almost never happens.
For engineers: The tradeoff is manufacturability, not properties. 2xxx-series aluminium alloys are effectively unweldable by fusion methods — they suffer solidification cracking. So aluminium-lithium's mass advantage is unusable unless you also have a solid-state joining process. Al-Li adoption and friction stir welding are a package deal; neither is worth much without the other.
↓ Which is why: Lithium buys lightness and stiffness in the same stroke. Now decide what shape to pour that advantage into — because on a rocket, shape is mass and mass is payload.
STAGE 04 · Hawthorne, California
The tank that shares a wall
A rocket is mostly empty space you're paying to carry. Deleting one wall shortens the whole vehicle — but it forces you to pipe cryogenic oxygen through a tank of kerosene, so a fuel-freezing problem is created by a mass-saving decision. And the second stage isn't a different design: it's the first stage's tooling, run shorter.
Inside the rocket are two tanks — one of cold liquid oxygen, one of kerosene. Instead of two separate tanks, they share a single wall between them, and a pipe carries the oxygen straight through the middle of the fuel tank to reach the engines.
For engineers: The common dome buys length and mass — one bulkhead instead of two plus an intertank barrel — at the cost of a thermal problem: cryogenic LOX now shares a wall with kerosene that must not freeze, hence the double-wall transfer tube. Structurally the two tanks answer different load cases, which is why they're built differently. Second-stage tanks are a shorter version of the first-stage tanks, reusing most of the same tooling.
↓ Which is why: You have drawn a thin-walled barrel with one shared wall and a tube of liquid oxygen threaded through the kerosene. Now join it all together — and discover that the alloy that made it light is an alloy you cannot weld by melting.
STAGE 05 · Hawthorne, California
The weld that never melts
Everyone assumes welding means melting. FSW joins metal that stays solid the entire time — you are essentially stirring two pieces of metal into one. And this isn't a nicety: the alloys that make the rocket light are exactly the alloys you physically cannot fusion-weld. The tank exists because of the welding process, not the other way around.
To join the pieces, a spinning tool is pushed into the seam and dragged along it. The metal gets soft and gooey like clay and squishes together — but it never actually melts. There's no flame and no puddle.
For engineers: FSW sidesteps the solidification physics entirely. No molten pool means no solidification cracking and no gas porosity — the two failure modes that make 2xxx and 7xxx aluminium alloys 'non-weldable' by fusion. Peak temperatures are lower, so distortion and shrinkage drop, which matters enormously when holding circularity on a 3.66 m thin-walled barrel.
↓ Which is why: The airframe is now one continuous piece of metal, light because of the alloy and possible because of the weld. It still has nothing whatsoever to push it.
STAGE 06 · Hawthorne, California
Nine engines, one room
The pintle injector at Merlin's heart is not new — it flew on the Apollo Lunar Module's descent engine, the thing that lowered astronauts onto the Moon. SpaceX went looking for the injector least likely to shake an engine apart, and found it in 1969. Meanwhile, the newest part of the engine is a valve body that was printed in under two days, versus months for a casting.
All nine of the rocket's engines are built in one factory in Los Angeles — a building so big two whole Falcon 9s could lie end to end across it. One valve inside the engine isn't machined at all. It's printed.
For engineers: Merlin is a deliberate exercise in choosing the simpler option and buying performance back through mass production. Gas generator over staged combustion; a single shaft carrying LOX pump, fuel pump and turbine; the turbopump also supplies high-pressure kerosene to the hydraulic actuators, eliminating a separate hydraulic system entirely. The pintle injector's inherent stability sidesteps the combustion-instability programmes that historically consumed years. SpaceX's own framing: flying ten engines per Falcon 9 means 'high product quality and repeatability through process control and continuous production'.
↓ Which is why: Nine engines exist, and they are beautiful. Built is not the same as trusted, and nobody has yet lit a single one of them.
STAGE 07 · McGregor, Texas
Every engine gets fired
The qualifier is everything: 'before first flight.' Acceptance testing screens manufacturing defects, so it belongs to the moment a unit is born — not to every mission it flies. A reused booster has already proven itself by flying and landing; sending it back to Texas would test nothing new. The test exists to catch the mistake, not to re-earn the trust.
Before a new engine is allowed to fly, it's bolted to the ground in a field in Texas and lit. Every single one. If it fails there, it fails somewhere safe.
For engineers: This is workmanship screening, not design qualification — the design is proven separately, and acceptance firing exists to catch build defects in this specific unit. That's why SpaceX distinguishes acceptance from qualification testing throughout its documentation, and why the qualifier in that sentence is doing so much quiet work.
↓ Which is why: It has passed, and it has passed in the only way that counts — as this specific unit, not as a design. It is also in a field in Texas, and the sea is a very long way away.
STAGE 08 · Hawthorne → McGregor → the Cape
A rocket lying down
The most expensive thing about a vertical rocket isn't the rocket — it's the building. Apollo needed one of the largest structures on Earth just to stand a rocket up indoors. SpaceX's answer was to never stand it up until the last possible moment. The rocket spends nearly its entire existence — factory, truck, test stand, hangar — lying on its side. It stands up once, and then it leaves.
The finished rocket doesn't fly to the launch pad — it goes by truck, on its side, down ordinary highways, for thousands of kilometres. Then it waits in a shed by the sea, still lying down.
For engineers: SpaceX's own stated rationale is safety, not cost: horizontal manufacturing, processing and integration 'reduces work at height during numerous manufacturing, processing, and integration procedures, and eliminates many overhead operations.' The scale contrast is the argument's real content — the VAB is 525 ft tall, covers eight acres, and its 456-ft doors take about 45 minutes to open. The constraint it imposes lands on the customer: payload access ends once the vehicle is vertical, and some national-security satellites must be mated vertically and cannot fly this way.
↓ Which is why: It has been horizontal for its entire existence: factory, truck, test stand, hangar. Everything left to do happens standing up, and it only happens once.
STAGE 09
Standing up, once
SpaceX quietly deleted a step from its own rulebook. The April 2020 user's guide promised customers 'a static fire test on the launch pad.' By September 2021 that clause was gone from the document — and it has never returned. You can watch a company's confidence change by diffing its paperwork.
A machine called the transporter-erector rolls the rocket out to the pad and lifts it upright, like standing a pencil on its end. Then it lets go — and the rocket leaves.
For engineers: The hold-down-then-release sequence is a genuine abort opportunity: the engines are running and being evaluated while the vehicle is still bolted to the ground, so a bad start is a scrub rather than a failure. The pre-launch static fire, once universal, is now selective — dropped as a routine step, retained for crewed and selected missions. That is a reliability dividend, and it is visible in the paperwork.