The Physics of Abundance: How Launch Deflation Reshaped the $630 Billion Orbital Economy
1. Cold Open & The Upstream Belief
At Space Launch Complex 40 on Florida's Atlantic coast, the interval between Falcon 9 launches in mid-2026 is measured in hours rather than months. A booster lands on a droneship, returns to port, is inspected, is stacked again. Three hundred miles west and one time zone away, on a flat scrap of South Texas coastline, a stainless-steel vehicle taller than the Statue of Liberty is stacked, fuelled, flown, and β increasingly often β caught. Between them, these two sites accounted for the majority of everything humanity placed into orbit last year.12
To understand how strange this is, compare it to the machine that defined the previous era. The Space Shuttle was reusable in the way a racing car is reusable: it came back, and then a small army took it apart. Depending on whether you amortise the programme's enormous fixed costs across its flights or count only the marginal expense of the next mission, the Shuttle delivered payload to low Earth orbit at somewhere between $18,000 and $54,000 per kilogram.1 That accounting ambiguity is itself the point. When a launch system costs so much that reasonable people can disagree by a factor of three about its unit economics, launch is not a logistics service. It is a national programme with a flight manifest attached.
Falcon 9 broke that. By the mid-2020s, commercially disclosed pricing put first-generation reusable heavy launch at roughly $1,400 to $2,500 per kilogram to LEO.1 Starship, Blue Origin's New Glenn, and Rocket Lab's Neutron are all aimed at the tier below that β $150 to $500 per kilogram, with SpaceX management targeting the bottom of that band and below.1 Each of those steps is roughly an order of magnitude. Two of them have already happened.
Why we looked here
The proposition that made this industry worth a dossier is narrow enough to be wrong: launch transport cost is the master variable of the space economy, and its collapse from over $10,000 per kilogram to under $1,500 β with a credible path below $200 β converts space from a capital-rationed strategic domain into an industrial frontier that can support commercial megaconstellations, sovereign orbital defence networks, in-space manufacturing, and lunar logistics. Everything that follows is a test of that sentence.
Three independent lines of evidence pushed us to take it seriously, and none of them is a market forecast.
The first is a physical census. Active operational payloads in orbit grew from roughly 1,400 in 2016 to more than 11,500 by mid-2026, and the annual mass placed in orbit expanded by more than 450% between 2020 and 2025.2 These are catalogued objects tracked by the US Space Force, not vendor projections. Whatever else is true, an enormous amount of hardware is now going up.
The second is a change in what customers buy, which is usually a better signal than what they say. The US Space Development Agency, the Space Force, and allied defence ministries have restructured procurement away from bespoke geostationary satellites costing hundreds of millions each and toward recurring tranches of standardised proliferated LEO spacecraft, awarded under fixed-price contracts.4 A defence establishment that has spent sixty years buying exquisite, irreplaceable satellites does not casually decide to buy hundreds of cheap ones. It does so when the transport cost stops being the binding constraint.
The third is industrial back-pressure in the supply chain. Order books in adjacent advanced manufacturing β high-efficiency III-V solar cells, laser communication terminals, radiation-tolerant microelectronics, carbon-fibre prepreg β are being filled by satellite constellation production runs and rocket stage fabrication rather than by one-off science missions.3 When a component maker's backlog changes shape, something real is happening downstream.
The transmission mechanism from cheap launch into industry economics is the interesting part, and it runs through design, not demand. When it costs $20,000 per kilogram to reach orbit, the rational engineering choice is to spend almost unlimited money on reliability: radiation-hardened parts, redundant everything, fifteen-year design life, exhaustive qualification. Transport dominates the budget, so you optimise everything else against it. When transport falls to $1,500 and heads toward $200, the calculus inverts. It becomes cheaper to fly automotive- and industrial-grade components with three-to-five-year lifespans and replace them continuously.1 That single substitution does three things at once: it collapses the unit cost of a satellite, it multiplies the number of satellites required, and β crucially for anyone holding these equities β it converts a one-time capital purchase into a perpetual replacement cycle. The same belief implicates several sibling industries we deliberately did not chase here: radiation-tolerant semiconductors, electro-optical and infrared defence sensors, and terrestrial mobile network operators whose spectrum and billing relationships now sit at the boundary of an orbital network.
What this article is trying to decide
There is a distinction that governs the whole piece, and it is the one that separates a correct industry forecast from a profitable investment. You can be entirely right that mass-to-orbit grows tenfold and still lose money, if you own the wrong layer of the stack, the wrong company inside that layer, the wrong capital structure, or the right company at the wrong price.
The dossier's decision context is deliberately plain: general institutional public-equity research, global listed expressions, a multi-year horizon, no position sizing and no security recommendations. That matters because a benchmark-relative long-only investor and an absolute-return long/short investor do not face the same problem here. The first is asking whether to own a newly listed mega-cap that has just become impossible to ignore. The second is asking which layer of a deflating industry has durable pricing power and which is being quietly hollowed out.
Both need the same causal map, and it has a specific shape. Reusability creates a natural cost monopoly at the launch layer, because the fixed cost of developing a reusable vehicle is enormous and the marginal cost of flying it again is small β so whoever gets there first can price below every rival's cash cost and still earn a margin. But that same deflation destroys the value of launch as a product. Transport becomes a utility. And utilities, historically, are not where the economic profit of an industry ends up.
So the question that organises this article is: where does the money go when getting to orbit stops being hard? The candidates are the vertically integrated operator that owns both the rocket and the network it carries, the subsystem suppliers who own bottlenecks no launch cost reduction can relieve, and the downstream data and bandwidth businesses that finally become viable at the new price. The evidence points at all three, in different proportions, and it points sharply away from one: the pure-play expendable launch provider, which the 2020s have already begun to bury.
The collapse of launch cost is not a cyclical price cut that competition will restore. It is a permanent re-platforming that invalidates the design logic of forty years of satellite engineering β and with it, the business models built on top.
Subscribe on LinkedIn2. Technical Foundations & The Launch Deflation Engine
To see why reusability is so financially violent, you have to look at a piece of arithmetic that has governed rocketry since 1903.
The Tsiolkovsky rocket equation says that the velocity change a vehicle can achieve equals its exhaust velocity multiplied by the natural logarithm of its mass ratio β the vehicle's starting mass divided by its ending mass. Written out: Ξv = vβ Β· ln(mβ/m_f). The logarithm is the cruelty. To reach orbital velocity, roughly 7.8 kilometres per second, a chemical rocket must be overwhelmingly propellant. In practice, a two-stage launch vehicle at liftoff is about 90% fuel and oxidiser, a few percent structure and engines, and something like 2β4% payload.
Now add the traditional business model on top: every kilogram of that structure and every engine is destroyed on each flight. A commercial airliner that was scrapped after one trip would need to charge each passenger the price of the aircraft. That is precisely what expendable launch did, and it is why the pre-reusability cost curve sat where it did β $8,000 to $14,000 per kilogram for the mature expendable commercial generation of Delta IV, Atlas V, and Ariane 5.15
Reusability attacks the numerator of that fraction rather than the physics. You spend performance β extra propellant reserved for the landing burn, the mass of landing legs and grid fins, a thermal protection system β to buy back the hardware. The trade only works if two conditions hold. The recovered stage must actually be re-flyable without a rebuild, and it must fly again often, because the whole point is amortising a fixed asset across many uses.
SpaceX proved both. Falcon 9's first stage returns under its own power, guided by grid fins that steer the vehicle through the upper atmosphere, and individual boosters have gone on to accumulate more than twenty flights.1 Once a booster is on its twentieth mission, the depreciation charge per launch has fallen to a rounding error, and the dominant costs become propellant, range services, refurbishment labour, and the expendable second stage. This is why SpaceX can list Falcon 9 at roughly $65 million and still earn launch gross margins in the mid-thirties.1 A competitor flying an expendable vehicle must recover 100% of its bill of materials on every single flight. There is no pricing strategy that closes that gap.
The engine question: why methane
The next order-of-magnitude step required changing the propellant. First-generation reusable rockets β Falcon 9's Merlin engines β burn RP-1, a refined kerosene. Kerosene is dense and energetic, but it cokes: burning it deposits carbon residue inside the engine's passages and turbopumps, exactly the way a wood stove clogs its own flue. Cleaning that out between flights is skilled, slow, expensive work. Liquid hydrogen, the other classical high-performance choice used by Ariane and the Shuttle, burns clean but is so low-density that it demands enormous insulated tanks, and it embrittles metal.
