The Falcon 9 Cannot Hover. That Is Why the Landing Looks Impossible.

Tim Dodd’s first launch was CRS-3, an early Commercial Resupply flight, maybe SpaceX’s fourteenth or fifteenth. The new feature on the Falcon 9 was landing legs. People were laughing. Why was this billionaire, Elon Musk, wasting time trying to land a rocket? It was not going to work. Dodd went to Kennedy Space Center as a photographer, left a remote-triggered camera on site, and went home after too many scrubs. The camera saw the liftoff. He watched the stream.

He had arrived disappointed. He did not know much beyond the Space Shuttle and the Saturn V, and he had been told this was not a big NASA rocket, not going to be that exciting, maybe three stories of skinny stick and a firecracker. At the pad he thought: this thing is huge. Seventy meters tall, 220 feet. The booster is 45 meters long and 3.7 meters, or 12 feet, wide. That is about as large as you can ship across the country without shutting down highways, built to lane privileges and bridges. People see a toothpick in photos and miss that it is a very large piece of machinery. Watching something that size leave the ground, he and Lex Fridman agree, is humbling.

Nine years later Dodd hosts Everyday Astronaut and is sitting with Fridman to tell the whole line from Falcon 1 to Starship. In 2022 China beat SpaceX by two launches, but many of those were tiny, on the order of 100 kilograms. SpaceX still put more mass on orbit, and the year before it had beaten the entire country of China. The company is now the number one launch provider in the world. Dodd no longer watches every flight. There are about two a week from SpaceX alone.

Falcon 1 is where it started. Musk wanted something on Mars, saw that NASA had no current Mars plan, and first tried to purchase a rocket from Russia. After a foiled attempt, he decided to develop his own. Dodd still does not know how Musk stumbled onto that team so quickly, but the people he assembled were amazing. They built a small-lift vehicle: one Merlin on the first stage and a Kestrel on the upper stage, tiny next to what SpaceX flies now. That Merlin kept evolving into the power plant for Falcon 9.

The hat they hung their business on was NASA. The COTS program, Commercial Orbital Transportation Services, evolved into commercial resupply contracts. SpaceX had to develop two things at once: Dragon, at first an uncrewed spacecraft that could dock to the ISS, and Falcon 9 to take it there. Dodd’s metaphor is plain. Falcon 9 is the semi truck. Dragon is the payload, the thing being dropped off, and the destination is the station. Cargo Dragon flew about twenty successful ISS missions, except CRS-7, when the rocket blew up. Dragon 2 then split into Crew Dragon and a cargo variant, a sleeker, ironically heavier update.

Flying people was a huge step. DM-2, Bob and Doug for NASA, was not a suborbital hop. Going into orbit and rendezvousing with the station is a different physics problem from crossing the Kármán line straight up. The heavier Crew Dragon also ended the old return-to-launch-site landings on CRS missions; those boosters come back to the drone ship. Reusability, Dodd thinks, was necessary from the start if you actually want an interplanetary mission to make financial sense. On Falcon 1 and the first two Falcon 9s they even wanted parachutes. They were not getting the booster through reentry long enough to pull them.

The first Falcon 9 had a square array, a three-by-three grid of Merlin 1Ds. Dodd thinks it lasted two or four flights. Then came the octaweb: a ring of eight engines and one in the center, still inside the same 3.7-meter fuselage, with a completely different thrust structure. The efficiency was manufacturing. You no longer had a corner engine, then an edge engine, then another corner. You could make eight of the same octaweb piece, and interchangeability got simpler.

After that, every time they touched the rocket it got longer and more capable, until today’s Block 5, even though it is something like the eighth or ninth iteration of the Falcon. Black landing legs, black interstage, a fleet of roughly ten boosters doing most of the work, a current reuse leader around fifteen flights. Fairings, the nose cones, are frequently if not always recovered. The only thing still expended is the upper stage. That, Dodd says, is where Falcon 9 is ending. It no longer makes sense to develop that infrastructure. The next step is physically bigger, so there is more margin for an upper stage you can actually reuse. That is Starship and Super Heavy.

