In January 2019, Elon Musk announced that SpaceX would build Starship — the largest rocket ever — out of stainless steel. The aerospace industry thought he’d lost his mind. Carbon fiber is lighter, stiffer, and SpaceX had already built a 9-meter carbon fiber test barrel. Musk scrapped it. The reason wasn’t nostalgia. The reason was math.

Here’s the math that changed everything: 304L stainless steel at room temperature has a tensile strength of about 515 MPa — nothing special. Cool it to liquid oxygen temperature, and it nearly triples to roughly 1,500 MPa. The steel gets stronger as it gets colder. Carbon fiber composites do the opposite — their resins become brittle at cryogenic temps, risking cracks that can spread through the laminate. Stainless steel doesn’t just survive the cold. It thrives in it.

That single property — cryogenic strengthening — restructured the entire Starship program. It changed how the vehicle is built, how the heat shield works, how fast SpaceX iterates, and how much each vehicle costs. This might be the most important material choice in modern rocketry.

The Decision

SpaceX had been publicly committed to carbon fiber since 2016. They built a full-diameter 9-meter test barrel at the Port of Los Angeles. The tooling alone cost millions. When Musk decided to switch, that barrel and all its composite infrastructure became instantly obsolete. Photos of the abandoned barrel sitting on the dock circulated on social media for months — a 9-meter monument to a killed engineering path.

The switch wasn’t impulsive. Musk later explained he’d been running the numbers for months. The case was overwhelming across five dimensions: cost, manufacturability, thermal performance, damage tolerance, and iteration speed.

Cost: 60-100x Cheaper

Raw 304L stainless steel: $3-5 per kilogram. Aerospace-grade carbon fiber prepreg: $135-200 per kilogram. That’s a 60-100x difference before you even start building anything.

But raw material is just the beginning. Carbon fiber also needs:

• Precision molds costing millions that wear out and need replacement
• Clean rooms where dust, humidity, and temperature are tightly controlled
• Autoclaves — giant pressure cookers for curing. One big enough for a 9-meter barrel would be among the largest ever built
• Ultrasonic inspection of every square meter for hidden defects
• Higher scrap rates — a contaminated batch or temperature glitch can condemn an entire section. Steel forgives. A bad weld gets ground out and re-welded

Total manufactured cost: 100-200x higher for carbon fiber. For a vehicle SpaceX wants to build by the hundreds, that’s disqualifying.

Manufacturing: Welders vs. Autoclaves

Stainless steel is one of the most widely fabricated metals on Earth. Every industrial city has shops that cut, bend, and weld it. The workforce exists. The supply chain exists.

SpaceX rolls 4mm-thick 304L sheet into ring segments and welds them together with standard MIG and TIG processes. They do this outdoors at Starbase in Boca Chica, Texas. No clean rooms. No autoclaves. Just wind, rain, Gulf Coast humidity, and the occasional pelican.

Compare that to Boeing’s 787 fuselage sections — smaller at 6 meters — which require billion-dollar facilities and weeks of automated fiber placement, autoclave curing, and inspection. SpaceX welds a Starship barrel ring in days.

The talent pool matters too. Certified stainless steel welders number in the hundreds of thousands in the U.S. alone. Certified composite layup technicians are scarce.

Thermal Performance: The Built-In Backup Heat Shield

304L melts between 1,400°C and 1,450°C and keeps useful structural strength up to about 800-900°C. Carbon fiber composite? Its resin fails at 150-300°C. The resin softens, the laminate loses strength, and the structure collapses. Suddenly. Totally.

This changes the entire thermal protection philosophy. If a heat shield tile falls off Starship during reentry — and tiles have fallen off every reentry vehicle ever built, including the Shuttle — the exposed steel can survive temperatures that would destroy carbon fiber or aluminum instantly. Musk calls it a “belt and suspenders” approach: tiles do the primary work, steel provides the backup.

The Columbia disaster happened when a breach in the Shuttle’s wing let superheated gas reach the aluminum structure underneath. Aluminum can’t take those temps. Stainless steel can — at least long enough to survive reentry.

Cryogenic Strengthening: The Physics That Sealed It

This deserves its own section because it’s the single most important reason Starship is made of steel.

When 304L gets cold, some of its crystal structure transforms from austenite to martensite — a phase that’s significantly harder and stronger. At liquid oxygen temperature (-183°C), ultimate tensile strength jumps from ~515 MPa to ~1,500 MPa. Yield strength climbs from ~205 MPa to over 1,000 MPa.

Think about what that means for a rocket. The propellant tanks are strongest precisely when they’re under the greatest stress — fully loaded with cryogenic propellants and subjected to launch forces. The steel is weakest when the vehicle is empty and loads are lowest. It’s an almost perfectly matched material response.

Carbon fiber composites don’t do this. The fibers handle cold fine, but the epoxy matrix becomes brittle and develops microcracks during thermal cycling. Those cracks can create leak paths for cryogenic propellants. NASA spent decades studying this problem. SpaceX sidestepped it entirely.

Damage Tolerance: Dents vs. Shatters

For a reusable vehicle that’s supposed to fly, land, and fly again — potentially within hours — damage tolerance is critical.

Steel dents. The dent absorbs energy, the surrounding structure stays intact, and you can see the damage during inspection. A dented steel tank can often still hold pressure.

Carbon fiber shatters. Impact creates hidden internal delaminations that can grow under load and cause sudden, catastrophic failure with no warning — a phenomenon called Barely Visible Impact Damage (BVID). Finding it requires ultrasonic inspection of every square meter after every event.