Methane sits between them. Liquid methane and liquid oxygen β "methalox" β burns essentially soot-free, is dense enough for reasonable tank volumes, and can be run in a full-flow staged combustion cycle where both propellant streams pass through preburners before entering the main chamber, extracting more energy without cooking any single turbine. The analogy to a clean-burning engine versus a coking one is useful, but it has a limit worth naming: methane's advantage is dominated by turnaround labour, not by raw performance. On specific impulse alone, hydrogen still wins. Methalox is chosen because it makes the vehicle cheap to fly twice, not because it makes the vehicle fast.
That is why the entire next generation converged on it. SpaceX's Raptor powers Starship. Blue Origin's BE-4 powers New Glenn β and, in a relationship that reveals a great deal about how power now flows in this industry, it also powers United Launch Alliance's Vulcan Centaur, the Boeing ($BA) and Lockheed Martin ($LMT) joint venture's flagship vehicle. Blue Origin is the sole source of the BE-4 for Vulcan's main stage.1 A privately held competitor supplies the beating heart of the two largest US defence primes' launch vehicle, and there is no qualified alternative engine on the shelf. Rocket Lab's Neutron uses the Archimedes engine on the same methalox architecture. In China, LandSpace's Zhuque-2 became the first vehicle anywhere to reach orbit on methalox, in 2023, and the company has been running vertical-takeoff-vertical-landing recovery tests for its larger Zhuque-3.17
The other half: manufacturing, not just recovery
Reusability gets the headlines, but a second and quieter change did comparable work. Rockets historically were fabricated the way cathedrals were: skilled hands, low volumes, long lead times. The new cohort industrialised the process.
Rocket Lab ($RKLB) built Electron's structure from carbon composite rather than machined aluminium, and replaced the conventional gas-generator turbopump with electric pumps driven by lithium-polymer batteries β a design that trades a small performance penalty for radically simpler, faster-to-build engines. Fifty-plus successful Electron flights through mid-2026 make it the most reliable dedicated small-satellite launcher in service, a leadership claim with a tight parameter: dedicated small-payload launch, measured by successful flights and mission success rate, as of mid-2026.3 Electron does not lead on cost per kilogram, and we will come back to why that distinction eventually cost the whole small-launch category its life.
Additive manufacturing pushed the same logic further. Agnikul Cosmos in Chennai flight-tested Agnite, a semi-cryogenic engine printed as a single piece β no assembly of hundreds of brazed components, no welds to inspect β and flew it on the Agnibaan SOrTeD demonstrator from its own private launchpad at Sriharikota.20 The engineering claim is genuine and the manufacturing lead time reduction is real. The commercial claim is more constrained, and we will test it in Section 6: Agnibaan's design payload is roughly 300 kilograms to LEO, which places it in exactly the category the cost curve has been destroying.
Here is what all of that produced.
Exhibit 1 β Global orbital launch attempts and mass to orbit, 2018β2025 Definition: attempted orbital launches by all operators worldwide; SpaceX reusable column counts Falcon-family flights using a recovered or recoverable first stage; mass to orbit is estimated total payload mass delivered, metric tons. Geography: global. Evidence status: observed/reported, compiled from launch logs and company disclosure.
| Year | Total global launches | SpaceX reusable launches | China total | Rest of world | Mass to orbit (t) |
|---|---|---|---|---|---|
| 2018 | 114 | 21 | 39 | 54 | ~450 |
| 2020 | 114 | 26 | 39 | 49 | ~550 |
| 2022 | 186 | 61 | 64 | 61 | ~1,020 |
| 2024 | 258 | 134 | 68 | 56 | ~2,100 |
| 2025 | 312 | 168 | 78 | 66 | ~2,950 |
Sources: SpaceX SEC disclosure; Space-Track.org catalogue.12
Read that aloud and one fact dominates. Global launch attempts nearly tripled between 2018 and 2025, from 114 to 312. But look at where the growth came from. The "rest of world" column β Europe, Russia, India, Japan, every legacy operator combined β went from 54 to 66 over seven years. That is essentially flat. China roughly doubled, from 39 to 78. SpaceX's reusable flights went from 21 to 168, an eightfold increase. And mass to orbit grew faster than flight count, from roughly 450 tons to 2,950 tons, because the vehicles doing the growing carry far more per flight. By 2025 SpaceX was placing more than 80% of all payload mass into orbit worldwide.1 The launch industry did not grow. One company grew, and the industry's arithmetic followed it.
Now the price series that caused it.
Exhibit 2 β Estimated launch cost per kilogram to low Earth orbit, by vehicle generation Definition: representative commercially disclosed or programme-accounted cost per kilogram of payload delivered to LEO, nominal USD. Evidence status: the first three rows are observed contract or programme accounting ranges; the heavy-reusable row is a forward target, not an achieved price.
| Era | Representative vehicles | Cost/kg to LEO | Economic driver |
|---|---|---|---|
| Shuttle era (1980sβ2000s) | Space Shuttle, Titan IV | $18,000β$54,000 | Cost-plus contracting, no reuse |
| Expendable commercial (2000sβ2010s) | Delta IV, Atlas V, Ariane 5 | $8,000β$14,000 | Commercial expendable competition |
| First-gen reusable (2016β2024) | Falcon 9, Falcon Heavy | $1,400β$2,500 | First-stage and fairing recovery |
| Dedicated small launch (2020β2026) | Electron, Alpha, Agnibaan, Vikram-1 | $12,000β$22,000 | Premium for dedicated orbit/schedule |
| Heavy reusable (2025β2030, target) | Starship, New Glenn, Neutron | $150β$500 (target) | Full reuse, rapid turnaround, methalox |
Sources: SpaceX filings and commercial contract disclosure; ESA Space Transportation Directorate reporting.15
The row that should stop a listener is the fourth one, not the fifth. Dedicated small launch β the entire venture-funded category of the early 2020s β sits at $12,000 to $22,000 per kilogram, which is to say above the expendable commercial generation it was supposed to disrupt and roughly ten times Falcon 9. Small-launch customers were paying a large premium for a specific orbit on a specific date. That premium is real and some customers still pay it. But it is a niche, and the capital markets funded it as though it were a market.
The forward row deserves its own caution. $150β$500 per kilogram is a target, and SpaceX's own management claim β long-run marginal cost below $10 million per Starship flight, implying figures near $100 per kilogram β is disputed by independent aerospace engineers who point to heat-shield tile replacement labour, methane boil-off losses, and launch-mount refurbishment as costs that management's figures appear to understate.1 Nothing in this article treats sub-$200 launch as achieved. It is the single largest open question in the sector, and Section 9 makes it the first thing to monitor.
Reusability converted launch from a bespoke engineering project into a logistics utility. The immediate consequence was that everyone with a satellite to fly suddenly had a much larger budget for the satellite.
3. The Megaconstellation Race & Ground Realities
Somewhere in the middle of the Pacific, a shipping container on a bulk carrier is streaming a gigabit. In eastern Europe, a mobile command post is doing the same. Neither is pointing a dish at a fixed spot in the sky. Both are using flat phased-array panels that steer their beams electronically, handing off between spacecraft crossing overhead at roughly 27,000 kilometres per hour.
That handoff is the entire architecture, and it comes from a choice about altitude with unavoidable consequences.
A geostationary satellite orbits at 35,786 kilometres, where its orbital period exactly matches Earth's rotation, so it appears to hang motionless above one spot. That is enormously convenient: you need three of them for near-global coverage, and your ground antenna never moves. It is also a physical trap. Radio waves travel at the speed of light, and the round trip to geostationary orbit and back imposes something on the order of 500β600 milliseconds of latency before any processing. For broadcast television, irrelevant. For a voice call, awful. For interactive data, mobile backhaul, gaming, or a video conference, disqualifying.
Low Earth orbit β 500 to 1,200 kilometres β cuts that to under 30 milliseconds, comparable to terrestrial broadband. The price is that nothing stays put. A LEO satellite crosses the visible sky in minutes. To keep a customer connected continuously, you need enough satellites that another is always rising as the first sets, everywhere on Earth you sell service. That requirement is why the number is not fifty or two hundred but thousands, and it is why the LEO broadband business could not exist until launch got cheap. The constellation is the product, and the constellation is a launch bill.
Starlink and the internal-customer flywheel
SpaceX ($SPCX) understood this earlier and acted on it harder than anyone. By mid-2026 Starlink operated more than 6,500 active satellites and served over 4.2 million active subscriber terminals globally.12
The mechanism that got it there is worth stating precisely, because it is the clearest example in modern industry of vertical integration converting into competitive advantage. SpaceX launches Starlink satellites on its own rockets, at internal cost. A Falcon 9 flight that a customer pays roughly $65 million for costs SpaceX something closer to $15 million to fly for itself.1 So while every rival constellation must buy launch on the open market at market prices β and, in most cases, buy it from SpaceX β SpaceX deploys its own network at roughly a quarter of that price. Starlink also solved Falcon 9's own economics from the other direction: reusability only pays if you fly often, and in the years when commercial demand alone could not fill the manifest, Starlink was the anchor customer that kept the cadence up. Each business made the other viable.
