Link to the code: Zae-Project / arkspace-core

The Physics of Reusability: What Erik Seedhouse's 2026 Analysis Gets Right About the Starship Economy

Every argument for orbital infrastructure at scale, whether a million-satellite computing constellation or a permanent Mars presence, eventually reduces to a single engineering variable: cost per kilogram to orbit. That number is not set by ambition or capital allocation. It is set by physics. Aerospace engineer Erik Seedhouse’s 2026 book SpaceX’s Dragon and Starship: Revolution, Reusability, and the Race to Mars is the most technically rigorous public accounting of where that physics currently stands, and where it is actually headed.

The book arrives as Starship is transitioning from test flights to operational missions. Three integrated flight tests in 2024 and a fourth in January 2025 demonstrated full-stack flight and ocean recovery. The question Seedhouse addresses is not whether Starship works, but whether the reusability economics that make it significant can actually be achieved at the margins that proponents project.

Reusability Is Not Reusability

The first clarification the book offers is definitional. “Reusability” covers a wide range of actual engineering achievements, and conflating them produces predictions that consistently miss reality.

Falcon 9 first stage reuse is a mature, high-cadence operation. Booster B1078, which completed its 26th flight in March 2026, represents genuine operational reusability: the vehicle is recovered, inspected, refurbished, and re-flown without significant reconstruction. The refurbishment cost per flight has fallen as SpaceX accumulated experience, and the marginal cost of each additional flight has declined accordingly.

Falcon 9 second stage is not recovered. This means roughly one third of the vehicle by mass is expended on every flight, and the economics that make Falcon 9 cost-effective depend on the first-stage savings being large enough to absorb second-stage expenditure.

Starship proposes full reuse of both stages. Seedhouse’s analysis identifies this as the critical variable in the cost equation. The projected $78–94 per kilogram figure that appears in most Starship economic projections assumes both the Super Heavy booster and the Starship upper stage are recovered and re-flown at high cadence. If upper stage recovery proves technically or economically impractical, the cost picture changes substantially.

The 2024 and 2025 test flights demonstrated booster catch by the mechanical arms of the Mechazilla launch structure, an engineering achievement that simplifies turnaround by eliminating the need for legs and post-landing transport. Upper stage ocean recovery was demonstrated at reduced fidelity. Full-rate upper stage recovery from orbital insertion, the actual operational scenario, remained as of early 2026 at the demonstration phase.

The Thermal Protection System: Where Physics Pushes Back

The most detailed section of Seedhouse’s book concerns the thermal protection system that allows Starship to survive reentry. This is where physics imposes its hardest constraints on reusability economics.

Reentry from orbital velocity exposes the vehicle to temperatures exceeding 1,400°C on the heat shield tiles. Falcon 9 first stage reentry, from a suborbital trajectory at much lower velocity and heating rates, is a comparatively mild thermal environment. Starship returning from orbit is in an entirely different regime.

The hexagonal silica tiles covering Starship’s windward surface serve the same function as Space Shuttle tiles, but at larger scale and with the constraint that they must be recovered, inspected, and re-flown at a cadence that makes economics work. Space Shuttle tile inspection and refurbishment was a primary driver of the Shuttle’s high operational cost, requiring thousands of person-hours per flight.

SpaceX has developed a tile formulation and application system designed to drive down that inspection burden, but Seedhouse’s analysis is appropriately measured on whether it can achieve the turnaround times the economics require. If each Starship reuse cycle requires days of tile inspection and replacement, the rapid reusability that justifies the cost projections is not achievable.

What the Cost Curve Means for Orbital Infrastructure

The reusable launch vehicle market analysis published earlier in 2026 documents the trajectory from the pre-Falcon 9 baseline of $9,000 per kilogram to Falcon 9’s current commercial rate of approximately $2,700 per kilogram. Starship’s target of under $100 per kilogram represents a further 27x reduction.

At that cost level, orbital infrastructure proposals that are currently economically incoherent become speculative-but-tractable. SpaceX’s FCC filing for a one-million satellite orbital data center constellation targets 100 gigawatts of AI compute capacity. Launching the satellite infrastructure for that constellation at Falcon 9 rates would cost hundreds of billions of dollars in launch costs alone. At Starship’s projected rates, it compresses into a range where projected compute revenue might justify it.