SpaceX’s biggest customer is now itself. One of the biggest reasons it has launched so much mass is Starlink, designed around the payload fairing and the capabilities of Falcon 9. Because the company is vertically integrated, it can start from a blank slate: here are the mass constraints, here are the volume constraints. Another manufacturer might have made a squarer, heavier satellite and then shopped for a rocket. SpaceX made a funky-looking one, folded up about the size of a table, built to fit as efficiently as possible inside that fairing.

They fly those dozens of times a year, forty or fifty, and put up mass on a scale the industry had not seen. That is the same instinct Dodd sees across the company: start from the constraint you already hold, then let the hardware grow into it. Falcon 9’s diameter was a road. Starlink’s shape was a fairing. The next vehicle, Starship, is the move to go physically bigger so an upper stage can afford the structure to come home.

Rocket Lab later made parachutes work on a much smaller booster, about a metric ton empty, carbon-composite tanks, snatched out of the air by a helicopter on a drag line. SpaceX, at the time, could not even get the first stage through reentry long enough to try. The mass fractions were wrong. The breakthrough was to slow down before hitting the dense air. They used to call it a reentry burn. Dodd still thinks that is the correct term. They now call it the entry burn. Three of the nine Merlins light.

The engines do two jobs. They cut velocity by almost half, and while they fire they create what Dodd calls a literal force field as the booster falls. The killer in the atmosphere is compression: air piled against the front of anything flying through it gets so hot it becomes plasma and can destroy the vehicle. Slow down enough and the air molecules stop winning. At some similar crossover, he thinks, they realized that if they were already lighting engines to survive the atmosphere, they could use the same engines to land. Stick landing legs on it and come down vertically. On December 21, 2015, they did exactly that for the first time.

Dodd cannot tell you a line of the landing code or the avionics. He can tell you the hardware. They shut down all but one of the nine engines. Even that one engine, at its minimum throttle, is still too much thrust to hover. If they light a little too early, the booster stops above the ground and cannot lower itself. Say it stops 200 feet up. The only option is to kill the engine and fall those 200 feet. The industry names are suicide burn and hoverslam, used interchangeably, because the thrust-to-weight ratio is never below one.

So they ride the throttle. If the engine can come down to 40 percent of rated thrust, you might start around 70, in the middle of the window. Coming in too hot, you have room to go up. Coming in a little early, you have room to go down. They are down to one engine out of nine and still on the finest margin, and every bit of propellant used to land is propellant that did not put something into space. They want that as efficient as possible. Fridman is sure the code is not JavaScript. Dodd, well outside his domain, remembers that Musk wanted a common language so SpaceX would not need people who spoke some archaic dialect, and could pull engineers off the street. Probably C, probably C++. Reliable systems, he thinks, have to be.

They are coming up on a hundred consecutive perfect landings. About 150 or 160 landings altogether. Five years earlier this was experimental and insane. Now the landing, which is not the primary mission and does nothing for the payload most of the time, is becoming more reliable than the early flights. It is a long-term investment in recovering boosters. From roughly five launches a year, when Dodd would wake at 3 a.m., to two a week, the workhorse has gone mundane. Fridman wonders when orbital launches will exceed airplane takeoffs. Dodd has a hard time extrapolating that far. He cannot even tell you when Starship’s orbital flight is going, and that is a month-or-two-scale question.

CRS-4 almost landed. CRS-5 almost landed. CRS-6 almost landed. CRS-7 blew up. Dodd was watching on a Saturday or Sunday morning and thought, now what. It was the first failure, he believes the eighteenth flight. A helium bottle inside the upper-stage tanks broke off on ascent, over-pressured the stage, and the whole rocket went, in his word, kaboom, in an uncontrolled manner. They came back with vengeance. The first mission back was the first time they landed a rocket.

Put those two facts next to each other and you see the tempo. There was no long freeze. Failure in those years looks, in Dodd’s telling, like an iteration cost rather than a reason to stop the line. Then the cadence itself changed the experience. When there are five flights a year, every one is an event. When there are two a week, even the host of Everyday Astronaut catches only the big ones. United Launch Alliance, Rocket Lab, and Chinese missions pile on top. It is hard to keep up.