The competition is real but structurally behind. Amazon ($AMZN) has been scaling Project Kuiper's operational deployment by procuring heavy launch β including, at times, from SpaceX. Eutelsat ($ETL) inherited OneWeb's constellation through consolidation. China's Guowang and Qianfan/G60 constellations are being built with state and municipal capital on national timelines. Each of these can eventually reach global coverage. None of them can reach it while paying a competitor's launch prices and expect the same margin structure. That is the durable part of Starlink's lead, and it is why the lead is best described narrowly: leadership in LEO consumer and enterprise broadband, measured by active subscriber terminals and satellites on orbit, as of mid-2026 β with the cost basis of deployment, not the technology, as the reason.
A word on the revenue figures, because two of them do not combine cleanly. The dossier records Starlink annualised subscriber revenue above $6 billion in one place and, elsewhere, notes that more than 70% of SpaceX's $7.81 billion of second-quarter 2026 revenue came from Starlink subscriptions and launch contracts combined.1 Those are different definitions measured at different times, and multiplying either into the other produces a number nobody disclosed. What can be said with confidence is that Starlink is now the majority of SpaceX's revenue and that its growth rate is the reason group revenue nearly doubled year on year. Anything more precise about the subscription line alone is not established by disclosure.
The bottleneck nobody expected: the thing on the roof
Here is the counterintuitive part of LEO broadband economics. The satellites are hard, but they are a solved manufacturing problem for anyone who can build them at volume. The binding constraint on subscriber growth has repeatedly been the user terminal.
A phased-array antenna is a flat panel containing hundreds or thousands of small radiating elements whose relative phase can be adjusted in microseconds, letting the panel form and steer a beam electronically with no moving parts. It is closer to a specialised radar than to a satellite dish. It is also expensive to manufacture, power-hungry, and sensitive to semiconductor supply. Consumer terminals have historically carried a several-hundred-dollar hardware cost that operators frequently subsidise to win the subscriber β which means each new customer consumes working capital before generating recurring revenue. Scaling to ten million subscribers is therefore a manufacturing and financing problem as much as an orbital one.
Which is exactly why the direct-to-cell approach is so financially interesting, and so risky.
AST SpaceMobile ($ASTS) is building BlueBird satellites with very large unfolding phased-array apertures designed to talk directly to unmodified 5G smartphones using mobile operators' existing licensed spectrum, under partnership agreements with AT&T ($T), Verizon ($VZ), and Vodafone ($VOD) covering an addressable base measured in billions of subscribers.13 If it works at scale, the user terminal cost falls to zero, because the terminal is the phone already in the customer's pocket. That is a genuinely different unit economic model from Starlink's.
The risks are correspondingly specific. First, physics: closing a link to a handset with a fraction of a watt of transmit power and a centimetre-scale antenna requires an enormous satellite aperture and generous spectrum, which is why the satellites are large and the constellation is expensive. Second, regulation: transmitting terrestrial mobile spectrum from orbit requires a framework β the FCC's Supplemental Coverage from Space rules established one in the United States, but equivalent authorisations must be won country by country.8 Third, and least discussed, bargaining power. The customer relationship, the billing system, the retail brand, and the spectrum licence all belong to the mobile operator. The satellite company supplies coverage into someone else's network. The revenue-share terms in those agreements have not been disclosed, and the spread between "operators pay 10% of incremental ARPU" and "operators pay 40%" is the difference between two entirely different companies.13 Anyone modelling AST's terminal value is, in effect, guessing at a number neither party has published.
Observation, not communication
Broadband dominates the revenue, but the same launch deflation created a second downstream industry.
Planet Labs ($PL) applied consumer-electronics thinking to Earth observation. Rather than one exquisite half-billion-dollar satellite tasked to photograph specific targets on request, Planet flew hundreds of shoebox-sized "Dove" satellites in a configuration that scans the entire terrestrial landmass every day at roughly three-metre resolution, and has done so for over a decade.11 The asset that produces is not the current image. It is the archive: a continuous, consistent, decade-long time series over every point on Earth, which is what makes automated change detection possible. You cannot retroactively photograph 2017. Planet's leadership claim is therefore narrow and strong β leadership in daily global optical revisit frequency, measured by coverage cadence, as of 2026 β while BlackSky ($BKSY) leads on rapid tasking of specific sites and Pixxel leads on spectral depth rather than temporal frequency.
Spire Global ($SPIR) took a third position: roughly a hundred LEMUR nanosatellites collecting radio-frequency signals rather than images β ship AIS transmissions, aircraft ADS-B, and GPS radio-occultation soundings that are fed into weather models.12 It is a smaller business, around $110 million of annual revenue, and it has carried compliance and restatement noise in its history, but it illustrates the category: once launch is cheap, entirely new sensing modalities become commercially viable because the constellation no longer has to be justified by a single flagship customer.
LEO broadband and LEO observation are global telecom and data networks that happen to run on an orbital substrate, and in both, victory goes to whoever can manufacture and deploy replacement hardware faster than orbital decay and obsolescence remove it. Which raises the obvious next question: who supplies the parts?
4. Value Chain Mechanics & Bottleneck Monopoly
In a cleanroom in Germany, technicians align optics to sub-millimetre tolerances on a device roughly the size of a carry-on suitcase. Its job is to fire an infrared laser at another such device up to two thousand kilometres away, moving at eight kilometres per second, and hold the lock while terabytes of data pass between them. No ground station, no radio spectrum licence, no downlink that an adversary can jam or intercept.
The company is Mynaric, and in 2026 Rocket Lab acquired it.3 Understanding why that transaction matters more than most people assumed requires walking the value chain from the bottom.
The five layers
Foundational inputs. At the base sit materials and semiconductors with no space-specific market of their own: aerospace alloys, carbon-fibre prepreg, and β most importantly β III-V compound semiconductor wafers used for space solar cells, along with radiation-tolerant microelectronics. Margins here run 20β30%, and bargaining power is moderate to high because qualification cycles are long and the supplier lists are short.
Enabling subsystems. This is where the economics get interesting. Optical inter-satellite link terminals, Hall-effect plasma thrusters, reaction wheels, star trackers, high-efficiency solar arrays. Margins run 25β40%, and bargaining power is genuinely high, because these are the parts that are hard to qualify, slow to second-source, and mandated by customers.
Launch vehicle assembly. Heavy reusable at one extreme, dedicated small expendable at the other. Reusable launch margins range from deeply negative during development to above 35% at scale; small expendable launch runs 0β15% and falling. Bargaining power is extreme for SpaceX and weak for everyone selling an expendable vehicle.
Satellite bus integration. Assembling standardised 100β1,000 kilogram spacecraft. Margins 15β20%, bargaining power moderate. This is the layer most exposed to commoditisation, because a bus is increasingly a configuration of purchased subsystems.
Downstream operations and data. Bandwidth sold by the month, imagery sold by subscription, defence intelligence sold by contract. Margins 30β60% at scale, with very high bargaining power for whoever aggregates the network.
Now overlay the deflation. When launch cost falls, the value of "being able to get to orbit" falls with it. But the value of the parts that determine whether your spacecraft works once it is there does not fall at all β if anything it rises, because there are now vastly more spacecraft competing for the same qualified suppliers.
Exhibit 3 β Migration of industry gross profit by value-chain layer Definition: estimated share of total space-industry gross profit by layer, global, comparing 2016 with 2026. Evidence status: dossier estimate synthesised from disclosed company economics; directional rather than precise, and not reconcilable to any single published statistical series.
| Layer | Share of industry gross profit, 2016 | Share, 2026 |
|---|---|---|
| Launch services | ~35% | <18% |
| LEO constellation bandwidth + specialised defence payloads | (minor) | >55% |
| All other (buses, ground, legacy GEO, services) | ~65% combined | ~27% |
Source: dossier profit-pool analysis derived from company disclosure.13
The interpretation is the thesis in one line: launch roughly halved its share of the industry's gross profit over a decade during which launch volume grew nearly threefold. That combination β surging volume, collapsing profit share β is the signature of a layer becoming a utility. Meanwhile the two layers that barely existed as profit pools in 2016, orbital bandwidth and proliferated defence payloads, now take the majority. Treat the exact percentages as directional. The direction is not in doubt. A caution belongs here too: this is a synthesis of company-level economics, not an audited statistical series, and no official body publishes space-industry gross profit by layer.
Who supplies whom, concretely
Take the SDA's Proliferated Warfighter Space Architecture, the US military's layered LEO network for missile tracking, data transport, and battle management. The government mandated that its satellites carry standardised optical inter-satellite links so the constellation can route data among itself as a mesh rather than dropping through vulnerable ground stations.4 That single architectural requirement created a procurement bottleneck.
