The Gravity Tax: Reusability, Megaconstellations, and the Battle for the Edge of Orbit

1. Cold Open: The Ghost in the Pad and the Twelve-Minute Window

At Space Launch Complex 40 on the Florida coast, the most valuable object on the pad is the one that looks worst. A Falcon 9 first stage stands in the pre-dawn dark with its flanks streaked black, the soot baked on by hypersonic re-entry through its own exhaust plume, its titanium grid fins discoloured from previous descents. Some boosters in the fleet have made more than twenty of these round trips.¹² In the industry that existed thirty years ago, this hardware would have been unthinkable, because in the industry that existed thirty years ago the hardware did not come back.

That is the whole story of the modern launch business compressed into a single image. For six decades, an orbital rocket was a bespoke, hand-fitted machine assembled over eighteen months, flown once, and dropped into the ocean. Cost discipline was almost beside the point, because the customer was usually a government buying national capability rather than transport, and the contract usually reimbursed cost plus a fee. The launch vehicle was a consumable, like a shell casing, and it was priced like a small building.

The physics behind that arrangement is unforgiving and worth stating plainly, because everything downstream follows from it. Rocket performance is governed by the Tsiolkovsky rocket equation, which links the velocity a vehicle can gain to the ratio of its fuelled mass to its empty mass. To reach orbital velocity of roughly 7.8 kilometres per second, a chemical rocket must be somewhere between ninety and ninety-five per cent propellant by weight at liftoff. What is left over for structure, engines, avionics and the actual customer payload is a sliver: the payload typically represents two to four per cent of the mass that leaves the pad.

Now add the argument that legacy aerospace made for forty years. If you want the booster back, you must reserve propellant for a boostback burn, an entry burn and a landing burn; you must carry legs, grid fins and thicker structure; and you must accept a payload penalty on the order of thirty per cent. Given a payload fraction already down at three per cent, giving up a third of it looked like engineering vandalism. Every performance-optimising instinct in the industry said: throw the stage away and carry more satellite.

The instinct was correct about the physics and catastrophically wrong about the economics. Throwing away a first stage means writing off roughly thirty million dollars of manufactured hardware on every flight. Recovering it, inspecting it, and refuelling it with a few hundred thousand dollars of liquid oxygen and kerosene means that the same asset can be amortised across a dozen or twenty missions. The payload penalty is real; it is simply much smaller than the capital destruction it was designed to avoid. Once a first stage flies fifteen times, its per-flight cost falls by something close to ninety per cent, and the cost of goods sold for an entire mission falls by roughly seventy per cent.

The consequence is visible in the flight logs, which is the useful thing about this industry: it produces an unambiguous, publicly countable output. Rockets either reach orbit or they do not, and every object placed in orbit is catalogued.

Definition: orbital launch attempts by launching state, all operators, civil, commercial and military. Units are launches per calendar year. Geography is global. The 2026 line is an annualised run-rate extrapolated from activity through 24 August 2026 and is an estimate, not observed full-year data. Source: BryceTech global space activity reporting, cross-checked against the public satellite catalogue.¹²

Read that table aloud and two things jump out. The first is the slope. Global orbital launch activity has more than doubled in four years, from 146 attempts in 2021 to 325 in 2025, and is pacing above 360 for 2026.¹ The second is the reliability line. Launch rates rose and the failure rate fell at the same time, which almost never happens in an immature industry and is the signature of serial manufacturing replacing artisanal fabrication. The growth is also lopsided: the United States alone added more launches between 2021 and 2025 than the entire world conducted in 2021.

The centre of that American number is one privately held company. SpaceX flew 165 orbital missions in 2025 and had passed one hundred again by August 2026, which means a single firm accounts for roughly half of everything humanity launches.¹²⁵ Its two institutional rivals in the West were built on the opposite model. United Launch Alliance, the joint venture formed by Boeing ($BA) and Lockheed Martin ($LMT), exists to deliver national-security payloads with extraordinary reliability at prices the government was willing to pay for certainty.¹⁰ Arianespace and its industrial parent ArianeGroup, jointly owned by Airbus ($AIR) and Safran ($SAF), exist so that Europe can put a European satellite in orbit without asking anyone's permission.⁸ Neither was designed to win a manufacturing throughput contest, and both are now in one.

A word on the boundaries of this story. What follows is about orbital launch: the design and manufacture of propulsion, stages and avionics; the integration of payloads; and the delivery of those payloads to low Earth orbit, geostationary transfer orbit and beyond. It excludes suborbital tourism, which reaches altitude but never reaches orbital speed and is a different business with different customers. It excludes satellite operators, earth-observation analytics firms and broadband retailers, who are the customers of launch rather than the providers of it. The boundary matters because a great deal of capital has been lost by investors who treated everything with a rocket in the logo as one theme.

The theme itself has also moved out of its speculative adolescence. The 2020 to 2022 window, when cheap capital and blank-cheque vehicles funded dozens of launch startups, is over; the shakeout that followed it is largely complete; and what remains is an infrastructure business with countable cadence, disclosed backlogs and unforgiving fixed costs. The interesting question is no longer whether orbital access gets cheap. It is who, if anyone, gets paid for making it cheap.

Subscribe on LinkedIn

↑ Back to Top

2. The Upstream Bet: Industrializing the 500-Kilometer Shell

Picture the Earth wrapped in a thin shell of moving objects, most of them between roughly 500 and 600 kilometres up, each circling the planet every ninety minutes or so. Until about 2019 that shell was sparsely occupied by scientific instruments, spy satellites and a space station. It is now becoming the most contested piece of infrastructure real estate outside a subsea cable landing station.

The belief that makes this industry worth investigating can be written as one falsifiable proposition:

The continuous deployment and generational replenishment of low Earth orbit satellite constellations constitutes a durable, non-discretionary utility capital-expenditure cycle that permanently decouples global launch demand from civil government prestige budgets, and establishes high-cadence orbital transportation as a distinct, cash-generative industrial layer.

Note what that proposition contains. It contains a claim about demand durability, which is testable against launch counts and constellation filings. It contains a claim about the source of demand, which is testable against the mix of commercial versus government customers. And it contains a claim about industry structure, which is testable against the cash flows of the companies that actually fly rockets. The first two are doing well. The third is where the money is won and lost.

Start with the demand claim, because three independent evidence streams converge on it.

The first is terrestrial and demographic. Roughly two and a half billion people live outside the economic reach of fixed-line or reliable cellular broadband, and the marginal cost of serving them with towers and trenches rises as density falls. Maritime, aviation and remote-industrial connectivity have the same shape. Satellite constellations invert the economics: the capital is spent once, in orbit, and coverage is close to free at the margin. That is why satellite capacity has become the fastest-growing line in global telecommunications capital spending rather than a niche adjunct to it. The regulatory record shows the ambition. Industry trackers count more than 65,000 satellites sitting in active national and international filings across commercial systems such as Starlink, Amazon's Kuiper, AST SpaceMobile and Telesat's Lightspeed, and sovereign systems including China's Guowang and Qianfan networks, the European Union's IRIS² programme, and the US Space Development Agency's proliferated warfighter architecture.¹⁵ Filings are applications, not funded programmes, and the gap between the two is where a lot of thematic money goes to die. But even a large discount to that number implies an order of magnitude more hardware in orbit than has ever flown.

The second stream is a change in how customers buy. The old geostationary satellite was a five-hundred-million-dollar asset designed to work for fifteen years, ordered once a decade, insured to the hilt, and treated as a capital monument. The low-orbit node is a mass-manufactured, software-defined box with a three-to-five-year working life, ordered by the hundred, and treated as a consumable. Small satellites now account for well over ninety per cent of all spacecraft deployed, and commercial operators — not civil space agencies — procure the large majority of orbital launches.¹ This is the single most important behavioural fact in the industry, and it is not a preference. It is arithmetic. A satellite in a 550-kilometre orbit is slowly braked by residual atmosphere and will re-enter within a few years unless it is actively maintained; a constellation designed around five-year hardware must therefore replace something like a fifth of its fleet every single year, forever, or it degrades. A seven-thousand-satellite network on a five-year cycle generates a standing order of roughly 1,400 spacecraft a year before it grows at all.

That is the transmission mechanism into launch economics, and it is worth dwelling on. Replenishment demand is not a decision; it is a maintenance obligation, closer to replacing rolling stock on a railway than to funding a moon programme. It converts launch from an episodic, politically cyclical purchase into a recurring one. Recurring demand allows fixed launch infrastructure — pads, integration halls, engine factories, recovery vessels, mission control staff — to be amortised across a high annual flight rate, and it is that amortisation, far more than any single technical trick, that separates a rocket company that generates cash from a rocket company that consumes it.