Publicly the big vehicle came out in 2016 as ITS, the Interplanetary Transport System. By the end of 2017 it was BFR, the Big Falcon Rocket. Around the end of 2018 they started calling it Starship. The whole system is confusing. The whole stack is kind of considered Starship, but technically Starship is just the upper stage, which is also the spaceship, and the booster is Super Heavy. That is what they are working full steam on. Falcon Heavy sits in the middle of the family tree: three Falcon 9-class boosters, and its 2018 demo is on Dodd’s milestone list.

Other milestones he would keep: first Falcon 1, first Falcon 9, first trip to the ISS, first booster landing, first reuse. The first reflown booster was SES-10 in 2017, about a year after they caught one. Catching it posed the next question: now what. Flying it answered. Then humans. Then Starship’s own early tests. The naming mess is not a footnote. If you hear “Starship’s first flight,” you still have to ask which piece.

The first Starship milestone, as Dodd classifies it, was Starhopper, a very rudimentary rocket and the first time they used the new Raptor engine to produce thrust and fly something. It first flew literally about three meters off the ground, tethered, then about 15, then 150 meters, in 2019. SN5 and SN6 did similar 150-meter hops with more elegant systems, proving out tank building and a lot of subsystems. The physically big ones were late 2020 and early 2021: SN8, 9, 10, 11, and 15. SN is serial number. They skip numbers when they scratch a test.

At Starbase, he says, one foot is always in front of someone else’s foot, and the arm is not always knowing what the leg is doing. They will tell someone to hurry up and build 40 tank sections, a bulkhead, a downcomer, a header tank, and then the design evolves and that header tank is not used anymore. Parts of certain rockets get built and then literally scrapped, not as a joke, actually sent to scrap. They iterate that quickly. The waste is visible. So is the learning speed.

Starship failures look epic, probably because of the amount of fuel, Fridman suggests. Dodd would measure them in explosive yield, like TNT, from how much propellant is left at the moment of the explosion. What has flown so far, even though the upper stage alone is one of the largest flying objects ever, has not been filled more than about 10 or 20 percent full of propellant. The fireballs are big and visual. In terms of spaceflight, they are still pretty small explosions. They could still go a lot bigger.

That gap matters once the cameras take over. Social media wants the picture. Engineering wants the load, the structure, and the failure mode. A prototype hopping with a mostly empty tank is not the same event as an orbital stack with full tanks. Dodd’s warning is almost a method: do not grade the program by how much of the sky turned orange.

In 2018 Falcon Heavy’s demo launched a Tesla Roadster. Dodd puts that flight on the milestone list without irony: first time a three-booster stack worked. The car is still out there. There is a site, whereisroadster.com. It orbits the Sun on a path that at its lowest point is back near Earth and at its highest goes beyond Mars, something like 2.5 AU if he remembers right.

Fridman asks whether a later mission could rendezvous and put extra things inside it. Dodd says it could absolutely be done. The hard part now is the eccentric orbit. Rendezvous and matching speed get messy. In principle there is no barrier. The stunt was also an engineering proof: three boosters in a row, first try.

Fridman’s observation is that Musk is exceptionally good at asking what the obvious way to do a thing is if you look out 20 years, and then making the robotics, brain-computer, and space decisions that only make sense on a 10-, 20-, 50-, or 100-year scale. The day-to-day still has to hit the next few months, because there are deadlines and missions. Dodd is honest that he cannot do that kind of extrapolation. He would not have predicted today’s situation five years ago. It is too hard. He cannot tell you when they land on Mars, or what the economy and the scale of launch operations for that would look like.

That mix, precise about the hardware in front of him and agnostic about the decade, is the useful one. Watch a landing with the telemetry up: speed, altitude, throttle. Read a rocket by diameter, length, and liftoff mass before you argue about the paint. When a Starship headline lands, decide whether it means the ship, the booster, or the stack. If you want to understand why reuse is hard, start with suicide burn. And treat public timelines as optimistic while treating the 10-year direction as, in Fridman’s framing, the thing Musk actually uses to choose the work.