The prime contractors on PWSA are the large defence integrators β Northrop Grumman ($NOC), L3Harris ($LHX), Lockheed Martin. They win the satellite contracts. But they must buy the optical terminals, and Mynaric's CONDOR terminals became one of the small number of qualified sources feeding those primes across Tranche 1 and Tranche 2.34 The direction of dependence is worth stating plainly: the primes need the terminals more than the terminal maker needs any single prime, because a prime who cannot source terminals cannot deliver satellites, while the terminal maker has three prime customers competing for the same production slots.
That is what Rocket Lab bought. And it fits a pattern that has been running for years.
Rocket Lab's chief executive Peter Beck concluded early that dedicated small launch, on its own, was a structurally low-margin business. The response was a sequence of acquisitions aimed squarely at the subsystem layer: Sinclair Interplanetary for reaction wheels and star trackers, SolAero Technologies for high-efficiency III-V space solar cells, Planetary Systems for separation hardware, Geost for defence electro-optical and infrared sensor payloads, Motiv Space Systems for robotics, and Mynaric for laser terminals β with an acquisition of Iridium ($IRDM) announced and pending as of the publication date.3 The result is that by mid-2026 Space Systems generated roughly two-thirds of Rocket Lab's revenue at gross margins above 30%, materially better than launch.3
SolAero deserves its own sentence because it demonstrates the same bottleneck logic on the power side. Space-grade solar cells are triple-junction III-V devices with efficiencies far above terrestrial silicon, and the qualified supplier base worldwide is essentially two companies: SolAero and Boeing's Spectrolab. SolAero supplies cells and panels into SpaceX, NASA, Lockheed Martin, and Northrop Grumman programmes.3 So Rocket Lab, a launch competitor to SpaceX, is also a component supplier to SpaceX. In an industry with this few qualified sources, competitors are routinely each other's vendors, and the exposure that creates cuts both ways: a customer concentration risk for the supplier, a single-source risk for the buyer.
Two more relationships complete the map of where leverage sits. Blue Origin supplies the BE-4 engine to United Launch Alliance as the sole source for Vulcan Centaur's main stage β a private company holding a chokepoint over two listed defence primes' launch vehicle.1 And Avio ($AVIO) manufactures the P120C solid rocket motor, which serves both as Vega-C's first stage and as the strap-on booster for Ariane 6, making a single Italian supplier a sole source across Europe's entire launch fleet.165 When Avio had to redesign the Vega-C nozzle after a failure, the consequence was not confined to Avio's income statement; it grounded European access to a class of orbits.
Beneath all of them sit the flight computers. Microchip Technology ($MCHP) and BAE Systems supply radiation-tolerant microcontrollers and processing units into Rocket Lab, OHB, Planet, and SpaceX spacecraft.3 This relationship is confirmed at the level of industry technical architecture rather than by itemised contract disclosure, and it should be read that way β these parts are on the qualified parts lists that spacecraft designs are built around, which is a durable position but not the same as a disclosed sole-source supply agreement.
Where the buses are built
Above the components sit the integrators. In Europe, OHB SE ($OHB) is the reference case: a Bremen-based prime that builds satellite buses and space infrastructure for the European Space Agency, most visibly the Galileo navigation constellation and meteorological satellites, generating more than β¬1 billion of revenue with space systems accounting for over 85% of the group.6 Ownership is unusual for a listed company β the Fuchs family holds more than 60%, with KKR taking roughly 20% following a June 2026 equity offering β which caps free float and, with it, index relevance.6
Pixxel sits at the opposite end of the same layer: a young integrator building the Fireflies hyperspectral constellation, which captures hundreds of narrow spectral bands rather than the three or four of conventional imaging, allowing mineral, crop-health, and pollution signatures to be identified directly from orbit. It has contracts with the US National Reconnaissance Office alongside commercial agriculture and mining customers.22 Its constraint is instructive and physical: hyperspectral data volumes are enormous, so the business is gated less by satellites than by downlink bandwidth and ground station density.
The pattern across all of this is consistent. In a launch-deflated world, the returns concentrate where a customer requirement is mandated, the qualification cycle is long, and the alternatives are few. Rocket Lab's deliberate rotation out of launch and into those positions is the clearest strategic playbook anyone in this industry has executed β and it exists because the alternative was demonstrated, expensively, by the companies that did not rotate.
5. The Capital Cycle: Boom, Bust, and Reusability Hegemony
In 2023, Virgin Orbit's assets went to auction. The company had built a genuinely clever system β a modified Boeing 747 carrying a rocket under its wing, releasing it at altitude to avoid the densest part of the atmosphere and to launch from any runway in the world. Its remains were divided among Rocket Lab, Stratolaunch, and Vast. The following year, Astra Space β which had gone public via SPAC with a plan for daily launches from anywhere β was taken private at a fraction of its listing value after repeated launch failures and a collapse in liquidity.
These were not accidents of execution. They were the predictable back half of a capital cycle, and the cycle ran with textbook fidelity.
The four movements
Capital floods in. Between 2020 and 2022, ultra-cheap money and the SPAC listing mechanism channelled well over $15 billion into speculative space ventures, including roughly twenty private launch startups. The thesis sold to investors was intuitive and wrong: small satellites were proliferating, therefore small satellites would need small rockets, therefore dedicated small launch was a large market.
Overcapacity meets a price the incumbents can't match. SpaceX began flying Transporter rideshare missions β a Falcon 9 packed with dozens of smallsats from different customers, sold by the kilogram at prices around $1,500 per kilogram against dedicated small launch at $12,000β$22,000.1 The value proposition of dedicated launch collapsed to a narrow band of customers who genuinely needed a specific orbital plane on a specific date and would pay a ten-times premium for it. That is a real business. It is not a business that supports twenty companies.
Rates rise and the music stops. Launch startups burn cash for years before first revenue, which makes them among the longest-duration assets in public markets. When real discount rates rose through 2023β2025, the present value of a business whose cash flows begin in 2029 fell sharply, and the equity windows through which those companies had been refinancing closed. Failed flight tests that would have been survivable in 2021 became terminal.
Consolidation. Assets were absorbed at distressed prices by survivors, capacity left the market, and entry barriers rose β not because regulation tightened, but because no rational investor would now fund a twenty-first small-launch company. As of 2026 the surviving Western launch platforms with credible scale economics number two.
Myth and reality: the small-rocket fallacy
The popular version of this story is that small-launch companies failed because their rockets failed. That is true as a proximate cause and misleading as an explanation, because it implies a better-engineered small rocket would have won.
The deeper problem is structural mass fraction. Every launch vehicle needs avionics, a flight computer, guidance, a flight termination system, tanks, plumbing, and engines. The cost and mass of many of those elements do not scale down proportionally with the vehicle. A small rocket therefore spends a much larger share of its total mass and cost budget on overhead per kilogram of payload delivered. Scale is not a preference in rocketry; it is baked into the physics of the mass ratio. A perfectly executed small expendable rocket still lands at multiples of Falcon 9's cost per kilogram, and against a rideshare option it is competing on schedule and orbit alone. Investors who funded twenty of them had mistaken a premium service niche for a volume market.
Rocket Lab survived this for two reasons, and only one of them was launch. Electron achieved a reliability record that made it the default choice for customers who genuinely needed dedicated small launch β the niche was small but Rocket Lab took most of it.3 The decisive move was the capital reallocation described in Section 4: pushing cash into subsystem acquisitions with better margins and defence lock-in, so that by the time small launch was fully understood as a niche, most of the company's revenue came from somewhere else. Rocket Lab entered mid-2026 with more than $500 million in cash and short-term investments and adjusted EBITDA approaching breakeven, with net income still suppressed by Neutron development spending.3 That is a survivable position. It was not available to a company that was only a rocket.
The remaining private launch entrant with genuine scale ambition is Firefly Aerospace, backed by AE Industrial, which has flown its Alpha small launcher multiple times, is developing a medium-lift vehicle, and has separately built a lunar delivery business around the Blue Ghost lander under NASA's Commercial Lunar Payload Services programme. Firefly's structure is the interesting part for investors: it is deliberately not a pure launch company either, pairing launch with government-funded lunar services in the same way Rocket Lab paired launch with subsystems. As a private company its financials are not disclosed, so no margin or cash-burn comparison against Rocket Lab is possible β a limitation worth stating rather than papering over with estimates.