The third stream comes from outside the industry, which is the most useful kind of corroboration. The US Department of Defense restructured its launch procurement around this belief, allocating more than $5.6 billion across the two lanes of its National Security Space Launch Phase 3 programme, and abandoning sole-source cost-plus contracting in favour of multi-provider, fixed-price manifests.⁶ Semiconductor and radio-frequency suppliers have retooled around direct-to-device standards that let ordinary handsets talk to satellites, which only makes sense if there is a permanent orbital layer to talk to. And in the specialty metals industry, the mills that pour nickel superalloys and aerospace titanium report multi-year order books dedicated to rocket propulsion lines, which is a physical, capital-committing vote on future flight rates.⁵

If this upstream belief is right, it implicates a family of neighbouring industries that this piece deliberately does not chase: satellite bus and payload manufacturing moving from cleanrooms to assembly lines, in-orbit servicing and orbital logistics, and the direct-to-device telecom layer where terrestrial carriers and constellation operators are converging.

What would break the belief? Three things, each observable. First, regulatory closure: if spectrum coordination or debris rules stop granting licences for large constellations, the standing order evaporates at the source. Second, longevity: if optical inter-satellite links, electric propulsion and edge computing push practical LEO spacecraft lifetimes past ten years, the replenishment cadence halves and with it the baseline demand. Third, and most subtle, oversupply: if heavy reusable capacity arrives faster than non-constellation commercial demand, the industry gets its volume and loses its pricing, which would confirm the demand half of the proposition while destroying the cash-generative half.

That third falsifier is the one that should worry an equity investor most, and it is a question about physics-driven cost curves. So the next thing to understand is why building a rocket engine is hard in a way that software is not, and what that hardness protects.

↑ Back to Top

3. Physics as Destiny: The Engineering Moats of Orbital Flight

Stand next to a rocket engine on a test stand at ignition and the numbers stop being abstract. Turbopumps spin at tens of thousands of revolutions per minute, forcing hundreds of kilograms of cryogenic liquid per second into a combustion chamber held at pressures that can exceed three hundred atmospheres, where the burning gas is far hotter than the melting point of the steel and copper alloys containing it. The chamber survives because a film of its own propellant is pumped through channels in the wall milliseconds before it burns. Everything in a launch business — cost curves, reliability, cadence, the ability to raise capital — is downstream of whether an organisation can make that arrangement work repeatedly.

Begin with the distinction that most people get wrong. Going to space and going to orbit are different problems separated by roughly an order of magnitude in energy. A suborbital vehicle climbs to a hundred kilometres, spends a few minutes above the atmosphere, and falls back down; it reaches perhaps Mach 3. An orbital vehicle must reach about 7.8 kilometres per second horizontally, near Mach 25. The useful analogy is an apple: throw it upward and it comes back; throw it sideways fast enough and it keeps falling, but the curve of the Earth falls away underneath it at the same rate, so it never lands. Where the analogy breaks is the return trip. All that kinetic energy has to be shed on the way back, and it is shed as heat — thousands of degrees of ionised plasma around the vehicle. That is why a reusable orbital upper stage is a fundamentally harder engineering problem than a reusable first stage, and why almost every "reusable" rocket flying today still throws away its second stage.

The engine cycle is the next fork in the road, and it determines both efficiency and reusability. A gas-generator engine, such as the Merlin that powers Falcon 9, burns a small side stream of propellant to spin its turbopumps and dumps that exhaust overboard; it is simple, robust, cheap to build, and slightly wasteful. An expander cycle, such as the RL10 that has flown on American upper stages for six decades, uses heat from the chamber walls to drive the pumps; it is beautifully efficient and thrust-limited, which makes it an upper-stage engine rather than a booster engine. Staged combustion routes all of the turbine exhaust back into the main chamber so nothing is wasted, at the price of feeding hot, chemically aggressive gas through the pumps. Full-flow staged combustion, used in SpaceX's Raptor, runs two separate preburners — one oxygen-rich, one fuel-rich — so that every molecule of propellant passes through a turbine and then into the chamber. It delivers the highest chamber pressures in service and, critically, keeps turbine seal temperatures lower, which is what makes an engine re-flyable rather than merely efficient.¹²

Fuel chemistry has undergone an equally consequential shift. Refined kerosene, the RP-1 that powers Falcon 9 and most Chinese and Russian boosters, is dense and easy to handle but leaves carbon deposits — coking — inside injectors and cooling channels, which means inspection and cleaning between flights. Solid propellant is simple and storable but cannot be throttled or shut down, and it deposits alumina in the upper atmosphere. Liquid methane burned with liquid oxygen, universally called methalox, sits in the sweet spot: cleaner than kerosene with essentially no soot, higher performing, easy to re-ignite in flight, and cold enough to help cool the engine on its way in. That is why methalox has become the default for nearly every next-generation reusable vehicle in the world at once — SpaceX's Starship, Rocket Lab's Neutron, Blue Origin's New Glenn, Relativity's Terran R and LandSpace's Zhuque-3.^1241119 When five independent engineering organisations on two continents converge on the same propellant, it is usually because the physics left them no choice.

Now the strategic point. None of this can be shortcut with simulation. Combustion instability — the phenomenon where pressure oscillations inside a chamber couple with the flame front and destroy the engine in milliseconds — remains only partly predictable from first principles. It is characterised empirically, by building injectors, firing them, breaking them, and building the next one. That accumulated test-stand record is a genuine cornered resource: it exists as data, tooling and the tacit knowledge of a few thousand engineers, and it cannot be bought, licensed or reverse-engineered from a competitor's photographs. It is also why engine development timelines are measured in five-to-eight-year increments regardless of funding, and why a well-capitalised entrant can be simultaneously rich and years behind.

Layer the other durable advantages on top and the moat structure of this industry becomes legible in Hamilton Helmer's vocabulary, though the mechanisms matter more than the labels.

Process power is the SpaceX story: not one invention but the compounding of manufacturing, test and flight-operations practice across hundreds of missions, encoded in tooling and procedures rather than patents. Scale economies operate through fixed research and development, because the several hundred million dollars needed to develop autonomous flight termination — the onboard system that decides for itself whether a vehicle has left its safe corridor and destroys it, replacing the human radar operators who used to gate range turnaround — is the same whether you fly twice a year or a hundred times. Cornered resources appear in unexpected places: coastal launch sites with clear downrange azimuths, deep-water port access and completed environmental approvals are geographically finite, and so, in the near term, are qualified engines. Counter-positioning explains why the incumbents did not simply copy reuse. For a contractor whose revenue is reimbursed on cost, an innovation that removes billions of dollars of recurring hardware manufacture from the income statement is not an opportunity; it is a subtraction. The economically rational incumbent response was to wait, and waiting proved fatal to their commercial share.

Read the same landscape through Porter's forces and the tension in this industry sharpens. Supplier power is unusually high in propulsion, where sole-source relationships bind entire vehicle programmes to one vendor's factory throughput. Buyer power is concentrating fast, because a handful of customers — the US Space Force, Amazon, and a small set of sovereign constellation programmes — represent most of the non-captive Western manifest. Barriers to entry are enormous in capital and in time, but they are not absolute, as several Chinese entrants have demonstrated. The most underrated force is substitution: for the small-satellite customer, the substitute for a dedicated rocket is a seat on somebody else's much larger rocket, and that substitute turned out to be roughly a fifth the price. Rivalry, meanwhile, is muted in the only place it would matter — nobody is currently able to contest SpaceX's cost position — and vicious everywhere else.

Physics, in other words, sets the entry fee, and the entry fee is paid in years and in test stands. The question that follows is where, along the chain of firms that turns metal into orbital velocity, the resulting profit actually settles.

↑ Back to Top

4. The Value Chain: From Superalloys to the Orbital Drop

Follow a single ingot. It begins in a vacuum arc remelting furnace at a specialty mill in Pennsylvania or Ohio, where nickel, chromium and a short list of expensive additions are melted under vacuum to strip out the inclusions that would become crack initiation sites under load. It is shipped as bar or powder to a factory in California, where it is either machined or laser-fused, layer by layer, into an injector head with hundreds of internal passages that no conventional machining could produce. That injector becomes part of an engine; the engine is bolted to a tank; the tank is trucked to a coastal integration hangar; and about nine minutes after ignition, that ingot is travelling at 7.8 kilometres per second. Every stage of that journey has different economics, and understanding which stage keeps the money is the practical core of investing in this theme.

