At the beginning of October, SpaceX released new footage from Starship’s fourteenth flight—the one in which the rocket first reached a stable orbit and deployed twenty-six next-generation Starlink satellites. Among the clips was a moment that lasts seconds in the live broadcast and usually slips past the eye: the booster falls away from the ship in a cloud of its own fire. The upper stage lights its engines while still physically attached to the lower one, and the plume itself flings the stages apart.
The technique is called hot staging. I went looking into where it came from—expecting either a recent-years innovation or an old standard like orbital refueling. It turned out otherwise. The technique is more than sixty years old, invented in the USSR for a rocket that still carries crews to the ISS today, while the largest machine in human history adopted it only in 2023—for the sake of not losing part of its payload.
A short digression is needed here, without which the argument makes no sense.
A rocket is rarely single-stage, and not out of engineering modesty. The velocity an engine can impart depends logarithmically on the ratio between initial and final mass—which means hauling empty tanks and dead engines into orbit is monstrously expensive. A stage is discarded at the exact moment its fuel runs out, so it does not have to drag dead metal any farther. Between the moment the lower-stage engines shut down and the moment the upper-stage engines ignite, every rocket has a pause of several seconds.
The physics of that pause is ruthless in its simplicity. Until the rocket has built up a speed close to orbital, gravity pulls it down, and the engines must burn fuel merely to keep it from falling—this surcharge is called gravity losses. During those seconds of silence, the rocket loses velocity that must later be regained from an already depleted reserve. For a rocket that separates soon after launch, still on the steep, almost vertical part of its ascent, that loss is at its greatest.
There is another, less obvious reason not to go quiet. In zero gravity, liquid propellant drifts away from the tank walls and no longer covers the outlets leading to the pumps; if a pump gulps gas, the engine stalls or breaks apart. Before a restart, the propellant must be settled—give the rocket a gentle push forward so the liquid gathers at the bottom again. Usually this requires small solid-fuel ullage motors: separate hardware, its own mass, its own failure modes. Hot staging removes both the pause and that hardware in one stroke—the upper stage’s thrust does both the pushing and the settling.
The classic separation scheme is as cautious as a well-managed divorce through a lawyer. The lower stage shuts down, a ring of pyrotechnics cuts the connection with millisecond precision, springs or pushers separate the bodies, the upper stage drifts to a safe distance, and only then ignites. Hot staging does the opposite: the upper engines start first, straight at the lower stage, and it is kicked away by that same fiery shove.
What matters here is where the plume goes. The exhaust from several engines, directed into the enclosed compartment between stages, would tear the structure apart in fractions of a second. That is why the interstage is left open: on Soviet rockets it is a lattice truss, which the flame passes straight through; on the Titan II, it is a ring of sixteen exhaust ports, grouped four by four in four quadrants. Because of those ports, the rocket in launch photographs seems to be missing a chunk of its body.
The gain from hot staging is not just seconds. The upper stage is accelerating the entire time, which means Newton’s law works in both directions: the plume strikes the lower stage and pushes it away, and it answers with an equal shove—and that shove gives the upper stage a free boost. The initial kick does not have to be paid for out of the main propellant reserve. By the same logic, the hot version fits returning boosters perfectly: the kick itself starts the stage tumbling, turning it around for the flight back to the pad.
On this side of the Atlantic, the technique remained an exception for a long time and never became the norm. The one regular American trace is the Titan II, an intercontinental missile NASA took for the Gemini program and launched twelve times between 1964 and 1966. Everything in it was subordinated to simplicity: the rocket stood fueled directly in its silo on storable hypergolic propellant, with no separate hardware for pulling the stages apart. The designers deliberately made the second-stage engine burn while the stages were still joined.
Hence the oddity in Gemini launch photos: holes in the first-stage body. Those are the very exhaust ports without which gas from the second stage would have ripped the enclosed interstage apart from within. The fading but still-live thrust of the lower stage forced propellant into the feed lines, the ports vented the excess gas, and then the pyrotechnic bolts released the stage—and it fell away. A neat grid of openings flashed in the body where any other rocket would have had solid metal.