The lunar layer: a different cycle, earlier stage
Two listed companies illustrate the capital-cycle stage below launch. Intuitive Machines ($LUNR) put the first commercial lander on the lunar surface with IM-1 and followed it with IM-2, and has expanded into NASA's Near Space Network communications contracts.14 ispace ($9348), listed in Tokyo, has built and flown HAKUTO-R landers and participates in the CLPS ecosystem.15 Both are attempting something with a genuinely binary technical outcome β a soft landing on another body, where the difference between success and total loss can be a few seconds of sensor behaviour.
Financially they belong to a distinct category: capital-intensive, largely pre-cash-flow, and dependent on equity and warrant issuance for funding.1415 The implication for a shareholder is that the relevant risk is not only whether a landing succeeds, but how many shares exist by the time the business generates cash. Dilution is the mechanism through which a correct thematic view about lunar logistics can still produce a poor per-share outcome. This is the single most reliable trap in early-stage thematic investing, and it does not show up in any addressable-market chart.
The capital cycle punished narrative and rewarded arithmetic. The next question is what happens when the buyer is a government that does not care about either.
6. Sovereign Access & The Geopolitical Space Race
Between 2022 and 2024, Europe experienced something it had spent forty years and tens of billions of euros to prevent: it could not reliably get to orbit.
The sequence was almost comically compounding. Ariane 5, the workhorse, was retired on schedule. Ariane 6, its replacement, slipped. Vega-C, the smaller vehicle, was grounded after a failure traced to its nozzle. And Russia's Soyuz, which had been launched from Europe's own spaceport in French Guiana under a cooperation agreement, became unavailable after the invasion of Ukraine and the sanctions that followed. Four routes to orbit, all closed at once.
The consequence was that Europe bought launches from SpaceX for flagship sovereign payloads β including the Euclid space telescope and, most pointedly, satellites for Galileo, the European navigation system explicitly built so that Europe would not depend on America's GPS.5 A programme whose entire rationale was strategic autonomy reached orbit on an American commercial rocket.
That episode explains most of European space policy since. But the deeper reason Europe found itself there is institutional, and it is worth understanding because it constrains the equity case for every European name in this article.
Juste retour: the price of a coalition
The European Space Agency operates under a principle known as juste retour β geographic return. Each member state receives industrial contracts roughly in proportion to what it contributes to the agency's budget.5 The political logic is unarguable: no government funds a multilateral agency whose work all flows to a neighbour's factories. The industrial consequence is that a European launch vehicle's supply chain is designed partly around a map of contributions rather than solely around cost and capability, and that consolidation which would be obvious commercially is politically impossible.
So Avio remains Italy's propulsion champion and the sole supplier of the P120C motors used by both Vega-C and Ariane 6.16 OHB remains Germany's satellite prime alongside Airbus Defence and Space and Thales Alenia Space, with European prime-contractor leadership genuinely contested and allocated tender by tender rather than won on cost.6 Both companies have visible, long-dated backlogs and defensible national positions. Both also operate at margins that reflect fixed-price government satellite manufacturing rather than commercial pricing power β OHB runs EBIT margins in the 6β8% range on gross margins around 18β22%.6 That is a defensive, high-visibility business with limited scope for return-on-capital expansion, and an investor should price it as such rather than as a space-growth asset.
Europe's response to the access crisis included subsidising a domestic micro-launcher cohort β Isar Aerospace and Rocket Factory Augsburg in Germany, PLD Space in Spain β through European Commission and ESA competition mechanisms.5 The strategic reasoning is sound. The commercial reasoning runs directly into Section 5's arithmetic: these are small expendable launchers entering a market whose cost curve has already destroyed the category commercially. Their survival will depend on whether sovereign demand is large and durable enough to substitute for a commercial market that does not exist at their price point.
India: deregulation as industrial policy
India's transformation is the cleanest natural experiment in the sector. Before 2020, space activity in India was effectively the Indian Space Research Organisation and nothing else. The government then created IN-SPACe, the Indian National Space Promotion and Authorisation Centre, as a single-window authorising body, alongside NewSpace India Limited as the commercial arm, permitting private firms to build launch vehicles, use ISRO facilities, and operate ground infrastructure.7 The 2023 Indian Space Policy formalised the roles, and 2024 foreign direct investment reforms allowed up to 100% foreign ownership in satellite manufacturing.7
Capital followed. Skyroot Aerospace flew Vikram-S, the first privately built Indian rocket to reach space, backed by investors including Singapore's GIC, and has been preparing the orbital Vikram-1.21 Agnikul Cosmos flew its 3D-printed engine from a private launchpad at Sriharikota β the first private launch facility on an ISRO range.20 Pixxel built its hyperspectral constellation with a US presence and US government customers.22
The genuinely interesting economic feature is cost structure. Indian engineering labour and ISRO's existing range infrastructure allow these companies to reach flight hardware on a fraction of the venture capital that comparable Western startups consumed. That is a real and durable advantage in development cost. It does not, however, repeal the mass-fraction arithmetic. A 300-kilogram-class Indian launcher faces the same competitive geometry against Falcon 9 rideshare as an American one did. The likelier path to value for the Indian cohort runs through domestic and regional sovereign demand, ISRO-adjacent contracts, and β in Pixxel's case β the downstream data layer, where cost advantages translate directly into gross margin rather than being competed away by a rideshare price list.
China: state capital with a commercial veneer
China's space sector is organised around two enormous state-owned enterprises, China Aerospace Science and Technology Corporation and China Aerospace Science and Industry Corporation, reporting through the State Council and supervised by SASAC. In 2014, State Council Document No. 60 opened the sector to private capital, and a commercial cohort emerged: LandSpace, Orienspace, Galactic Energy, and iSpace.10
In 2024, commercial space was formally designated a strategic emerging industry in the central government work report β an official statement of Party-state priority that reliably precedes the mobilisation of capital.10 Municipal funds in Shanghai, Beijing, and Wuhan have since financed the two national megaconstellations, Guowang under China Satellite Network Group and Qianfan/G60 under Shanghai Spacecom Satellite Technology.10 It is important to read that designation for what it is: authoritative evidence of what the Chinese state intends and how it frames the sector, and not by itself evidence of what has been achieved.
On achievement, the record is mixed and partly unverifiable. LandSpace's Zhuque-2 reaching orbit on methalox in 2023 was a genuine world first, confirmed by independent tracking.17 Orienspace's Gravity-1, launched from a sea platform with roughly 6.5 metric tons of capacity, is the largest privately built Chinese rocket to fly.18 Galactic Energy has the longest commercial operating record of the private Chinese cohort, with more than ten successful Ceres-1 flights and a kerosene-LOX Pallas-1 in development β the same solid-to-liquid transition that has broken other companies.19 Chinese commercial launch attempts reached roughly 22 in 2025.2
Where the evidence thins is reusability. Vertical-landing hop tests by LandSpace and others have been reported through company and state channels, but orbital-class stage recovery has not been independently corroborated in Western tracking data as of August 2026.217 An investor should treat Chinese reusability claims as unresolved rather than as either propaganda or fact. The distinction matters, because if a Chinese methalox reusable vehicle reaches operational cadence, the launch-cost floor becomes a two-supplier question rather than a one-supplier question β and geopolitical exclusion under US export control rules means those two suppliers serve largely non-overlapping customer sets.
That last point generalises. Export controls under ITAR and the EAR restrict what US space companies can sell abroad, which simultaneously caps their addressable market and protects their domestic pricing.1 Space is the purest dual-use domain there is, and sovereign defence budgets function as a floor under national champions everywhere β which is precisely why the public-market question is harder than it looks.
7. Public-Market Expressions & The Expectations Gap
For twenty years, the central frustration of thematic space investing was that the most important company in the industry was not purchasable. That ended in 2026, when SpaceX listed on Nasdaq. The listing valuation was around $1.8 trillion, and by early August 2026 the market capitalisation stood near $1.7 trillion.1 Overnight, every global growth portfolio acquired an implicit position on the space economy, whether or not it had an opinion about one.
A liquid mega-cap benchmark changes the analytical problem. It stops being "how do I get exposure" and becomes "what does the price already assume, and is there any layer of this industry where consensus is wrong about the mechanism rather than merely the multiple."