Stage one: raw materials and cryogenics. Aerospace-grade titanium, nickel superalloys such as Inconel 718 and 625, lithium-aluminium plate for propellant tanks, and carbon fibre prepreg are supplied by a small set of qualified mills — ATI ($ATI) and Carpenter Technology ($CRS) in specialty metals, Hexcel ($HXL) in composites — into essentially every Western launch programme, SpaceX, Rocket Lab and Blue Origin included.⁵ Gross margins here run in the high teens to high twenties. The qualification process is real, but buyers can usually dual-source over time, so pricing power is moderate rather than structural. The other input is the propellant itself. Air Liquide ($AI) and Linde ($LIN) supply bulk liquid oxygen, liquid methane, liquid hydrogen and helium into the propellant farms at Cape Canaveral, Vandenberg, Kourou and other ranges.⁵ Their leverage is geographic rather than technological: once a cryogenic plant and pipeline serve a spaceport, the incumbent effectively holds a local monopoly, and helium in particular is a genuinely scarce commodity with recurring supply squeezes.

Stage two: propulsion, the chokepoint. This is where margins jump to somewhere between thirty and forty-five per cent, and where the industry's real dependencies live. Blue Origin supplies the BE-4 methalox engine to United Launch Alliance, two per Vulcan Centaur first stage, under a long-term original-equipment agreement; this is a confirmed commercial relationship, not an inference, and it is sole-sourced.¹⁰¹¹ Aerojet Rocketdyne, now a business of L3Harris ($LHX), supplies the RL10 hydrogen-oxygen upper-stage engine for Vulcan's Centaur V, a relationship stretching back decades and with no qualified Western alternative.¹⁰ Northrop Grumman ($NOC) supplies the GEM-63XL solid strap-on boosters that Vulcan bolts on for its heavier configurations, and also builds the large solid motors for NASA's Space Launch System and for American strategic missiles.¹⁸ In Europe, Avio ($AVIO) manufactures the P160C solid motor that serves as both the Ariane 6 strap-on booster and the first stage of Vega C, supplied to ArianeGroup and Arianespace through the Europropulsion industrial arrangement, and protected by European Space Agency policy rather than by patent.⁹⁸

Look at that list again and note what it means for one company. United Launch Alliance buys its first-stage engines from one supplier, its upper-stage engines from a second, and its solid boosters from a third, and cannot substitute any of them on a useful timescale. Its vehicle is, in cash-flow terms, an integration and mission-assurance business wrapped around three external monopolies. The propagation runs downhill: constrained BE-4 deliveries limit Vulcan production, Vulcan production limits ULA's flight rate, and ULA's flight rate is one of the three pillars — alongside Ariane 6 and New Glenn — carrying Amazon's Kuiper deployment schedule, which is itself governed by regulatory milestones. A turbopump problem in Washington State becomes a constellation problem in Seattle and a licensing problem in Washington DC. This is the single most important chokepoint map in the Western industry, and it is why "who supplies whom" is not a due-diligence formality here.

Stage three: airframes and structures. Tanks are the largest single piece of a rocket and the most process-dependent. The American and Chinese mainstream builds them from aluminium-lithium alloy joined by friction stir welding — a solid-state process that stirs two plates together below melting point, avoiding the porosity of fusion welds — while SpaceX moved Starship to stainless steel precisely because it retains strength when cryogenically cold and when re-entry-hot. Rocket Lab took the other road entirely, using automated fibre placement to lay carbon composite for Electron and Neutron, trading material cost for mass and for the ability to build a stage without an enormous metal-forming shop.⁴ Margins here run twenty to thirty-five per cent and are almost entirely a function of how many units the tooling produces per year.

Stage four: avionics, guidance and flight software. Radiation-tolerant flight computers, inertial measurement units, navigation receivers and actuator control come partly from established aerospace electronics houses such as Honeywell ($HON), BAE Systems and Moog, and increasingly from in-house teams at the vertically integrated players. Margins are the highest in the physical chain, thirty-five to fifty per cent, because the differentiating asset is software. Propulsive landing is a control problem before it is a hardware problem: the guidance algorithm that solves, in real time and repeatedly during descent, for a fuel-optimal trajectory to a moving barge is proprietary, and it is the thing a competitor cannot photograph.

Stage five: ground infrastructure. Flame trenches, cryogenic farms, umbilical towers, transporter-erectors, catch mechanisms and recovery vessels are engineered and built by large construction and engineering contractors, and they are low-margin work — fifteen to twenty-five per cent — attached to an extremely high-value asset. The scarcity sits with the site, not the builder. There are only so many coastlines with a clear eastward or polar launch corridor, an existing range safety organisation, and a completed environmental record. Cape Canaveral and Vandenberg in the United States, Kourou in French Guiana with its equatorial velocity bonus, Mahia in New Zealand, Tanegashima in Japan, Sriharikota in India, and Jiuquan and the Hainan commercial site in China are the fixed points around which everything else arranges itself.

Stage six: launch services. This is the revenue-recognition event — the contracted delivery of a payload to a specified orbit — and the range of possible gross margins is enormous, from ten per cent to forty per cent, depending almost entirely on how many times the fixed asset base flies. It is also where regulation binds. In the United States, every commercial launch requires a licence under the Federal Aviation Administration's Part 450 framework, and every new or modified site drags in environmental review.⁷ Payload integration, cleanroom encapsulation, range clearance and mission control are the visible work; the invisible work is a reliability record, which is the actual product being sold to anyone flying a four-hundred-million-dollar spacecraft.

Stage seven: the spacecraft itself. Here is where the profit pool has been quietly migrating. Satellite buses, solar arrays, reaction wheels, star trackers, separation systems and optical crosslinks carry twenty-five to thirty-eight per cent gross margins and, crucially, are sold into the same customer budget as the launch. Rocket Lab is the clearest case: it sells engines, reaction wheels, star trackers, solar panels and separation mechanisms both into its own rockets and, as a merchant supplier, to third parties including defence satellite programmes and in-space manufacturing customers.³⁴ SpaceX's equivalent is Starlink, which is not a merchant business at all but a captive one. Lockheed Martin and Northrop Grumman build the large defence spacecraft that fly on other people's rockets.

Two directions of travel emerge. Value is flowing toward super-heavy reusable logistics, where sheer scale drives a cost position nobody else can match. And value is flowing toward the spacecraft layer, where high-margin component and bus sales can subsidise an aggressive launch price. The middle of the chain — building and flying a medium rocket, and nothing else — is the worst place to stand. That was not obvious in 2021, and the reason it became obvious is a history worth telling.

↑ Back to Top

5. History and the Three Eras: From State Prestige to the SPAC Hangover

In January 2006, NASA under administrator Mike Griffin announced a programme called Commercial Orbital Transportation Services. It was, on its face, a modest procurement reform: rather than paying a contractor's costs plus a fee to develop a cargo vehicle, NASA would pay fixed sums on completion of defined milestones, and the contractor would own the resulting vehicle. The agency put seed money behind two firms, one of which was a four-year-old startup that had not yet reached orbit. That decision reorganised the industry more thoroughly than any technology of the following decade, because it changed who bore development risk and who kept the upside.

Era one: the cost-plus state monopoly, 1957 to 2006. Launch vehicles descended directly from ballistic missiles and inherited their procurement culture. Governments specified, contractors built, and cost was reimbursed. Nobody involved had an incentive to reduce the recurring cost of the hardware, because the recurring cost was the revenue. The era's closing act was structural: in 2006 Boeing and Lockheed Martin merged their competing Delta and Atlas rocket lines into United Launch Alliance, a fifty-fifty joint venture, on the argument that the American national-security market was too small to support two suppliers.¹⁰ The result was reliable access to orbit at prices in the region of $150 million to $350 million per flight, and a genuine capability that the country depended on. It was also a duopoly of two firms that had merged into one, facing a single customer, with no competitive pressure on cost whatsoever.

Era two: the reusability revolution, 2006 to 2020. SpaceX reached orbit with Falcon 1 in 2008 and flew Falcon 9 for the first time in 2010. The vehicle was competitive on price from the start, but the structural break came in December 2015, when a Falcon 9 first stage delivered eleven ORBCOMM satellites to orbit and then flew itself back to a landing pad at Cape Canaveral.¹²⁵ The immediate significance was not the cost saving, which took years to materialise; it was that the argument was settled. Propulsive recovery of an orbital-class booster was demonstrably possible, and every expendable vehicle in development anywhere in the world instantly became a depreciating asset. Two and a half years later, in early 2018, Rocket Lab put its Electron into orbit from a private launch site on New Zealand's Mahia Peninsula, using 3D-printed engines driven by electric turbopumps instead of gas turbines, and proved that a small company could build a small rocket that worked.⁴

Era three: the bubble and the hangover, 2020 to 2024. Cheap capital met a plausible story. If satellites were getting smaller and more numerous, the reasoning went, thousands of them would want dedicated rides to precise orbits, and dozens of small launch vehicles would be needed to carry them. More than forty launch startups were funded globally, many via blank-cheque acquisition vehicles that allowed pre-revenue companies to list on public markets with projections attached. The reasoning contained one fatal omission: the customer's alternative. SpaceX began flying dedicated rideshare missions that sold space on a Falcon 9 by the kilogram, at a price around a fifth of what a dedicated small rocket had to charge to break even.¹² For roughly nine payloads in ten, orbital precision was worth less than an eighty per cent discount.