For a vehicle SpaceX wants to turn around as fast as an aircraft, the inspection burden of composites is a deal-breaker. Steel wears its damage on the surface. Composites hide theirs.

Iteration Speed: Weeks, Not Months

Between 2019 and 2024, SpaceX built and tested over 20 full-scale Starship and Super Heavy prototypes. Many exploded. Each failure fed data into the next build. The gap between prototypes? Weeks.

This pace is only possible with steel. Cutting, rolling, and welding a new barrel section takes days. Changing a dome profile or adding stiffeners can happen between prototypes with minimal retooling.

Carbon fiber operates on a fundamentally different timeline. New layup schedule: weeks of analysis. New molds: months. Layup, cure, inspect: more weeks. Total cycle from design change to finished hardware: 3-6 months. SpaceX would have built maybe 4-5 prototypes in the time they actually built 20+.

Musk has said iteration speed was just as important as cost in the material decision. Building, breaking, and rebuilding on a weekly cadence compressed a decade of development into a few years.

301 vs. 304L: The Details Matter

SpaceX has used both 301 and 304L on Starship prototypes. The “L” in 304L stands for Low carbon (0.03% max vs. 0.08%). Lower carbon prevents chromium carbide precipitation during welding — a problem called sensitization that creates corrosion-prone zones along grain boundaries.

Since Starship is almost entirely welded, sitting at a coastal launch site exposed to sea spray, rain, and rocket exhaust, 304L’s corrosion resistance is a practical necessity. It welds cleanly, resists corrosion, and still gets that cryogenic strength boost.

How Steel Enables the Heat Shield

The Shuttle used ~24,300 unique silica tiles bonded to aluminum. Each tile was hand-fitted. Each was irreplaceable. The aluminum underneath couldn’t handle more than about 175°C — so every single tile had to work perfectly. One tile failure in a critical area could destroy the vehicle. And did.

Starship uses hexagonal ceramic tiles that are interchangeable — any hex tile fits any hex position of the same size. They attach to steel studs welded to the skin, not adhesive. This simplifies manufacturing, replacement, and robotic installation.

If a tile falls off, the exposed 304L steel survives temperatures up to 800-900°C. An aluminum substrate would fail almost immediately. A composite would pyrolyze. The steel endures — buying seconds to minutes depending on location, which can be enough to complete reentry.

Material Comparison: 304L vs. CFRP vs. Aluminum 2219

Property	304L Stainless Steel	CFRP (Quasi-Isotropic)	Aluminum 2219-T87
Density	8,000 kg/m3 (499 lb/ft3)	1,550-1,600 kg/m3 (97-100 lb/ft3)	2,840 kg/m3 (177 lb/ft3)
Tensile strength (RT)	515 MPa (74,700 psi)	800-1,000 MPa (116,000-145,000 psi)	455 MPa (66,000 psi)
Tensile strength (cryogenic)	~1,500 MPa (217,500 psi)	800-1,000 MPa (matrix microcracking risk)	~530 MPa (76,900 psi)
Specific strength (cryo)	188 kN-m/kg	500-645 kN-m/kg (with microcracking risk)	187 kN-m/kg
Melting / max service temp	1,400-1,450 C (2,552-2,642 F)	150-300 C resin limit (302-572 F)	660 C (1,220 F) / ~175 C service
Raw material cost	$3-5/kg ($1.36-2.27/lb)	$135-200/kg ($61-91/lb)	$15-30/kg ($6.80-13.60/lb)
Fabrication method	MIG/TIG welding (outdoor capable)	Autoclave cure (clean room required)	Friction stir welding (controlled environment)
Damage tolerance	Ductile (dents, visible damage)	Brittle (delaminates, hidden damage)	Ductile (dents, fatigue cracking)
Prototype turnaround	Days to weeks	Months	Weeks to months
Notable vehicles	SpaceX Starship, Atlas (1950s-60s)	SpaceX BFR (abandoned), proposed concepts	Saturn V, SLS, Falcon 9, Centaur

At room temperature, CFRP dominates on strength-to-weight. But at cryogenic temps — where Starship’s tanks actually operate — 304L closes the gap dramatically, without any of the microcracking concerns. Factor in cost, build speed, thermal tolerance, and damage tolerance, and the overall case shifts decisively to steel.

The Weight Penalty (And Why It Doesn’t Matter)

At 8,000 kg/m³, 304L is five times denser than carbon fiber. How is the heaviest option the right choice?

First, cryogenic strengthening allows thinner walls than room-temperature specs would suggest. Second, steel’s thermal tolerance enables a lighter heat shield — the Shuttle’s tiles weighed roughly 9,000 kg, while Starship’s hex tiles backed by thermally tolerant steel target a lower mass fraction. Third, Starship is enormous. At 9 meters wide and 121 meters tall, a few extra tonnes of airframe matter less when you’re lifting 100-150 tonnes to orbit.

The Abandoned Carbon Fiber Barrel

The 9-meter carbon fiber barrel SpaceX built in 2017-2018 deserves a moment. It was a genuine engineering achievement — a flight-quality composite cylinder at a diameter that pushed the state of the art. When the steel switch happened, it was abandoned at San Pedro and eventually scrapped.

It wasn’t a failure. It was data. SpaceX learned enough about large-scale composites to confidently conclude that steel was better. That willingness to walk away from millions in sunk costs based on engineering evidence is arguably as important as any specific material property.

Frequently Asked Questions