Photo: NASA/MSFC
How Many Rockets Should a Rocket Be?
A look at why almost every rocket is staged, why staging stops at two or three stages, and what that tells us about engineering trade-offs
The question
Rockets discard much of their mass on the way to orbit. Why do they stop at two or three stages? If dropping dead weight is good, why not drop it ten times?
The answer lies in a fundamental engineering trade-off. Adding a stage gives performance benefits in the form of discarded inert mass, potentially better mass ratios, the ability to optimise propulsion for different flight regimes, and more efficient acceleration of the upper vehicle after lower stages are discarded. However, every additional stage introduces more tanks, engines, interstage structures, interfaces, separation systems, manufacturing and testing requirements, and potential failure points.
The competing effects of staging raise the central questions of this investigation. For a given mission Δv and payload, what number of rocket stages provides the best overall performance when the marginal benefits of staging are weighed against the additional structural and architectural penalties? How does the optimal number of stages in a launch vehicle change with mission Δv, structural efficiency, and propulsion performance? When does the benefit of another stage stop outweighing its additional penalties?
Why it matters
A rocket has a nasty problem. To go fast, it needs fuel. The fuel needs tanks, the tanks need engines, plumbing, and other equipment, and all of that has to be pushed upward too. As a result, most of a rocket’s liftoff mass is propellant, and only a few per cent ends up as payload.
Staging is the trick that keeps this from being impossible. It shapes every launch vehicle today. How well you do it decides whether a mission is feasible, how much a launch costs per kilogram, and whether a rocket can be reused.
The frontier: who is doing what
Almost all operational rockets that launch satellites, cargo, or astronauts into orbit are multistage launch vehicles. SpaceX’s Falcon 9 and Rocket Lab’s Electron are two-stage vehicles [1,2]. Saturn V, which took astronauts to the Moon, used three main stages [3]. SpaceX’s Starship is also two stages, but both are designed to be reused, which changes the whole question [4]. Some vehicles like Electron also add a small kick stage on top for precise orbit insertion [2]. So the industry has more or less converged on two, sometimes three. The interesting part is why.
The deep dive: how staging works
The rocket equation
Everything starts with the Tsiolkovsky rocket equation:
The logarithmic term captures the fundamental challenge: achieving incremental increases in velocity requires disproportionately greater propellant mass. Since Isp × g₀ represents the effective exhaust velocity, an engine with a specific impulse of 350 s expels propellant at approximately 3.4 km/s. Reaching low Earth orbit takes roughly 9.4 km/s once you include gravity and drag losses, so the rocket has to be about e^(2.74) ≈ 15 times heavier at liftoff than when it is empty (Figure 1). That brings us to the structural problem. The rocket’s final mass is not just its payload. It also includes structures that must remain after the propellant burns, such as the engines, tanks, and other hardware still attached at that point.
Consider a purely illustrative rocket with an initial mass of 100,000 kg. A mass ratio of 15.5 would imply:
That final mass must accommodate the payload and the relevant remaining hardware. The rocket cannot devote all of it to the payload.
This creates the structural problem: more propellant increases the mass available to generate thrust, but that propellant must itself be accelerated. Stronger or heavier structures increase the mass that must be accelerated. Once a stage is empty, its structure is dead weight if the rocket continues carrying it. This is why rockets discard empty stages. They shed structures that are no longer needed, allowing the remaining stages to accelerate a smaller mass.
Rearranging the rocket equation gives:
Taking Δv = 9.40 km/s:
The structural coefficient
The structural coefficient, denoted by the Greek letter ε (epsilon), is the fraction of a stage’s mass that is structure rather than propellant [5]:
ε = structure / (structure + propellant)
A good modern stage might have ε around 0.08 [6], so 8% of the stage is tanks and engines, and 92% is fuel. In a single stage, the empty rocket has to weigh less than 1/15 of its fueled self. With ε = 0.08, that’s simply impossible, because the structure alone is already 8%. A single stage to orbit doesn’t work with these numbers, no matter how hard you try.
Why splitting helps
Consider a two-stage rocket. The first stage propels the vehicle through the initial part of its ascent before separating, taking its empty tanks and engines with it. The second stage then accelerates a substantially lighter vehicle, no longer burdened by hardware that has served its purpose. Each stage is responsible for a smaller share of the mission’s total velocity change, and the mass it must propel is progressively reduced. The exponential relationship in the rocket equation, however, remains unchanged; staging does not circumvent it. Rather, it makes the equation work in our favour by allowing the rocket to shed dead weight between burns, improving the efficiency with which the remaining propellant produces the required change in velocity.
I built a simple model where every stage gets an equal share of the Δv. For each stage, the mass ratio is R = e^(Δv_stage / exhaust velocity).
Our example has:
Stage payload fraction
Since ε=m_L/m_0, the result is λ = (1/R − ε) / (1 − ε), which gives the fraction of a stage’s liftoff mass that is left over as “payload” for the stages above it.
Multiplying λ across all stages gives the final payload fraction, the share of the rocket’s liftoff mass that actually reaches orbit.
What the numbers say
My baseline: mission Δv of 9.4 km/s, Isp of 350 s, and ε of 0.08.