The correction was swift and total. Virgin Orbit, which air-launched rockets from a modified airliner, filed for Chapter 11 bankruptcy in 2023 and liquidated its launch assets; part of its Long Beach facility and tooling ended up with Rocket Lab.⁵⁴ Astra Space burned through its public listing, delisted, and was restructured away from orbital launch.⁵ Small-launch venture funding effectively closed. What survived did so by being something else as well: Rocket Lab had a components and spacecraft business, and Firefly had a lunar lander programme and a defence customer.

Overlaid on all of this was a geopolitical rupture that reshaped the Western manifest overnight. When Russia invaded Ukraine in February 2022, Soyuz and Proton disappeared from the Western commercial market. OneWeb lost its rides mid-deployment. Europe lost its backup option at precisely the moment its own capability was at a low ebb: Ariane 5 was retiring, Ariane 6 was late, and Vega C had been grounded after a failure.⁸ For roughly two years, Europe and much of the Western commercial world had one realistic path to orbit for anything heavy, and it belonged to a private American company. Sovereign launch autonomy stopped being a talking point in ESA ministerial documents and became a live strategic problem.

Institutions are not scenery in this industry; they are causal actors, and three of them deserve naming.

The Federal Aviation Administration's Office of Commercial Space Transportation licenses every American commercial launch. Its Part 450 rule was designed to streamline approvals by replacing prescriptive requirements with performance-based safety criteria.⁷ In practice, streamlining licensing collided with a cadence increase nobody had planned for, and with environmental review obligations that attach to new sites and modified operations. The bottleneck in American launch has repeatedly been paperwork and environmental litigation rather than hardware, which is an unusual sentence to write about rocketry and an important one for anyone modelling flight rates.

The European Space Agency operates on a principle known as geographical return, or juste retour: each member state receives industrial work roughly in proportion to what it contributes. It is the reason ESA exists as a functioning multinational body, and it guarantees political durability of funding. It also means a European launch vehicle is designed around a workshare map — engines from France, solid motors from Italy, structures from Germany — rather than around minimum cost. Against a vertically integrated competitor manufacturing eighty-five per cent of its vehicle under one roof, that is a permanent handicap, and everyone involved knows it. The response has been to protect the institutional market: European public payloads fly on European rockets, which converts a commercial disadvantage into a guaranteed revenue floor.⁸

China runs a deliberate dual-track system. A 2014 State Council policy document opened the space sector to private capital, and subsequent five-year planning designated commercial aerospace a strategic emerging industry.⁵ The state group CASC retains the crewed programme, lunar exploration and national security missions and flies the Long March family. Alongside it, a cohort of private firms — LandSpace ($LANDSPACE), Space Pioneer ($SPACEPIONEER), Galactic Energy ($GALACTICENERGY), OrienSpace ($ORIENSPACE) — receives state-linked venture funding, access to launch ranges including the new commercial site on Hainan, and anchor contracts to deploy the sovereign Guowang and Qianfan constellations.¹⁹²⁰ The framing in official policy is civil-military fusion and industrial upgrading; the independently verifiable facts are the flight logs and the hardware. Financial disclosure is the weak point. These companies do not file audited accounts, and their capitalisations and timelines are reported through domestic technology press rather than regulators, which means capability claims should be weighted toward demonstrated flights and away from announcements.²⁰

Finally, and quietly, export control shapes everything. American rocket propulsion and guidance hardware is regulated as munitions under ITAR, and the Missile Technology Control Regime constrains international transfer of launch technology generally. The practical effect is that there is no single global launch market. There is a Western market and a Chinese market, and they do not trade with each other.

Every structural break in this industry has come from a change in how a government buys, or from a geopolitical shock that removed a supplier. What the breaks left behind is the field as it stands today.

↑ Back to Top

6. The Field in 2026: Hegemons, Sovereign Champions, and Challengers

Take any ordinary week in 2026. Several Falcon 9s lift off from Florida and California, most of them carrying batches of the operator's own broadband satellites. An Electron departs from New Zealand or Virginia with a single customer's spacecraft. A Vulcan Centaur carries a national security payload. Somewhere in Gansu or on a barge off Shandong, a Long March or a commercial Chinese vehicle goes up. Ariane 6 flies from French Guiana every few weeks. That distribution — one operator flying more often than everyone else combined, a scatter of sovereign programmes, and a thin layer of commercial challengers — is the market structure, and it shows up most starkly not in flight counts but in mass.

Definition: gross mass of spacecraft and cargo successfully delivered to orbit, in metric tonnes, by operator group; global. 2026 is an estimate annualised from activity through 24 August 2026, not observed data. Source: BryceTech global space activity reporting.¹

The line to read aloud is the third column. While one operator's delivered mass grew more than sixfold in five years, everyone else in the world outside China added almost nothing in aggregate — from about 180 tonnes in 2021 to an estimated 260 tonnes in 2026. Global upmass more than quadrupled, and roughly four-fifths of the increase belongs to a single company. Any thematic thesis that assumes a rising tide lifts the sector has to explain that column.

The hegemon. SpaceX leads on the two parameters that compound: annual flight cadence and delivered mass, as of August 2026. It flew 165 orbital missions in 2025 and had passed one hundred by late August 2026, delivering roughly eighty per cent of global upmass.¹²¹ Why it leads is a three-part answer, and each part reinforces the others. Propulsive recovery gives it a manufactured asset it can use twenty-plus times, with pad-to-pad booster turnaround compressed to a few weeks. Vertical integration — engines, avionics, tanks, software, and the recovery fleet, largely in-house — removes the margin that each external supplier would otherwise stack onto every unit, and shortens iteration from quarters to weeks. And Starlink supplies captive anchor demand: a hundred-plus missions a year that the company sells to itself, at whatever internal transfer price it chooses, which absorbs the fixed cost of pads, factories and staff before a single external customer is invoiced. That last element is the one competitors cannot replicate, because it requires simultaneously operating the largest satellite constellation in history. What could erase the lead is narrower than it looks: a sustained grounding after a failure, a regulatory constraint on cadence, or a strategic misstep on Starship. Competitive imitation is not on the near-term list.

The public challenger. Rocket Lab ($RKLB) is the only scaled, listed, pure-play launch company in the Western world, which makes it carry more thematic weight than its size warrants. It leads a defined and defensible parameter: commercial dedicated small-satellite launch, where Electron has flown more than fifty-five successful missions at an annualised cadence in the high teens.⁴³ Its closest rivals on that parameter are Chinese — Galactic Energy's solid-fuelled Ceres-1 and the Kuaizhou family — and they do not compete for Western commercial or defence payloads because of export control. The lead came from three founding choices: carbon composite structures that let a small company build stages without heavy metal forming, the Rutherford engine's electric turbopumps that traded a modest efficiency penalty for manufacturing simplicity and 3D printing, and a privately owned launch site that removed range scheduling as a constraint. The honest qualification is that leading the small dedicated launch niche is a limited prize, because the niche itself is capped by rideshare substitution. Rocket Lab's answer is Neutron, a thirteen-tonne-class reusable medium-lift vehicle powered by its own Archimedes methalox engine, with a maiden flight targeted for the fourth quarter of 2026.⁴ Everything about the company's forward economics turns on that vehicle.

