SpaceX stopped catching $6M fairings and just fished them out of the sea

SpaceX started landing Falcon 9 boosters in 2015 and reusing them within two years — the first stage holds nine engines and roughly half the rocket's manufacturing cost. That left the payload fairing, two composite halves laid up like papier-mâché, costing about $6 million a pair. Elon Musk challenged the team to catch them: 'You have six million bucks falling from the sky.'

Engineers assumed the halves could never touch seawater — saltwater corrosion would ruin the delicate electronics — so they built an elaborate recovery: each half flew a parafoil that steered itself while an automated control system turned the boat to close the gap. It caught the first fairing in January 2020. 'It worked. We did it,' said Kiko Dontchev, SpaceX's VP of launch — but not with regularity: most of the time the Atlantic is 'a choppy hot mess with 7- to 9-foot waves,' and the real recovery rate sat under 50 percent, around 40, slow enough to choke the launch cadence.

The fix came from SpaceX's five-step design algorithm — make the requirements less dumb, delete the part or process step, repeat, then optimize, accelerate, automate. Challenging the requirement showed the premise was wrong: composite fairings float like a sailboat, and moving sensitive parts higher meant a splash would not kill them. A standard vessel with a crane now fishes the halves out of the ocean. Recovery success rose from about 40 percent to 99 percent, refurbishment takes weeks, and the company now has more fairings than it has space for.

Step two of SpaceX's algorithm — delete the part — asked why catch at all: fairing halves are composite boats; move the electronics up and any crane ship recovers them.

The hard problem was self-inflicted: 'catch before splashdown' was a requirement, not a fact, and deleting it collapsed a parafoil-plus-chasing-boat automation stack into a crane.

Designing for survivable splashes cost a relocation of electronics — trivial next to a dedicated recovery ship and automated rendezvous.

A 40-percent recovery rate was a launch-cadence tax; at 99 percent, fairings stopped being a scheduling constraint at all.

When automation barely works, interrogate the requirement before tuning it: the cheapest subsystem is the one you delete, and 'recover fast' may not need 'catch mid-air.'

By 2023 Dontchev described fairings as a solved afterthought — always ready, more than the company has room for — and used the episode as his worked example of the five-step SpaceX algorithm, from a talk at the Summit at Sea program.