Going from one stage to two is the difference between impossible and possible. Two to three still gives a noticeable boost. After that, the returns collapse. Thus, splitting a rocket into stages can allow it to carry more payload, but each extra stage contributes less than the previous one.
Get Lakshya Mina’s stories in your inbox
Join Medium for free to get updates from this writer.
*The one-stage case is infeasible under the model because the required velocity change demands a mass ratio that is incompatible with the assumed structural coefficient. This does not establish that every conceivable single-stage-to-orbit vehicle is impossible.
Adding a penalty for complexity
The ideal model hides a lie, assuming the structural coefficient remains constant as stages are added. However, every extra stage needs its own engine, tank structure, avionics, interstage hardware, and a separation system that can fail. To capture that, I added a simple penalty rule: each extra stage makes every stage slightly heavier, raising ε by 0.01 per additional stage (ε = 0.08 for stage count N = 1, 0.09 for N = 2, 0.10 for N = 3, and so on). This is my own assumption, and I treat it as a model parameter rather than a measured fact.
Now the payload fraction curve peaks around four stages and starts to turn over at five. The gain from three to four is tiny, about 0.10 percentage points, and after that the marginal returns are negative. The optimum moved, and the exact penalty rule decides how far. A heavier penalty pulls it down to three, and in reality the penalty includes cost, reliability, and manufacturing that my model does not even capture. This is why the real answer lands at two or three.
Engine vs architecture
I also asked whether it is better to improve the engine or add another stage for the aforementioned parameters. Starting from a two-stage rocket (3.58%):
- Isp +10% (350 → 385 s): payload fraction goes to about 5.1%, a gain of about 43%.
- Add a third stage: payload fraction goes to about 4.3%, a gain of about 19%.
- Structural coefficient −10% (0.08 → 0.072): about 3.9%, a gain of about 8%.
- Mission Δv +10%: about 2.4%, a loss of about 34%.
Engine efficiency is by far the most powerful lever for the given constraints. Across a wider range, a two-stage rocket with 300 s engines delivers only 1.8% of its liftoff mass to orbit, while the same rocket with 390 s engines delivers 5.3%, almost triple:
Specific impulse is a powerful design variable in this model, and a better engine beats an extra stage by a wide margin. This matches the real world, where engineers spend enormous effort on engine efficiency (hydrogen-fuelled upper-stage engines like the RL10 reach around 450 s or more) rather than endlessly adding stages [7].
Sanity check against real rockets
Real rockets land in a similar zone. Falcon 9 lifts roughly 550 tonnes and delivers about 22 tonnes to low Earth orbit in expendable mode, around 4% [1]. Saturn V did similarly, roughly 4% [8]. My simplified model lands in that same ballpark, which is encouraging, but I would be careful not to claim it “predicts” real rockets. It is too crude for that.
Challenges and limitations
My model deliberately leaves out a lot, and each omission matters:
- Gravity and drag losses: I folded them into a single 9.4 km/s figure instead of simulating the trajectory.
- Engine thrust-to-weight: A stage has to lift itself. Real engines have mass that scales with thrust, which I treated as fixed.
- Different stages, different engines: Real first stages use sea-level engines with lower Isp, and upper stages use vacuum-optimised engines with higher Isp. I used one Isp for everything.
- Equal Δv split: The best split among stages is not equal, and it depends on each stage’s parameters.
- Cost, reliability, and reuse: Each stage adds failure points and production cost, and reusability changes the whole objective.
So this is an architecture-level trade study allowing for generalisations and predictions, not a flight-design tool. It shows where the trade space is, not where any particular rocket belongs.
My take
Things I believe after working through this:
- Two stages is a sweet spot, not a law of physics. The equation says more stages theoretically keep helping. The reason rockets stop at two or three is that the benefit curve flattens while cost, complexity, and risk keep growing. It is an economic decision as much as a physical one.
- The engine sometimes matters more than the architecture. The optimal rocket is not necessarily the rocket with the most stages. If I had to put money on one improvement, I would probably put it in propulsion efficiency, not extra stages. The numbers obtained earlier make that very clear.
- Rocket engineering rarely has a single variable that can simply be maximised. More stages can improve mass efficiency, but they also create more hardware. Lower structural mass improves the mass ratio, but it may affect the reusability due to the specified minimum amount of hardware that must survive a mission. Every improvement therefore creates another constraint somewhere else. This is why I think systems engineering is the deeper idea behind staging.
- The ‘best’ rocket for a single launch is not necessarily the best rocket for a hundred launches. It depends on the mission and the constraints, including the mission Δv, payload, propulsion technology, structural efficiency, recovery strategy, manufacturing capability, reliability, and economics.
Sources
[3] Season of Apollo — Part Three — Saturn V
[5] Mass Ratios — Glenn Research Center
[6] Introduction to Aerospace Flight Vehicles — Principles of Flight Vehicle Design
[7] National Academies of Sciences, Engineering, and Medicine. 2006. A Review of United States Air Force and Department of Defense Aerospace Propulsion Needs, Page 115. Washington, DC: The National Academies Press. https://doi.org/10.17226/11780.