The defence duopolist. United Launch Alliance leads on a different parameter entirely: assured access for the most demanding American national security orbits. Its Vulcan Centaur is in operational ramp, and the Centaur V upper stage — high-energy hydrogen-oxygen, capable of long coast phases and multiple restarts — does something that kerosene and methane upper stages do badly, which is deliver payloads directly into difficult high-energy orbits.¹⁰ The customer cares because the payloads are irreplaceable and the orbits are specified by mission requirement rather than convenience. ULA leads there because of accumulated institutional capability: classified payload handling, vertical integration facilities, security infrastructure, and a mission success record that is the industry's benchmark. Its lead is protected by policy — the Department of Defense has consistently maintained more than one certified provider for its most critical missions — and it is vulnerable to exactly one thing, which is its own cost structure and its dependence on three external propulsion suppliers.⁶

The deep-pocketed challenger. Blue Origin occupies two positions at once. It is a launch operator, having reached orbit with New Glenn's first flight in January 2025 and holding a substantial commercial and civil backlog including Kuiper missions and NASA lunar work.¹¹⁵ And it is a merchant propulsion supplier, selling BE-4 engines to its competitor ULA. Its constraint has been operational velocity rather than money or engineering ambition; damage at its Cape Canaveral pad complex has weighed on the manifest, and Blue Origin does not publish detailed repair schedules, so return-to-flight estimates in the trade press are inferences drawn from customer manifest changes rather than company disclosure.⁵

The European sovereigns. ArianeGroup, through the Arianespace marketing entity, operates Ariane 6 from Kourou with a multi-year institutional and commercial book including European public payloads and Kuiper missions.⁸ It leads on one parameter — guaranteed European access to orbit independent of any other state — and that lead is entirely institutional, conferred by treaty and budget rather than won on cost. Ariane 6 is expendable, and against a reusable competitor it cannot win an open commercial tender on price. Development work on the Prometheus reusable methalox engine and the MaiaSpace light launcher represents the acknowledgement of that. Avio, the listed Italian firm at Colleferro, is a cleaner expression of the same institutional protection: it is prime contractor for Vega C and the sole European source of large solid rocket motors, and it is simultaneously expanding into American and European tactical missile propulsion, where NATO restocking demand is independent of the launch cycle.⁹

The Asian field. Mitsubishi Heavy Industries ($7011) operates Japan's H3 in its several configurations from Tanegashima, with a target cadence of a handful of launches a year serving JAXA and commercial customers; it leads Japanese sovereign access and is constrained by range capacity and unit cost rather than by capability.¹⁷ India's NewSpace India, the commercial arm of ISRO, leads on a genuinely distinctive parameter — lowest cost per mission among reliable Western-accessible providers — flying PSLV, the heavier LVM3 and the small SSLV, and having deployed OneWeb spacecraft commercially. Its constraint is throughput, not price.

China's structure is two-tier. CASC flies the Long March family at fifty to sixty missions a year and carries the space station, lunar and national security programmes on an effectively unlimited state balance sheet, walled off from Western commercial markets by export control.¹ The commercial tier is where the interesting engineering is happening. LandSpace put the world's first methalox rocket into orbit with Zhuque-2 and has been developing the stainless-steel, reusable Zhuque-3, conducting booster recovery testing during 2026.¹⁹²⁰ Space Pioneer flew the kerosene Tianlong-2 and brought the much larger Tianlong-3 — a vehicle roughly in Falcon 9's payload class — to a first launch in 2026. Galactic Energy has built the highest cadence of any Chinese private firm on the solid-fuelled Ceres-1, launching from land and from sea, and is developing the reusable kerosene Pallas-1; its vulnerability is that solid propellant unit economics do not work for megaconstellation deployment. OrienSpace flies Gravity-1, an unusually large solid vehicle launched from a sea platform off Haiyang, bypassing congested inland ranges, with liquid Gravity-2 and Gravity-3 in development. On the parameter of commercial methalox reusability, leadership is genuinely shared and contested between SpaceX and LandSpace, with the American company far ahead on operational flight experience and the Chinese company having demonstrated the core recovery technology.¹⁹

The medium-lift aspirants. Three American companies are attempting the transition that killed the small-launch cohort. Firefly Aerospace flies the small Alpha at low cadence, holds responsive-launch contracts with the US Space Force, and is co-developing a Medium Launch Vehicle with Northrop Grumman, which needs a successor to its retired Antares.⁵¹⁸ Relativity Space abandoned its small Terran 1 after a single test flight and is building the much larger reusable methalox Terran R around its proprietary large-format 3D printing; it holds commercial launch agreements and has been burning cash for years without flying.⁵ Stoke Space is attempting the hardest version of the problem, a fully reusable vehicle with a regeneratively cooled metallic heat shield on the upper stage, and remains in development.⁵ None of the three is yet an operating business in the sense that matters, which is flying paying payloads repeatedly.

That is the field. Now the question that decides who among them survives: what does it actually cost to fly?

↑ Back to Top

7. The Economics of Fire: Unit Costs, Fixed-Cost Absorption, and the Big-Market Delusion

Two invoices, same destination. The first is for a traditional expendable medium-lift rocket: the customer pays roughly $65 million, and every component on that invoice — first stage, second stage, fairings — ends up on the seabed. The second is for a flight-proven reusable booster rolling out for its fifteenth mission, where the largest single item on the bill of materials was expensed years ago and is being amortised a fifteenth at a time. Laying the two side by side is the most clarifying exercise in this industry.

Definition: modelled per-flight cost of goods sold for representative vehicle architectures, in current US dollars, excluding company fixed overhead, research and development, and depreciation of facilities. This is an analytical construction built from disclosed list prices and published cost estimates, not audited segment reporting; SpaceX does not publish financial statements. The third column represents a design target for a next-generation medium reusable vehicle, not an achieved result. Source: assembled from public list pricing and specialist trade reporting.¹²⁵⁴

Read that aloud and the temptation is to say reusable rockets earn seventy-seven per cent gross margins. They do not, and the difference between contribution margin and gross margin is where most of the analytical errors in this sector live. Every number in that table is a variable cost. None of it includes the pads, the factories, the test stands, the engineering payroll, or the depreciation of a billion dollars of development. Those fixed costs do not scale with flights; they arrive whether you launch twice or fifty times.

Put a number on them. An active launch provider with its own site, integration facility and engine production carries somewhere between $150 million and $300 million a year in fixed operating cost before it lights anything. Divide that by two launches a year and you have added over $100 million per flight to a mission that sells for $65 million. There is no unit-economics improvement that recovers from that. Divide it by ten launches and the fixed burden per flight falls to $20 to $30 million, which a reusable vehicle's contribution margin can just about carry. Divide it by twenty and the burden falls to $10 million or less, at which point almost the entire contribution margin drops to the operating line, and cash generation goes vertical.

That is the flywheel, and it is the reason this industry has a violent, discontinuous relationship between cadence and profitability. There is a threshold — call it eight to twelve flights a year for a medium vehicle — below which the business is structurally loss-making regardless of how elegant the rocket is, and above which it becomes a very good business quite suddenly. Every company in the sector is either above that line, walking toward it, or dying quietly below it. Virgin Orbit and Astra died below it. It also explains why captive demand is so valuable: an operator that can generate a hundred internal missions a year has crossed the threshold before selling anything, and can then price external launches at whatever level maximises share, because those flights carry no fixed-cost burden at all.

A necessary caution about the industry's favourite statistic. "Dollars per kilogram to orbit" is quoted constantly and defined inconsistently. Divide a Falcon 9's list price by its maximum expendable payload capacity and you get something near $3,000 per kilogram. Divide it by the payload it can actually carry while recovering the booster, on a real mission with real orbital requirements, and the figure rises. Ask what a small satellite operator actually pays for a rideshare seat and the answer in 2026 is roughly $5,500 to $6,500 per kilogram.¹² Marginal internal cost is different again; external estimates put SpaceX's internal Falcon 9 mission cost in the mid-teens of millions of dollars, but the company publishes no financial statements and the transfer price it charges its own Starlink division is undisclosed, so any figure quoted for its launch-segment profitability is an inference.⁵ When one source says access costs $300 per kilogram and another says $6,000, they are usually both right about different things: theoretical bulk capacity versus the price an actual customer pays for an actual seat. Structural cost deflation is real, and it has been of the order of seventy-five per cent or more since the early 2000s; the precise figure depends entirely on which of those definitions you choose.

Which brings us to the belief that has cost thematic investors the most money. Call it the big-market delusion.

The myth: the space economy is heading toward a trillion dollars in annual value, launch is the gateway to all of it, and therefore any competent launch company will be carried to prosperity by the size of the market.

The reality, tested against the numbers already on the page: total addressable commercial and unclassified civil-defence launch revenue in 2026 runs somewhere around $15 billion to $16.5 billion globally.¹ That is the entire industry — every rocket, every country, every customer — and it is smaller than the annual revenue of a mid-sized speciality retailer. A trillion-dollar space economy is a statement about downstream services: broadband subscriptions, navigation embedded in every device, imagery analytics, defence communications. Launch is the toll road underneath it, and toll roads capture a small fraction of the value of the commerce that travels on them.

Where does the surplus go? Overwhelmingly, downstream. Every dollar of launch cost deflation is a dollar that ends up with a rural broadband subscriber, a shipping company, a cloud provider, or a defence ministry that can now afford a proliferated architecture instead of a handful of exquisite satellites. That is a wonderful outcome for society and a difficult one for the equity holder of the company that produced it. The classic industrial parallel is airlines: consumers captured almost all of the value created by jet aviation, and the operators, over the industry's history, destroyed capital.