Exhibit 4 β Listed exposure and financial read-through, dated August 2026 Definition: disclosed revenue base and margin profile for the principal listed expressions; periods differ and are labelled. Currency as reported. Evidence status: company disclosure and regulatory filings; segment and consolidated figures kept distinct.
| Company (ticker) | Revenue base (period) | Gross margin | Profitability | Balance sheet | Exposure type |
|---|---|---|---|---|---|
| SpaceX (SPCX) | $7.81bn, Q2 2026 (+92% YoY); ~$18.7bn FY2025 | ~40β45% blended; Starlink >60%, launch ~35% | Q2 2026 net loss $541m; positive operating EBITDA | Strong post-listing liquidity | Pure-play flagship |
| Rocket Lab (RKLB) | $601.8m FY2025 (+38%); ~$680m TTM | ~28β32% blended; Space Systems ~32%, launch ~24% | Adj. EBITDA near breakeven; net loss on Neutron R&D | >$500m cash + short-term investments | Pure-play infrastructure |
| OHB SE (OHB) | >β¬1.0bn annual; 5β8% organic growth | ~18β22% | EBIT margin ~6β8% | Modest net debt; Fuchs family >60%, KKR ~20% | Diversified prime |
| Planet Labs (PL) | ~$240m TTM (+12β15%) | ~52β58% GAAP; >70% non-GAAP | Adj. EBITDA near breakeven | ~$280m cash, no long-term debt | Pure-play data |
Sources: SEC filings; Rocket Lab annual report; OHB investor disclosure; Planet Labs filings.13611
Reading this table aloud, the first thing to say is that these four companies are not comparable on any single line, and pretending otherwise produces nonsense. SpaceX's figure is a quarter; Rocket Lab's headline is a fiscal year; OHB reports in euros; Planet's growth is organic while a meaningful part of Rocket Lab's is acquired. What is comparable is the shape. SpaceX is growing revenue at 92% year over year while running a quarterly net loss of $541 million β a company deliberately converting a profitable core into future capacity. Rocket Lab is growing 38% at roughly half SpaceX's gross margin, with the mix improving as Space Systems grows faster than launch. OHB grows at single digits with EBIT margins around 7%, which is a defence-industrial profile rather than a technology one. Planet has software-like gross margins and hardware-like growth, which is the specific tension in its equity story.
What the prices appear to require
SpaceX. Consensus, as reflected in sell-side commentary since the listing, extrapolates consumer Starlink expansion toward a revenue run-rate approaching $1 trillion by 2030.1 The variant view worth stating is not that Starlink fails, but that the composition of that growth is likely to change in ways that affect margin. Consumer ARPU in developed markets tends to plateau: the customers who most value satellite broadband β rural households beyond fibre, maritime, aviation β are acquired first, and each subsequent cohort has lower willingness to pay. Growth beyond that point has to come from enterprise and government, from direct-to-cell, and from developing markets where ARPU is structurally lower and where spectrum licensing and local-partner requirements are slow and country-specific.8 Add the terminal manufacturing and working-capital constraint from Section 3, and the plausible bear mechanism is not a demand shortfall β it is that the marginal subscriber becomes progressively more expensive to acquire and less profitable to serve, exactly while Starship capital expenditure remains heavy.
The first smart reason a senior investor rejects SpaceX at this price is straightforward: at $1.7 trillion, the equity already capitalises a decade of flawless execution across two separate hard problems β reusable heavy launch and global consumer telecom β in a company that reported a loss last quarter. What would make it worth deeper work is evidence that defence and enterprise revenue, which carries different pricing dynamics, is growing faster than consumer. What would kill the thesis is a sustained plateau in terminal growth combined with Starship cadence stalling.
Rocket Lab. Here the dossier supports an actual variant view rather than a valuation quibble, and it is about classification. A meaningful part of the market still frames Rocket Lab as a small-launch company facing pressure from SpaceX rideshare β a framing that made sense in 2021 and is now roughly a third of the business. The evidence says something different: over 65% of revenue comes from Space Systems, at gross margins above 30%, with the Mynaric optical-terminal and Geost sensor positions embedded in mandated defence architectures and the Iridium transaction announced and pending.34 If that is right, a large share of Rocket Lab's earnings power is insulated from Neutron's schedule, because it does not depend on Neutron at all.
The counterargument a good short-seller would make deserves equal weight. Rocket Lab's growth has a substantial acquired component, which flatters the headline rate and complicates organic comparison. Serial acquisition of subsystem businesses carries integration risk and can mask flat underlying performance. Neutron consumes cash on a schedule that has already moved. And a company valued near $45 billion on roughly $680 million of trailing revenue is being priced on the completion of a strategy, not on its current cash flow.3 The observable evidence that would resolve this is mix disclosure: whether Space Systems margins hold above 30% as volume scales, and whether optical-terminal deliveries convert into recurring programme revenue rather than one-off tranche awards.
Planet Labs. The bull case treats Planet as a subscription data business with a decade-long proprietary archive and 70%-plus non-GAAP gross margins that should eventually re-rate toward software multiples.11 The constraint the market sometimes underweights is the sales motion. Planet's customers are governments, defence agencies, and large enterprises, and those procurement cycles run for quarters or years regardless of product quality. Revenue growth in the 12β15% range against gross margins above 50% describes a company whose limitation is distribution, not technology. That is a solvable problem, and it is not solvable quickly.
AST SpaceMobile. The valuation embeds aggressive assumptions about global cellular penetration.13 Section 3 laid out the mechanism risk; the securities point is that this is an option, and options have expiry dates set by financing. The company must fund a full constellation before it earns meaningful revenue, which means the number of shares outstanding at the moment of commercial success is itself a variable. A correct view on direct-to-cell can still lose money if it is expressed through the entity that has to issue equity to reach the finish line.
OHB and Avio. Both are legitimate exposures to European sovereignty spending with visible backlogs, and both are structurally capped. OHB's free float is small given the Fuchs family and KKR holdings, and its margins are set by fixed-price public procurement.6 Avio's revenue base of roughly β¬400 million is over 90% solid rocket motors and Vega-C, which makes it a genuine sole-source enabler and simultaneously a company whose fortunes hinge on the technical performance of one product family and the budget decisions of one agency.165 European defence missile demand provides a second, partially uncorrelated leg. Neither is a launch-deflation beneficiary; both are sovereignty-premium beneficiaries, which is a different trade with a different driver.
False positives
Two categories deserve explicit warning, because both are common in thematic screens.
The first is the legacy defence prime marketing itself as a space leader. Several large contractors derive under 5% of group EBIT from space activities while featuring space prominently in investor materials. Boeing is the sharpest illustration: its space segment is a small and dilutive part of group earnings, weighed down by fixed-price cost overruns on Starliner. Owning such a company as a space expression means owning, overwhelmingly, something else β and inheriting the cost structure and cost-plus institutional habits that proliferated LEO procurement was designed to bypass. Available disclosure does not establish measurable thematic exposure at the group level for these names.
The second is any remaining pure-play expendable small launcher without subsystem revenue or a reusability roadmap. Section 5 explained the arithmetic. The screen catches these companies because their revenue is 100% "space." The economics say their revenue is 100% exposed to a price that keeps falling beneath them.
The general lesson is that ticker-level enthusiasm and thematic exposure are only loosely related. The names most likely to produce excess return are the ones consensus has classified into the wrong industry.
8. Future Game Changers & Value Migration
In the Utah desert, a capsule roughly the size of a large kitchen appliance comes down under parachute. Inside it are pharmaceutical crystals grown in orbit by Varda Space Industries. The reason to grow them there is that in microgravity, without convection and without sedimentation, crystals form with a uniformity that is difficult or impossible to achieve inside Earth's gravity well β which for certain drug formulations changes bioavailability and shelf stability.
That capsule is the physical proof of a proposition that only becomes interesting at low launch prices: some things are worth more made in space than made on the ground. Three developments could move the industry's profit pools materially over the next several years, and each has a different mechanism, a different set of beneficiaries, and a different observable milestone.
Direct-to-cell becomes a standard rather than a product
The 3GPP standards body β the organisation that defines what "5G" means β has been incorporating Non-Terrestrial Network specifications into successive releases. If NTN support becomes a default feature of mainstream smartphone modems from Qualcomm ($QCOM), MediaTek, and Apple ($AAPL), the addressable market for satellite connectivity stops being "people who buy a terminal" and becomes "people who own a phone."
The mechanism matters more than the market-size arithmetic. Today, satellite broadband requires a customer to make a purchase decision and install hardware. Standardised NTN removes both steps: the capability is simply present, and the mobile operator activates it as a service tier. That collapses customer acquisition cost toward zero and shifts the commercial contest entirely onto the operator relationship β which, as Section 3 established, is where the satellite company has the least leverage.
The beneficiaries are the constellation operators with sufficient aperture and spectrum access to actually close the link: SpaceX's Starlink direct-to-cell service and AST SpaceMobile most directly. The chipmakers are gatekeepers rather than beneficiaries; NTN support is a feature in a modem they were selling anyway. The parties whose position improves most quietly are mobile network operators, who gain coverage without capital expenditure and retain the customer.
The adoption hurdles are regulatory and specific. Country-by-country authorisation to radiate terrestrial mobile spectrum from orbit is required, and it must be negotiated against incumbent terrestrial operators who may prefer to extend their own networks.8 The observable milestone is not a technology demonstration; it is bilateral spectrum approvals outside the United States β the European Union, Japan, and India in particular. Until those exist, direct-to-cell remains a US-centric service with an emergency-messaging fallback everywhere else.