Apollo 5 stands somewhat apart. It was an uncrewed flight of the lunar module, and its program included a test called fire-in-the-hole. The name came from miners’ jargon: that was how they warned their comrades before blasting a charge. The point of the test was exactly the same as hot staging, only at minimum distance: shut down the descent engine, blow the bolts connecting the descent and ascent stages, switch control over, and ignite the ascent engine while it is still sitting inside the landing stage. This tested whether an abort would work in the event of a landing failure.
If you trace hot staging’s genealogy by dates, its roots are neither American nor recent. A Soviet upper stage ignited while the lower one still held on from the very beginning—already on Vostok, which carried Gagarin into orbit. Anatoly Zak, whose RussianSpaceWeb remains perhaps the most detailed independent source on Soviet rockets, puts it plainly: the open lattice structure between the second and third stages of Soyuz allowed the third-stage engine to begin ignition before separation—at 282.2 seconds into flight—and only then did pyrotechnics cut the connections. Today it looks standard: the lattice trusses on Soyuz are recognizable in any photograph.
It is even clearer with Proton. Its first and second stages are joined not by an enclosed interstage but by a genuine mesh, through which the exhaust burns. This too is hot staging, and its value is formulated in terms of engineering economy: it frees the second stage from ullage motors. The same argument recurs in every examination of the Soviet school—fewer parts, fewer failures. The logic of designers working in a country where repairs in orbit cost lives was simple: do not install extra hardware if existing thrust can do its job.
The separate irony is that the school’s grandest creation—the lunar N1—was not saved by hot staging. N1’s upper stages were meant to ignite before separation, but none of the four test launches stayed in the air long enough to reach that phase.
The technique has a bill, and it is charged to the metal. The shield covering the top of the lower stage must withstand, for several seconds, the barrage of the upper engines almost at point-blank range: heat and acoustic hammering. The interstage grows heavier, and part of the gain slips away again. The designer must also choreograph the motion: the shove has to be clean and straight, because a spent stage tumbling back into the upper stage’s path turns the whole trick into a lost mission.
And this is one of those cases where you cannot do the math on a napkin. The plume from Starship’s engines in a tight interstage strikes the booster shield and spreads around the vehicle—the environment is complex and, at the time of flight, poorly understood. NASA and SpaceX modeled hot staging with the Loci/CHEM solver: meshes of roughly five hundred million cells, around forty million hours of processor time on the Pleiades supercomputer, many terabytes of data. At separation altitude, the pressure is so low that the exhaust expands far beyond the rocket and envelops it completely. The data from the second flight matched the simulation—but it was the calculation that made it possible to understand what exactly was visible in the footage.
This is where the part absent from official press releases begins. When Musk announced the move to hot staging in June 2023, he called the gain “conservatively around ten percent” of payload. The figure made every headline and stayed there. The question “ten percent of what, and compared with what?” did not make the headlines.
The Space Exploration Stack Exchange community dug up the original wording, and it is slippery: Musk himself qualified it in the interview, saying that the increase depended on what you compared it to. The argument there was not contrarianism. Veteran user Russell Borogove put it plainly: the line “hot staging could add 10%” is equally honest if it added one percent and if it took away three. Another participant proposed a version in which the ten percent was not a gain over the classic scheme, but over the original plan with no separation mechanisms at all: by this hypothesis, SpaceX wanted to spin up the entire stack and use centrifugal force to throw off the upper stage, the way it deploys Starlink satellites, ran into problems, and adopted hot staging as a quick solution.
There is no way to verify it.
Neither SpaceX nor Musk has published a comparison methodology, and without a baseline, any “ten percent” is a number with no denominator. What remains is what can be seen: hot staging works, and without any exact percentage it has become part of the flight profile. The pedantic caveat about the unknown comparison is the only thing that truly follows from the sources, and it is not nitpicking: the question is not whether anyone is lying, but that the figure cannot be checked.
As for the fourteenth flight, whose footage started all this: there, the moment of hot staging had already become a checklist item. Fire in the hole has ceased to be an experiment and become a procedure, and that is perhaps the most honest conclusion to the story. A technique invented by Soviet designers to save hardware returned sixty years later on the world’s largest rocket for the same reason. Only now, no one can say exactly how many kilograms that saving is worth.