There are two further drains specific to this sector. The first is the reinvestment treadmill. Rocket architectures obsolete each other. A company that finally reaches cash generation on a vehicle immediately faces the choice of harvesting it or spending the proceeds on a larger, cheaper successor before someone else does — Falcon 9 cash funding Starship, Electron cash funding Neutron. The cash exists; it just does not become distributable for a very long time. The second is dilution. Development programmes that run $500 million to $2 billion, financed in equity markets by companies with negative free cash flow, transfer a meaningful share of any eventual success from existing holders to new ones. Per-share value, not enterprise value, is the metric that matters, and the two can diverge for years.

So the correct question for a thematic investor is not whether orbital access is becoming an industrial utility. The flight logs settle that. The question is which layer of the chain, in which capital structure, at which price, actually keeps a durable share of the economics. That question has a short list of publicly available answers.

↑ Back to Top

8. Public Expressions and the Portfolio Paradox: Where Can Equity Capital Go?

Here is the awkward shape of the problem. An investor becomes convinced that orbit is being industrialised, that the evidence is strong, and that the trend has a decade to run. They then discover that the company delivering roughly eighty per cent of the world's orbital mass is private, valued in mid-2026 secondary transactions somewhere in the range of $210 billion to $250 billion, and shows no interest in a listing.⁵ The theme's protagonist is unavailable. What remains is a short list of partial expressions, each with a different defect.

Before drawing security-level implications, the decision context has to be stated, because the same facts imply different actions for different mandates. What follows assumes a general institutional public-equity thematic allocation, a multi-year horizon, a global listed opportunity set, and no ability to access private rounds. It contains no position sizing and no buy or sell recommendations. A long-only investor measured against a broad benchmark and a long/short investor targeting absolute return face genuinely different versions of this problem: the former's main risk is owning an expensive stock that de-rates, the latter's is being short a stock whose optionality resolves favourably in a single quarter.

Definition: most recent reported figures as of 24 August 2026, mixing one quarterly period (Rocket Lab) with three annual periods, and mixing segment reporting (Lockheed Martin, MHI) with consolidated company reporting (Rocket Lab, Avio). Currencies are as reported and not converted for the growth and margin columns. These are not rankings; the businesses are not economically comparable on any single line. Sources: SEC filings, Avio investor disclosure, Lockheed Martin and MHI investor reporting.³⁹¹³¹⁷

Reading that table honestly requires saying what it does not show. Rocket Lab's 31.5 per cent gross margin and Avio's 14.2 per cent are not evidence that one company is twice as good as the other; they are evidence that a components-and-spacecraft business and a solid-propulsion manufacturing business have different cost structures, and that Avio's margin was compressed by raw material inflation. Lockheed Martin's Space segment number includes equity earnings from a joint venture, not consolidated launch revenue. And SpaceX, the company that matters most, appears nowhere, because it discloses nothing.

The pure play: Rocket Lab. The exposure link is unusually clean. The company reports Launch Services and Space Systems as separate revenue lines, disclosed backlog is contracted, and there is no ambiguity about whether the theme touches its cash flows.³ Second-quarter 2026 revenue of $234 million grew sixty-two per cent year over year, with roughly seventy per cent of it from Space Systems. Backlog stood at a record $2.36 billion, weighted toward spacecraft rather than launch. The company is loss-making at the operating line, with the deficit substantially explained by research and development running near $45 million a quarter on Neutron, and it is burning roughly $60 million a quarter in free cash flow against more than $850 million of cash and short-term investments.

The market's stated case against the stock is straightforward and should be taken seriously: it trades at a very high multiple of sales — a market capitalisation in the mid-$40 billion range in August 2026 against roughly $1 billion of trailing revenue implies something close to forty times forward sales — for a company that does not yet make money and whose flagship vehicle has never flown.³ That is the first reason a senior investor rejects it, and it is a good one. A single valuation observation makes the point about dispersion: on these figures Rocket Lab commands multiples of revenue that are roughly two orders of magnitude above what Avio commands on sales, and the entire gap is a claim about the future of one unflown rocket and one scaling spacecraft business.

The variant view, where the dossier evidence supports one, sits on the composition of the business rather than its growth rate. Consensus discussion frames Rocket Lab as a rocket company facing an existential threat from Starship. The disclosed revenue mix says something different: it is predominantly a spacecraft and subsystems manufacturer with a launch business attached, and the spacecraft contracts — defence constellation work, commercial bus programmes, merchant components sold to other primes — are multi-year, sticky, and largely indifferent to the price of a kilogram to orbit.³⁴ Cheaper launch is an input cost improvement for that business, not a competitive threat. What would have to become true for the thesis to merit deeper work is narrow and observable: Neutron must fly, Archimedes must reach serial production rates, and Space Systems margins must hold as volume scales. What would kill it is equally specific: an Archimedes redesign pushing Neutron beyond 2027, or Space Systems gross margin compressing below twenty per cent on fixed-price spacecraft overruns, which would indicate the high-margin half of the business is not actually high-margin at scale.

The propulsion chokepoint: Avio. The exposure is confirmed in audited European filings, with Ariane and Vega programme contributions disclosed and a backlog of €2.16 billion covering more than three and a half years of revenue.⁹ The business is structurally protected in a way that no commercial launch company is: it is the sole European source of large solid rocket motors, and European institutional payloads fly on European vehicles by policy.⁸ Growth of 22.7 per cent in FY2025 to €541.7 million, with FY2026 guidance of €560 to €590 million, comes with thin margins — 4.8 per cent at the operating line — because the company is investing in United States solid propellant capacity while absorbing input cost inflation.

The reason a sharp investor rejects Avio first is that it is a small-capitalisation, low-margin European defence supplier with a single dominant customer set, a history of Vega C qualification failures, and an earnings multiple flattered or distorted by a capital expenditure cycle. The variant framing available from the evidence is that the solid rocket motor business has two demand curves converging on one factory: European launch cadence, and NATO munitions restocking, where solid propellant capacity is a documented Western bottleneck independent of anything happening in orbit. That second curve is not a launch story at all, which is precisely why it may not be priced as one.

The diluted expressions. Lockheed Martin and Boeing each own half of ULA, which they account for as equity earnings rather than consolidated revenue; Boeing is additionally prime contractor for NASA's Space Launch System core stage.¹³¹⁴ In both cases launch is a small single-digit percentage of consolidated economics, and in Boeing's case it is invisible beneath commercial aircraft recovery and fixed-price defence charges. Airbus and Safran each own half of ArianeGroup, and for both the joint venture is a sovereign capability commitment rather than an earnings driver; Safran's cash flows are dominated by commercial jet engine aftermarket work.¹⁵¹⁶ Mitsubishi Heavy Industries has the most direct sovereign launch role of the conglomerates, running H3, but its earnings are set by gas turbines, shipbuilding and Japanese defence rearmament.¹⁷ Northrop Grumman is the most interesting of the diversified names for this theme, because its exposure is propulsion rather than integration: solid motors into Vulcan and SLS, strategic missile motors, and the medium launch vehicle it is co-developing with Firefly.¹⁸ For all of these, buying the stock to express a launch view means accepting that ninety per cent or more of the outcome will be determined by something else.

The enablers. ATI, Carpenter Technology and Hexcel sell into essentially every Western launch programme, and Air Liquide and Linde supply the cryogens. Their attraction is indifference to which rocket wins; their limitation is that launch is a modest slice of aerospace demand for all of them, and available disclosure does not break out launch-specific revenue, so measurable exposure cannot be established from public filings. Honeywell, BAE Systems and Moog occupy the same position in avionics and actuation. These are real relationships, not ecosystem associations, but the financial link is not quantified in company reporting and should not be presented as though it were.

The false positives. Virgin Galactic ($SPCE) flies suborbital tourism, never reaches orbital velocity, and shares no customer, no economics and no demand driver with this industry. Planet Labs ($PL), Spire Global ($SPIR), BlackSky ($BKSY) and AST SpaceMobile ($ASTS) are customers of launch, not providers of it; falling launch prices reduce their input costs and simultaneously lower the barrier for their competitors, which is a more ambiguous benefit than it first appears. Varda Space Industries manufactures in microgravity and buys rides. Legacy geostationary satellite operators sit on the other side of the same trend: SES and Eutelsat face the reality that proliferated low-orbit capacity is deflationary for the enterprise connectivity revenue their long-lived fleets were built to earn. And the cautionary cases remain instructive: Astra Space and Virgin Orbit were both, at their peaks, presented to public investors as ways to own this theme.

Constructed as a portfolio, exposure to this theme therefore separates into three distinct pathways rather than one: a small number of high-thematic-beta pure plays whose value depends on execution milestones; sovereign and defence-protected cash generators whose revenue is non-discretionary but whose launch exposure is diluted; and upstream material and cryogenic suppliers whose participation is broad, unquantified, and indifferent to which vehicle wins.