More immediately relevant are the LEO mega-constellation deployments already underway in 2026. Logos Space’s FCC-approved 3,960-satellite constellation, Amazon Kuiper’s full deployment plan, and SpaceX’s Starlink Gen 2 expansion all depend on launch economics that make deploying thousands of satellites a commercial activity rather than a government program.

The Manufacturing Bottleneck

Seedhouse identifies an underappreciated constraint: rocket reusability changes the economics of launch, but if satellite manufacturing does not scale in parallel, the bottleneck simply moves.

SpaceX’s Starlink factory in Redmond, Washington, currently produces approximately five satellites per day, targeting a rate needed to maintain and expand the constellation at Falcon 9 launch cadence. Deploying a million satellites requires either sustained production over decades or manufacturing scale that does not currently exist. The on-orbit servicing programs emerging in 2026, including Astroscale’s ELSA-M and the Tetra-5 refueling demonstration, hint at a future where satellites are maintained in orbit rather than replaced, which would reduce the manufacturing demand but introduces its own complexity.

Starship’s superior payload capacity, approximately 100,000 kg to LEO versus Falcon 9’s 22,800 kg, means each launch can carry significantly more satellite mass. But it also means that satellite manufacturing must produce satellites fast enough to fill those larger fairings at a pace that justifies the Starship production and operational investment.

Dragon: The Proven Infrastructure Beneath the Headlines

The book’s treatment of Dragon, while receiving less attention than Starship in the current space media environment, provides useful context. Dragon has completed 49 flights as of early 2026, including both cargo and crew configurations, with a demonstrated reliability record that matters for how we evaluate Starship’s trajectory.

Dragon’s development traced a path from demonstration flights through operational certification across approximately eight years. The structural lessons visible in that path, the gap between initial test success and operational reliability, the unexpected failure modes that only appear at higher flight rates, and the refurbishment complexity that emerges from real operational experience, are directly relevant to projecting Starship’s path.

Seedhouse applies Dragon’s track record as a calibration lens for Starship claims. The enthusiasm around Starship’s test successes is warranted. The projection of rapid operational capability from test flight data is the same reasoning that led to consistently optimistic Starship timelines in earlier years.

The Mars Mission Mass Budget

The final third of the book addresses Mars mission architecture, where the physics of reusability intersects with the extreme requirements of planetary transit. A Mars mission using Starship requires propellant production on Mars for the return journey, since carrying return propellant from Earth eliminates the mass advantage that makes the mission viable.

This drives the case for in-situ resource utilization: using Martian CO2 and water ice to produce methane and liquid oxygen, which are Starship’s propellants, on the surface. The future spacefaring society requirements analyzed elsewhere include this dependency explicitly. Permanent Mars settlement cannot be accomplished with Earth-supplied propellant any more than Antarctic stations could operate if every liter of fuel had to be shipped from the northern hemisphere.

The mass budget that Seedhouse documents makes this concrete. A crewed Starship configured for Mars transit carries approximately 100-150 metric tons of cargo. The return propellant requirement for that mission is approximately 900 metric tons of liquid methane and liquid oxygen. Producing that on Mars requires industrial-scale chemical processing equipment that currently does not exist, powered by nuclear reactors or large solar arrays, operating autonomously for months before crew arrival.

That scale of autonomous operations requires the edge computing architecture being developed for current satellite applications, adapted for surface deployment in a radiation and dust environment that is harder than the orbital environment current systems address.

Path Forward

Seedhouse’s conclusion is neither optimistic nor pessimistic, which is the intellectually honest position given the data. Starship’s reusability proposition is physically achievable. The thermal protection system challenges are engineering problems, not physics impossibilities. The cost projections are reachable if the engineering solutions perform as modeled.

The uncertainty is not whether the physics works, but whether the operational tempo needed to achieve the economics actually materializes and at what timeline. The history of launch system development, including Dragon’s own path, suggests that the gap between technical feasibility and operational reliability is measured in years rather than months.

For the orbital infrastructure ecosystem, that timeline gap matters. The edge AI satellite systems already operational in 2026 have been designed for current launch economics. The distributed AI coordination protocols being developed for multi-satellite constellations assume a cost environment that makes deploying hundreds or thousands of nodes viable. Whether Starship moves the economics to the point where millions of nodes become viable, and when, is the variable that will determine whether orbital computing scales to the level the most ambitious proposals envision.

Official Sources