In-space manufacturing: the freight-rate threshold
The economic test for orbital manufacturing is simple to state. A product is worth making in space if the value added by microgravity exceeds the round-trip transport cost. At $10,000 per kilogram up and an expensive, risky return, that test excluded almost everything. At a few hundred dollars per kilogram up with routine reentry capsules, the list of candidates lengthens.
Three candidate products recur. ZBLAN optical fibre β a fluoride glass that theoretically transmits with far lower loss than silica but which crystallises during terrestrial drawing, a defect microgravity suppresses. Semiconductor crystals grown without convective flows. And bioprinted tissue structures that collapse under their own weight on the ground. Varda and Redwire ($RDW) are the visible commercial actors, with Redwire operating manufacturing payloads on orbital platforms and Varda flying dedicated reentry capsules.
The honest assessment is that this remains at demonstration stage. The disputed question is whether microgravity ZBLAN production yields are actually high enough, at achievable volumes, to beat improving terrestrial fibre manufacturing on a total cost basis β evidence on that is company-reported and not independently established. Separate announced capability from scaled commercial adoption here rigorously: a returned capsule containing product is a milestone, and it is not a production line. The observable threshold is repeat commercial orders from a non-governmental customer at a stated price, which nobody has published.
If it does scale, the value migration is significant: orbital freight becomes an input cost to a manufacturing business rather than the business itself, and the profit pool moves to whoever owns the process IP β which is a chemicals and pharmaceuticals competency, not an aerospace one.
Kessler syndrome: the risk that reprices everything
The counter-scenario is physical and it is the reason the bull case is not unconditional.
In a sufficiently crowded orbital shell, a single collision between two large objects generates thousands of fragments, each capable of causing further collisions, in a cascade that can render a band of altitudes unusable for decades. This is Kessler syndrome, and the 500β600 kilometre shell where most megaconstellation satellites operate is where the object density is highest.
The financial mechanism by which this damages investors is not primarily the loss of satellites. It is insurance and regulation. A material rise in conjunction events would raise premiums across every constellation operator, and β more decisively β would put launch licensing itself into play. The FAA licenses US launches and the FCC imposes orbital debris rules, including a five-year post-mission disposal requirement.98 A regulator responding to a debris cascade has an obvious lever: slow or halt new deployments. For a business model built on continuously replacing satellites every three to five years, a licensing pause is not an inconvenience; it is an existential interruption of the replacement cycle that the entire thesis depends on.
The beneficiaries in that world are space domain awareness and debris remediation specialists β Astroscale in removal, LeoLabs in tracking β and the losers are precisely the vertically integrated constellation operators that the base case favours. That inversion is the most useful thing about this risk: it is one of the few developments that would require rotating away from the names the rest of this analysis points toward.
Where value goes if launch becomes free
Run the thesis to its logical end. Suppose heavy reusable launch reaches genuine utility economics below $100 per kilogram. What happens to the launch layer?
It becomes infrastructure with utility-like returns: high volume, essential, and structurally unable to capture much of the value it enables. Exhibit 3 already shows this process underway, with launch falling from roughly 35% to under 18% of industry gross profit while volumes tripled. The extrapolation is that launch continues toward a low-teens share.
Value migrates in two directions. Upward, into the businesses that only become possible at the new price β orbital manufacturing, large-aperture direct-to-cell, orbital compute if power and thermal constraints can be solved. And laterally, into the bottleneck components that no amount of transport deflation makes cheaper: optical terminals, high-efficiency solar, radiation-tolerant processing, and the ground segment. The consistent pattern in industrial history is that when a transport cost collapses, the transport provider rarely keeps the surplus. The customers and the irreplaceable suppliers do.
The destination of launch deflation is an extraterrestrial industrial economy where orbit is a location on a supply chain rather than a destination. Whether the current owners of that supply chain capture the value is a separate question, and it turns on a small number of measurable things.
9. Crux KPIs, Monitoring Dashboard & Thesis Kill Criteria
Everything above compresses into a small number of observable quantities. The discipline that matters is choosing indicators that sit upstream of financial results rather than restating them. Aggregate market size, revenue growth, and historical share are outcomes; by the time they move, the decision has already been made for you.
Five measurements sit directly on the binding constraints of this industry.
Exhibit 5 β Crux KPIs, latest readings and thresholds, August 2026 Definition: leading physical and procurement observables selected for causal proximity to the thesis. Evidence status: latest readings are dated observations or disclosed estimates as noted; thresholds are analytical judgments, not forecasts.
| KPI | Latest reading (Aug 2026) | Source / cadence | Confirms thesis | Breaks thesis |
|---|---|---|---|---|
| Heavy reusable launch cost per kg | ~$1,400/kg (Falcon 9); Starship targeting <$300 | SpaceX filings, contract disclosure / semi-annual | <$300/kg on a commercial payload flight | >$1,000/kg sustained through 2028 |
| Starlink active terminals | ~4.2 million | SpaceX 10-Q / quarterly | >7 million by end-2027 | Sequential decline or plateau below 5 million |
| Defence proliferated-LEO backlog | ~$6.5bn (SDA PWSA Tranches 1β2) | DoD/USSF budget justification / annual | Combined US+EU LEO backlog >$10bn | Tranche 3 cancellation or budget freeze |
| Optical terminal unit cost / yield | ~$150,000/unit; ~85% yield | Rocket Lab and DoD procurement disclosure / quarterly | <$80,000/unit at 1,000+ units/yr | Yield sustained below 70% |
| Asian commercial launch frequency | India 3 private test flights; China ~22 commercial attempts (2025) | IN-SPACe, CNSA registers, Space-Track / monthly | >40 combined private launches/yr | Private failure rate above 40% |
Sources: SEC filings; SDA and DoD procurement disclosure; Rocket Lab disclosure; IN-SPACe and CNSA registers; Space-Track.1347102
Now the reasoning behind each, because a threshold without a mechanism is just a number.
Launch cost per kilogram is the master variable, and it leads everything else by roughly two years. It is upstream of satellite design choices, which are upstream of constellation size, which is upstream of revenue. The bull-bear disagreement it settles is precise: bulls assume Starship's refurbishment burden falls with iteration the way Falcon 9's did; bears assume heat-shield replacement and pad refurbishment impose a floor around $1,000 per kilogram that no cadence fixes. The measurement problem is real and should be acknowledged β SpaceX has no incentive to publish true internal cost, so the honest proxy is commercially disclosed pricing on third-party payloads, which lags internal cost and includes margin.1 Use it as a limited proxy rather than pretending to precision. If commercially quoted heavy-launch pricing remains above $1,000 per kilogram through 2028, the upstream belief that started this article is substantially wrong.
Starlink terminals measure whether the demand side absorbs what the supply side produces. This is the single cleanest test of downstream monetisation, and it now has a quarterly disclosure cadence through SEC filings.1 It leads revenue because terminals are installed before subscriptions compound. The disagreement it settles is whether the addressable consumer market saturates near six million households or extends toward ten million and beyond via enterprise, maritime, aviation, and direct-to-cell. A plateau below five million with flat ARPU would indicate that the profitable cohorts were acquired early and the remainder are uneconomic to serve.
Defence LEO backlog is the indicator that decouples the thesis from consumer behaviour entirely. Governments do not buy proliferated LEO architectures because they are cheap; they buy them because distributed constellations are harder to destroy than a handful of exquisite satellites. That logic is insensitive to interest rates and consumer sentiment, which makes this backlog the sector's most stable revenue foundation and the reason subsystem suppliers have pricing power at all.4 The kill signal is a cancellation or freeze of the next tranche, which would indicate the US defence establishment reverting to concentrated high-orbit architectures β and would remove the mandate that gives optical terminals their bottleneck position.
Optical terminal cost and yield is the most granular indicator here, and the most diagnostic. It sits exactly on the chokepoint described in Section 4. If unit costs fall toward $80,000 at volumes above a thousand units a year, mesh networking becomes standard on commercial as well as defence constellations, and the profit pool at the subsystem layer widens. If yields fall below 70%, deliveries slip, prime contractors miss milestones, and constellation deployment schedules across the industry move to the right. This is the KPI most specific to the Rocket Lab variant view, because it directly tests whether the Mynaric position is a durable bottleneck or a temporary scarcity that competitors will compete away.3
Asian commercial launch frequency tests whether the cost floor is set by one company or several. SpaceX's pricing power exists because no one else can fly reusable heavy launch at scale. If Chinese and Indian private launchers reach meaningful combined cadence β and, more importantly, if Chinese methalox reusability is independently confirmed β the structure changes from a single global cost leader to a bifurcated market with a Western supplier and a Chinese one serving largely separate customers under export-control lines.210 The reading here is deliberately conservative because Chinese reusability claims remain uncorroborated in Western tracking data.