The common factor risks across all three are worth naming because they are not independent bets. Long-duration, cash-consuming development companies carry heavy interest-rate sensitivity through the discount rate, so a hawkish rate shock hits the pure-play layer and leaves the defence layer untouched. Single-point failure risk is severe and specific: one upper-stage anomaly grounds a fleet for six to twelve months pending mishap investigation, which for a single-vehicle operator eliminates near-term revenue entirely. Regulatory and environmental review can freeze cadence with hardware sitting ready and customers waiting.⁷ And customer concentration is acute in a way that is easy to miss: a substantial share of the entire non-SpaceX Western commercial backlog traces to two buyers, Amazon's Kuiper programme and the US Space Force. An investor holding ULA exposure through Lockheed Martin, Ariane exposure through Airbus, and propulsion exposure through Avio has not diversified; they have bought the same customer three times.

↑ Back to Top

9. Scenarios to 2030: Three Worlds for the Launch Highway

Forecasting this industry by extrapolating the launch curve is easy and mostly useless, because the curve's slope is set by a handful of discrete, binary events: whether a specific engine reaches production rate, whether a specific constellation keeps writing cheques, whether a specific regulator issues a licence. What follows are three worlds distinguished by causal path rather than by percentage adjustment.

Bear: launch bottleneck and constellation rationalisation. Assign this roughly a one-in-five weight. The mechanism starts outside the industry: a broad macroeconomic downturn forces commercial satellite operators to defer capital spending, and the marginal megaconstellation — the one whose broadband business case was always the thinnest — slows its build. Amazon trims Kuiper's satellite volume. Regulators, responding to congestion in the most crowded orbital shells, tighten collision-avoidance and disposal requirements in ways that raise per-satellite cost and slow licensing. Simultaneously, the supply side disappoints: Neutron slips, New Glenn's cadence stays low, and Vulcan remains engine-constrained. Global cadence stalls in the 280 to 310 launches a year range, well below the 2026 run-rate. Because non-SpaceX capacity never arrives in volume, commercial prices stay stubbornly high — $3,500 to $5,000 per kilogram — which sounds good for incumbents and is not, because the pricing reflects scarcity of supply rather than strength of demand. Operating margins across the challenger cohort compress below five per cent, capital-starved developers write down tooling, and SpaceX's commercial share moves above ninety per cent. Note the shape of this outcome: the upstream belief about orbital industrialisation is only partly falsified, but the investable theme is destroyed, because volume never reaches the level at which anyone but the hegemon absorbs their fixed costs.

Base: bifurcated reusable infrastructure. Assign this roughly three-in-five. Nothing spectacular happens; several ordinary things happen on schedule. The US Space Force executes NSSL Phase 3 as designed, sustaining a dual-source Lane 2 and a competitive commercial Lane 1.⁶ Amazon continues Kuiper deployment and moves into replenishment. China scales Guowang and Qianfan, with commercial Chinese vehicles taking a growing share of domestic manifests. Ariane 6 reaches steady cadence on institutional and Kuiper missions; Vulcan clears its backlog as BE-4 output stabilises; Neutron flies and enters commercial service. Global cadence reaches 450 to 500 launches a year, and upmass reaches 5,000 to 6,500 tonnes. Prices on medium reusable vehicles settle at $1,500 to $2,500 per kilogram, and top-tier reusable operators run gross margins in the twenty-five to thirty-five per cent range once fixed costs are absorbed, while the sovereign programmes continue earning low-double-digit margins on protected institutional work. This is the world in which a genuine multi-provider Western oligopoly exists alongside a dominant leader, and it is worth emphasising that it requires no exotic new space economy: ordinary telecommunications replenishment plus sovereign defence procurement is sufficient to sustain it.

Bull: mass abundance and an industrial space economy. Assign this roughly one-in-five. The mechanism runs through one vehicle. Starship achieves routine tower catch and rapid re-flight of both stages, and demonstrates cryogenic propellant transfer between vehicles in orbit — the capability that turns a rocket into a logistics network, because it allows a fully fuelled stage to depart from low orbit rather than requiring an enormous single-shot vehicle for every deep-space mission. Cost per kilogram falls below $300 and then below $200. At that point the constraint on orbital activity stops being transport and becomes imagination: commercial space stations replacing the ISS, orbital compute clusters, space-based solar power demonstrators, large-aperture instruments that were previously unaffordable. Chinese methalox fleets scale in parallel, and a second, separate high-cadence launch order emerges serving Chinese and non-aligned customers. Global cadence reaches 650 to 800 launches a year and upmass exceeds 12,000 tonnes; leading networks run gross margins above forty-five per cent on volume and automation.

The trap in the bull case is that it is simultaneously the best world for the industry and a dangerous one for most of its equity. Radical cost deflation expands the market, but it expands it through the operator that caused the deflation. The companies that benefit unambiguously in that world are the ones selling what goes on top of the rocket and into it: spacecraft, subsystems, propulsion components, and the industrial capacity to build them at rate. That asymmetry — adoption exceeding expectations while most manufacturers fail to earn their cost of capital — is the standard signature of a deployment-phase capital cycle, and it is the pattern this industry has already run once, at small scale, in 2021 to 2024.

Which of these three worlds is unfolding will be visible long before it shows up in anybody's revenue.

↑ Back to Top

10. The Crux KPIs and Falsification Dashboard

The distillation of everything above is a short list of observables. They are chosen on one criterion: each sits on a binding physical or commercial constraint, so it moves before revenue, share or margin move. Lagging measures — total market size, aggregate sector revenue growth, historical share — are excluded on purpose, because by the time they change the decision has already been made for you.

One: non-SpaceX Western commercial launch throughput. Measured in successful commercial launches per year by Vulcan, Ariane 6, New Glenn, Electron and Neutron combined. The latest reading, annualised through August 2026, is roughly eight to twelve. This leads because commercial constellation deployment schedules are written six to twelve months in advance against expected slot availability, so throughput today determines what operators can contract for next year. It discriminates the central disagreement in the sector: the bull case requires a functioning multi-provider Western oligopoly, while the bear case is a world where every rival becomes a state-subsidised utility serving only its own government. Confirmation would be more than twenty-five non-SpaceX Western commercial launches during 2027. The kill threshold is fewer than twelve combined in 2027, which would mean the capacity shortage is structural rather than transitional. Source: FAA commercial space launch records and the public orbital catalogue, updated monthly.⁷²

Two: reusable booster turnaround time, pad to pad. Measured in days between a booster's landing and its next launch. SpaceX currently averages roughly eighteen to twenty-two days; Rocket Lab and LandSpace are still qualifying recovery procedures rather than reflying at rate.¹²¹⁹ This is the cleanest available proxy for the true marginal cost of reuse, because refurbishment cost is overwhelmingly touch labour, and touch labour shows up as elapsed time. It leads reported margins by a year or more, and it discriminates the argument about whether reuse is genuinely cheaper or merely differently expensive. Confirmation: Falcon 9 sustaining under fourteen days, and first-generation Neutron or Zhuque-3 boosters achieving under thirty. Kill: turnaround stretching beyond ninety days on any reusable programme, which would indicate structural or thermal fatigue requiring teardown rather than inspection. Source: company flight manifests and specialist launch logs, tracked continuously and best assessed quarterly.⁵

Three: medium-lift liquid engine production run-rate. Measured in flight-qualified engines delivered per month. Current readings: BE-4 at roughly one and a half to two per month, Archimedes in qualification batches, and Raptor at approximately one per day.¹¹⁴¹² This is the hardest physical constraint in the entire industry. A rocket programme cannot fly faster than its engine factory, and engine factories scale slowly because every unit must be hot-fire accepted. It leads launch cadence by roughly the vehicle assembly cycle, six to twelve months, and it discriminates whether the challengers' cadence plans are credible. Confirmation: Rocket Lab reaching more than two flight Archimedes engines a month during 2027, and BE-4 output rising enough to unblock Vulcan. Kill: engine delivery shortfalls forcing manifest postponements of more than six months. Sources: company disclosures in SEC filings and defence programme reporting, quarterly.³⁶

Four: Space Systems revenue share and gross margin at Rocket Lab. Measured as the non-launch share of quarterly revenue and its GAAP gross margin. The latest reading is 70.1 per cent of revenue, roughly $164 million, at a company gross margin of 31.5 per cent, against $2.36 billion of backlog.³ This is the single most informative disclosed number in the listed universe, because it tests the central structural claim of this article — that value is migrating out of launch services and into the spacecraft layer. If the migration is real, the margin holds or expands as volume scales. It discriminates the bull case, in which vertically integrated platforms capture end-to-end constellation budgets, from the bear case, in which spacecraft assembly turns out to be commoditised contract manufacturing with fixed-price risk. Confirmation: Space Systems revenue growing more than forty per cent year over year with gross margin above thirty-two per cent. Kill: gross margin compressing below twenty per cent. Source: quarterly SEC filings.³

Five: the commercial smallsat rideshare price floor. Measured in dollars per kilogram to a standard sun-synchronous orbit on aggregated rideshare missions, currently $5,500 to $6,500.¹² This sets the global clearing price for access, and therefore the survival threshold for every dedicated small launcher on Earth. It leads because pricing decisions precede the demand response by a booking cycle. It discriminates whether dedicated launch retains a defensible premium for bespoke orbits and responsive timelines — the bull case is a premium around three times the rideshare rate — or whether aggressive rideshare pricing eliminates the category. Confirmation: dedicated small-launch pricing holding above roughly $7.5 million per flight. Kill: dedicated list prices falling below $5 million without a compensating increase in flight volume. Source: published rideshare pricing and commercial launch quotations, revised roughly twice a year, with the caveat that dedicated launch pricing is negotiated and only partially observable, so this indicator is a bounded proxy rather than a precise measurement.

