Starship's first orbital attempt: the most

Starship launched for the first time today from Boca Chica, Texas.

All 33 Raptor engines on the Super Heavy booster lit at T-0. The thrust was staggering. The loudest thing I’ve ever heard through a livestream. The launch pad was engulfed in a cloud of concrete debris and fire (the pad didn’t survive, which is a problem they’ll need to fix).

The stack cleared the tower. It climbed. It punched through Max-Q.

Then engines started shutting down. One. Then another. Then more. The vehicle began to tumble. At T+4 minutes, the flight termination system activated and Starship broke apart over the Gulf of Mexico.

SpaceX called it a success. NASASpaceflight called it a success. I agree with both.

Why an explosion is a success

The primary objective of this flight was to clear the launch pad. Not orbit. Not stage separation. Just: light the engines and get off the pad without destroying the launch infrastructure (they partly failed this one, given the crater, but the vehicle cleared).

Everything after pad clearance was bonus data. And they got four minutes of it. Four minutes of flight data on 33 engines, on the booster’s structural performance, on the aerodynamic behavior of the full stack, on the flight termination system. Four minutes of data that would have taken years of ground testing to approximate.

SpaceX’s approach to rocket development is: build, fly, collect data, iterate. They’re not trying to get it right the first time. They’re trying to fail informatively.

Watching SN8, SN9, SN10, SN11 all explode before SN15 landed successfully taught me this pattern. The explosions aren’t failures in the traditional sense. They’re iterations. Each one teaches something the next one incorporates.

The pad problem

The launch pad took significant damage. SpaceX reportedly chose not to install a flame diverter or water suppression system for this flight, relying instead on a steel plate. The steel plate did not hold up. Chunks of concrete were thrown hundreds of meters. A car in a parking lot was damaged. Environmental concerns about debris entering the Gulf are already being raised.

The FAA will need to sign off on the next flight. That process might take months. The pad needs to be rebuilt.

This is the one part of the flight that wasn’t a success. The rocket left, but it also destroyed the thing it left from. For the iterative approach to work, you need to be able to iterate quickly. A damaged pad slows the iteration cycle.

The explosion itself

I watched the moment of flight termination several times. The vehicle was tumbling, engines firing at random angles, the booster and ship still connected (stage separation had failed). Then the charge detonated and the whole thing came apart.

From the ground cameras, it looked like a second sun. A sphere of orange light in the upper atmosphere, expanding, then fading. Debris arcing through the sky in every direction.

The most beautiful explosion I’ve ever seen. The biggest rocket ever built, carrying the most engines ever fired simultaneously, coming apart 40 kilometers above the Gulf of Mexico. Not a failure. A data point. A very expensive, very loud, very bright data point.

SpaceX will fly again. The pad will be rebuilt (with a flame trench this time, presumably). The engines that shut down will be investigated. The stage separation failure will be analyzed. The next vehicle is already being assembled in Boca Chica.

The biggest rocket ever built flew. For four minutes. And then it exploded, beautifully, and the team that built it started planning the next one before the debris hit the water.

That’s how the future gets built. One explosion at a time.

Related thinking: