Why Does the Cheapest Rocket Keep Winning?

Where the other 98% went

Before 2006, the global launch industry operated on a stable assumption: rockets are expensive because rockets are inherently expensive. The vehicles were complex, the failure modes catastrophic, the customers few, and the suppliers entrenched. A single launch to low Earth orbit cost between $10,000 and $30,000 per kilogram. The industry had optimised within those parameters for decades and concluded the parameters were fixed.

Then a company built a rocket whose material cost was roughly 2% of the prevailing launch price — and asked where the other 98% went. The answer turned out to be structural, not physical. The cost lived in contracting models, supplier margins, interface complexity, and institutional habits that had compounded over fifty years of cost-plus procurement. The raw materials for an orbital rocket cost less than a mid-range business jet. The price was a product of how the industry was organised, not what the vehicle was made of.

That question — why does the gap between material cost and market price exist? — is a method, not just an observation. It applies wherever an industry's pricing reflects accumulated institutional structure rather than underlying physics. The method is called first-principles costing: decompose to physical inputs, price each, measure the gap, and investigate what created it. The investigation matters more than the arithmetic; the gap itself is the diagnostic.

Vertical integration as architecture

But cost analysis alone builds nothing. The structural follow-through was vertical integration as competitive architecture — manufacturing roughly 80% of components internally, including engines, avionics, flight computers, and fairings. The aerospace industry's standard practice was to outsource most subsystems to specialist suppliers. Each supplier added margin, compliance overhead, and interface complexity. More critically, each supplier imposed its own development timeline. An engine vendor with an eighteen-month lead time makes rapid iteration impossible regardless of how fast the vehicle integrator moves.

Bringing production in-house collapsed three constraints simultaneously. Cost fell because supplier margins disappeared. Lead times shortened because design changes no longer required contract renegotiation. And iteration speed increased because the engineers who designed a component could walk to the factory floor and watch it being built. The Raptor engine — the most advanced rocket engine ever constructed, using full-flow staged combustion — was developed at a pace that would have been structurally impossible with outsourced subsystems.

Vertical integration is not universally correct. It demands enormous capital, deep expertise across many domains, and tolerance for internal failures that an outsourcing model externalises. The decision depends on whether the constraint being removed — supplier lead time, interface complexity, cost opacity — is the binding constraint on the system's performance. In aerospace, it was.

Iteration funded by cheap failure

The iteration methodology itself deserves separate examination. Orthodox aerospace development designs exhaustively on paper, tests components individually, and aims for the complete vehicle to work on its first integrated flight. Failure is career-ending and budget-ending. The alternative approach — build fast, test the whole system, expect failure, treat each failure as data — was considered reckless in an industry where vehicles cost billions.

The enabling condition was cost. When a test vehicle costs $60 million instead of $1.5 billion, the economics of controlled failure invert. The Starship prototype sequence — SN8 through SN15, each flying once, each teaching one specific thing — represents the most compressed hardware learning curve in modern aerospace. The iteration rate per dollar spent was radically higher than any government programme achieved over the same period.

This methodology has a prerequisite that is institutional, not technical: an organisation that treats explosions as engineering data, not as crises. That is a genuinely different psychology from traditional aerospace, and it cannot be adopted by policy memo. It requires hiring practices, incentive structures, and leadership norms that are consistent with treating failure as information.

Cheap reusability

Reusability extended the cost logic from manufacturing to operations. A first stage that costs $30 million to build and flies once costs $30 million per launch in stage cost alone. The same stage flown twenty times amortises to $1.5 million per launch. The arithmetic was obvious for fifty years; the engineering was not. Landing a rocket requires reserved fuel (reducing payload), precision guidance, grid fins, landing legs, and engines capable of deep throttle. Each was a solved problem individually. Combining them into a reliable, economical system was the hard part — and "economical" is the key word. The Space Shuttle was reusable. It required months and hundreds of millions in refurbishment per flight, making it more expensive than expendable alternatives.

The breakthrough was not reusability itself but cheap reusability — refurbishment cost plus amortised build cost well below the price of a new expendable stage, with turnaround times measured in weeks rather than months. The existing fleet of roughly twenty flight-proven boosters now handles most of the launch manifest without building new vehicles. This converts a manufacturing business into something closer to an airline: capital asset utilisation replaces unit production.

Building the demand

Reusability, however, only pays off at high flight rates. Flying a booster twenty times requires twenty missions. Before 2019, total global launch demand was roughly thirty missions per year. At that volume, fleet utilisation economics barely close. The response was to build the payload: a satellite internet constellation that simultaneously functions as a telecommunications business and a mechanism to generate the launch cadence that reusable rockets need to become profitable. Approximately 80% of recent launch activity has been internally generated demand. Without it, the flight rate would match competitors — around twenty launches per year.

This is vertical demand integration: manufacturing the vehicle, operating the launch, and running the end-user service. The company that began by selling launch services is becoming a telecommunications provider that happens to own its own launch vehicle. The principle generalises: in capital-intensive industries with high fixed costs, creating demand can be more valuable than competing for existing demand.

One vehicle, every mission

The deepest competitive advantage, though, may be architectural. One vehicle serves cargo resupply, crewed spaceflight, military payloads, commercial satellites, constellation deployment, and deep-space science missions. One production line, one launch operations team, one accumulated learning curve across more than three hundred flights. A new entrant must match not just the vehicle's performance but the compounding operational learning embedded in that flight history — the same dynamic that makes it difficult to compete with a mature commercial aircraft platform.

The principles at work here — cost restructuring from first principles, vertical integration as a speed enabler, iterative development funded by cheap-enough-to-lose hardware, reusability economics, demand creation, and platform architecture — are not aerospace-specific. They describe a general pattern: an entrant that restructures an industry's cost basis, development methodology, and demand assumptions simultaneously, rather than competing on any single axis. The reason the cheapest rocket keeps winning is that cost was never the only variable being optimised. It was the most visible result of a system designed to optimise all of them at once.