Sitting above these company and industry indicators are two theme-level kill triggers, which are different from security-level ones and should not be confused. The thematic hypothesis fails if global launch attempts fall below 250 a year for two consecutive quarters, or if a major constellation operator — Kuiper or Guowang being the two that matter most — pauses deployment for more than nine months.¹ Those would falsify the demand claim itself. A security-level failure, by contrast, can occur in a thriving industry: an engine redesign, a fixed-price contract overrun, a grounded fleet, or simply a valuation that already assumed the good outcome.

↑ Back to Top

11. Game Changers, Tail Risks, and the Final Verdict

Three developments could move the value pools described here, and each is at a different stage of maturity. It is worth separating what has been demonstrated from what has been announced.

Full-stack reusability. If Starship achieves routine recovery and rapid re-flight of both stages, rather than the booster alone, launch cost falls below $200 per kilogram and the expendable upper stage — currently the single largest variable cost on every reusable mission — disappears from the industry's cost structure.¹² The technical hurdle is upper-stage thermal protection surviving orbital re-entry repeatedly without refurbishment, which is a materials problem nobody has yet solved at scale; Stoke Space is attempting a different answer with an actively cooled metallic heat shield.⁵ The beneficiaries would be spacecraft builders, whose mass and volume constraints relax dramatically, and the operators of anything that becomes affordable only at that price: commercial stations, large-aperture instruments, orbital compute. The losers would be every expendable vehicle still in production, and the medium-lift entrants whose business cases assume prices five to ten times higher. The observable milestone is not a test flight; it is a second flight of the same upper stage.

Orbital propellant transfer. Moving cryogenic methane and oxygen between vehicles in microgravity is unglamorous and transformative, because it decouples departure energy from launch vehicle size. A stage refuelled in low orbit can deliver far more mass to the Moon or beyond than the same stage flown direct. The hurdle is fluid management without gravity to settle the propellant, plus boil-off control over the days a depot must hold cryogens. This is the enabling technology for lunar industrial supply chains and the reason methalox architecture matters strategically as well as thermodynamically.

Chinese commercial fleet scaling. If Zhuque-3 and Tianlong-3 reach weekly cadence, China gains the capacity to deploy Guowang and Qianfan on schedule and to offer low-cost launch to countries outside the Western export-control perimeter.¹⁹²⁰ The result would be two parallel launch orders with almost no trade between them, and a competitive dynamic in which Western providers never face Chinese price pressure directly but lose access to a growing share of the world's satellite operators.

Against these sit three tail risks with genuinely different mechanisms. Orbital debris cascade — the Kessler scenario — is the one that could close the business outright: a high-velocity collision in the congested 500 to 600 kilometre shells generating debris that triggers further collisions, rendering specific inclinations unusable and prompting licensing moratoria. Its probability is unquantifiable with current data, but its consequence is total. Atmospheric regulation is slower and more likely: research groups continue to disagree about the stratospheric effects of kerosene soot, solid-motor alumina and water vapour injection at altitude, no quantitative regulatory threshold yet exists, and the direction of any eventual rule is clear enough — it would penalise kerosene and solids and advantage methane and hydrogen, which happens to be the direction the industry is already moving for engineering reasons. Space-domain conflict is the risk that reallocates budgets rather than destroying them: anti-satellite weapons or persistent electronic warfare would shift government spending from commercial launch procurement toward hardened, defended architectures, which is bad for commercial cadence and ambiguous for the defence primes.

So return to the proposition this article set out to test: that continuous LEO deployment and replenishment constitutes a durable, non-discretionary capital cycle that decouples launch demand from prestige budgets and establishes orbital transportation as a distinct, cash-generative industrial layer.

Two-thirds of that belief is holding and, on the evidence, strengthening. Demand has decoupled from civil space budgets — commercial operators now procure the large majority of launches, and the replenishment obligation embedded in short-lived satellites is a maintenance liability rather than a discretionary purchase.¹ Volume has more than doubled in four years while reliability improved, which is what industrialisation looks like from the outside.

The third clause is where the belief frays. Orbital transportation has become a distinct industrial layer, but it has not yet become a broadly cash-generative one. The economics concentrate viciously: one private operator captures roughly eighty per cent of delivered mass and has crossed the fixed-cost absorption threshold using demand it manufactures for itself; a handful of sovereign programmes earn protected but modest returns on institutional payloads; and the open commercial middle remains a place where companies burn capital waiting for a flight rate that arrives slowly or not at all. The entire addressable launch market, worldwide, is around $15 billion a year — smaller than the market's valuation of its single listed pure play.

For an equity investor, that resolves into a specific and slightly uncomfortable conclusion. Being right about the industrialisation of low Earth orbit is necessary and nowhere near sufficient. The surplus created by cheap launch flows overwhelmingly downstream, to subscribers, defence ministries and satellite operators. What remains inside the industry accrues to two positions: the operator whose scale nobody can contest, and the suppliers of the things every rocket must have and cannot easily substitute — engines, solid motors, superalloys, cryogens, and increasingly the spacecraft that ride on top. The rocket is the most visible object in this story and the least reliable place to own the economics. The tollbooth is real; the question worth asking of any expression of this theme is whether you are buying the toll, or merely the road.

↑ Back to Top

Glossary

AFTS (autonomous flight safety or termination system). An onboard system that tracks a vehicle's trajectory and destroys it automatically if it leaves the approved corridor, replacing ground radar operators and human range safety officers. It matters because it dramatically shortens range turnaround between launches, which is a direct input to achievable cadence.

Contribution margin versus gross margin. Contribution margin is price less variable cost per unit; gross margin additionally absorbs fixed manufacturing and facility cost. In launch the gap between the two is enormous, and conflating them is the most common analytical error in the sector.

Dry mass and wet mass. Wet mass is the fuelled vehicle on the pad; dry mass is structure, engines and avionics without propellant. Because propellant is ninety to ninety-five per cent of wet mass, small improvements in dry mass translate into large changes in payload capacity.

Full-flow staged combustion (FFSC). The most thermodynamically efficient liquid engine cycle, routing all propellant through separate oxygen-rich and fuel-rich preburners before the main chamber. It permits very high chamber pressure while keeping turbine temperatures lower, which is what makes an engine practical to re-fly.

GEO and GTO. Geostationary orbit is the ring 35,786 kilometres above the equator where a satellite's period matches Earth's rotation; geostationary transfer orbit is the elliptical path used to get there. Reaching them demands high-energy upper-stage performance, which is a distinct capability from bulk delivery to low orbit.

LEO (low Earth orbit). Altitudes between roughly 160 and 2,000 kilometres. Satellites here experience residual atmospheric drag and de-orbit within years unless replaced, which is the physical basis of the replenishment demand cycle.

Methalox. Liquid methane burned with liquid oxygen. It leaves no soot in engine passages, re-ignites easily, and performs better than kerosene, which is why it has become the default propellant for reusable vehicles worldwide.

NSSL (National Security Space Launch). The US Space Force programme that procures launch for national security payloads. Its structure — multiple certified providers, split lanes, fixed-price awards — is the single most important determinant of the American non-commercial manifest.

Part 450. The FAA's performance-based licensing framework for commercial launch and re-entry in the United States. Licensing timelines and associated environmental review are a real, recurring constraint on flight rate.

Rideshare. Selling capacity on a large rocket by the kilogram to many small customers at once. It is the substitute that destroyed the dedicated small-launch business case, at roughly a fifth of dedicated pricing.

Upmass. Total mass of spacecraft and cargo successfully delivered to orbit over a period. It is a better measure of industrial activity than launch count, because vehicles differ by more than an order of magnitude in capacity.

VTVL (vertical takeoff, vertical landing). Propulsive recovery of a booster, decelerating through the atmosphere and touching down under engine thrust. It is the technique that converted rocket first stages from consumables into capital assets.