Theme kill criteria versus security kill criteria
These are different, and conflating them is a common error.
The theme dies if launch cost stalls above $1,000 per kilogram through 2028, or if a debris cascade triggers a launch licensing moratorium, or if direct-to-cell spectrum sharing is denied broadly enough to confine satellite cellular to emergency messaging.89 Any one of those breaks the causal chain from cheap transport to abundant orbital infrastructure.
A security dies for narrower reasons. Rocket Lab's specific thesis breaks if Space Systems gross margins compress below the mid-twenties as volumes scale, indicating that the subsystem positions were scarcity rents rather than structural power β regardless of what happens to launch costs. Planet's breaks if revenue growth decelerates below ten percent while gross margins hold, which would indicate a distribution problem no product improvement solves. AST's breaks on financing terms rather than on technology. The theme can be right while any of these is wrong.
Common factor risk: the exposures you own twice
Three shared exposures run underneath every name in this article, and an investor holding several of them is less diversified than the ticker count suggests.
The first is duration. These are long-dated cash flows, and pre-revenue constellation operators in particular behave like the longest-duration assets in the equity market. A rise in real discount rates hits them simultaneously and hard, independent of any operational news. A portfolio holding four space names holds one interest-rate position four times.
The second is a physical single point of failure. An anomaly at Cape Canaveral's SLC-40 or LC-39A, or at Starbase, could ground a substantial share of global launch capacity for months.9 That would freeze deployment for every satellite operator waiting in the manifest β including operators who compete with SpaceX and buy launch from it. The correlation runs through concrete, not through markets.
The third is regulatory concentration. ITAR and EAR export controls, FCC spectrum and debris rules, and FAA launch licensing are all US instruments applying to most of the sector's economic value.189 A single policy shift propagates across the entire holding set.
The analytically interesting consequence is that the pathways within this theme are not all pointed the same direction. Launch deflation is a cost tailwind for satellite operators and a revenue headwind for launch providers. Sovereignty spending benefits European primes largely independently of the launch cost curve. Debris risk is a direct negative for constellation operators and a direct positive for tracking and remediation specialists. Those are genuinely offsetting causal exposures rather than correlated ones, and mapping them is more useful than counting positions.
What the evidence now says about the upstream belief
Return to the proposition this article set out to test: that collapsing launch cost converts space from a capital-rationed strategic domain into an industrial frontier.
On the evidence available in August 2026, the belief is holding on its first link and unresolved on its second.
The first link β that transport cost governs everything downstream β is confirmed about as firmly as an industrial proposition can be. Payloads in orbit grew roughly eightfold in a decade, mass to orbit grew from 450 to 2,950 tons in seven years, and the entirety of that growth traces to one company's reusable vehicles and to state programmes imitating them.12 Satellite design has genuinely shifted from exquisite to disposable. Defence procurement genuinely restructured around it. Profit pools genuinely migrated out of launch, which is the paradoxical proof that launch became a utility.
The second link β that the next order of magnitude arrives β is not yet established. Starship has not demonstrated the operational cadence or the refurbishment economics that sub-$300-per-kilogram requires, and management's cost claims remain disputed by independent engineering assessment.1 The heavy reusable row of Exhibit 2 remains a target. Everything downstream of it β orbital manufacturing, orbital compute, genuinely mass-market direct-to-cell β is contingent on a number nobody has yet observed.
That leaves a specific and unglamorous conclusion. The abundance thesis has already paid out once, in the transition from expendable to first-generation reusable launch, and that payout is now visible in the price of the companies that delivered it. Whether it pays out a second time depends on a heat shield, a regulator's tolerance for orbital density, and a spectrum negotiation in a dozen capitals. Those are the things to watch. Everything else is a consequence.
Glossary
ADCS (Attitude Determination and Control System) β The subsystem of reaction wheels, magnetorquers, and star trackers that orients a spacecraft. It matters commercially because pointing accuracy determines whether an imaging satellite or a laser link works at all, and it is one of the subsystem niches where a handful of qualified suppliers hold pricing power.
Cislunar space β The region from low Earth orbit out to and including the Moon's orbit, roughly 384,400 kilometres. It defines the operating domain of lunar logistics companies and is where technical risk shifts from launch to landing.
Direct-to-cell (D2C) β Satellite service to an unmodified consumer smartphone using the mobile operator's own spectrum. It matters because it eliminates the user terminal, which has been the practical bottleneck on satellite broadband adoption, and because it hands commercial leverage to the mobile operator.
Geostationary orbit (GEO) β 35,786 kilometres altitude, where a satellite's orbital period matches Earth's rotation so it appears fixed in the sky. Convenient for coverage, disqualifying for latency-sensitive applications β which is why the industry's growth moved to LEO.
IN-SPACe β India's single-window authorising body for private space activity, created in 2020. It converted Indian space from a state monopoly into a licensable market and is the reason a private Indian launch cohort exists.
Juste retour β The European Space Agency principle allocating industrial contracts in proportion to member-state contributions. It preserves the political coalition that funds European space and simultaneously blocks the industrial consolidation that would lower costs.
LEO (Low Earth Orbit) β 160 to 2,000 kilometres altitude. Low latency and cheap to reach, but satellites move relative to the ground, which is why continuous coverage requires thousands of them rather than three.
Methalox β Liquid methane and liquid oxygen propellant. Chosen for next-generation vehicles because it burns without soot, making engines fast to turn around between flights β a manufacturing and labour advantage rather than a performance one.
OISL (Optical Inter-Satellite Link) β Laser terminals letting satellites pass data directly to one another without touching the ground. Mandated in US defence constellations, which converted a niche component into one of the industry's genuine bottlenecks.
Phased array β A flat antenna that steers its beam electronically via hundreds of individually phase-controlled elements, with no moving parts. It is what makes tracking fast-moving LEO satellites practical, and its manufacturing cost has repeatedly gated subscriber growth.
PWSA (Proliferated Warfighter Space Architecture) β The US Space Development Agency's layered LEO constellation for missile tracking, data transport, and battle management. It is the single largest structural demand source for proliferated satellite hardware and the reason certain subsystems have mandated demand.
Rideshare launch β Selling capacity on one rocket to many small customers by the kilogram. Falcon 9 rideshare pricing near $1,500 per kilogram is what made dedicated small launch commercially unviable for most payloads.
SCS (Supplemental Coverage from Space) β The FCC framework permitting satellites to transmit on terrestrial mobile spectrum. It is the regulatory precondition for direct-to-cell, and equivalent authorisations must be obtained country by country.
Structural mass fraction β The share of a launch vehicle's mass consumed by structure, engines, and avionics rather than propellant or payload. It scales badly downward, which is the physical reason small rockets cost far more per kilogram than large ones.
VTVL (Vertical Takeoff, Vertical Landing) β Propulsive landing of a rocket stage on its engines. The technique underlying first-stage reuse, and the specific capability whose independent verification in China remains unresolved.
References
-
SpaceX (SPCX) Form 10-Q, IPO prospectus and related filings, 2026 β US Securities and Exchange Commission, EDGAR ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩
-
Satellite catalogue and orbital object tracking data β Space-Track.org, US Space Force ↩↩↩↩↩↩↩↩↩
-
Rocket Lab USA FY2025 Annual Report and 2026 acquisition disclosures β Rocket Lab Investor Relations ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩
-
Proliferated Warfighter Space Architecture, Tranche 1 and Tranche 2 award documentation β US Space Development Agency ↩↩↩↩↩↩
-
Ariane 6, Vega-C and Space Transportation Directorate reporting, 2024β2026 β European Space Agency ↩↩↩↩↩↩↩
-
OHB SE financial disclosures and June 2026 capital increase β OHB SE Investor Relations ↩↩↩↩↩↩
-
Indian Space Policy implementation and private authorisation registry β IN-SPACe ↩↩↩
-
Supplemental Coverage from Space framework, spectrum and orbital debris filings β US Federal Communications Commission ↩↩↩↩↩↩
-
Commercial space transportation licensing and launch records β US Federal Aviation Administration, Office of Commercial Space Transportation ↩↩↩↩
-
Chinese national space programme and commercial space policy notices β China National Space Administration ↩↩↩↩↩
-
Planet Labs PBC constellation, product and financial disclosures β Planet Labs ↩↩↩
-
LEMUR constellation and data services disclosures β Spire Global ↩
-
BlueBird constellation, mobile network operator agreements and disclosures β AST SpaceMobile ↩↩↩
-
IM-1 and IM-2 lunar missions and NASA Near Space Network contracts β Intuitive Machines ↩↩
-
HAKUTO-R lunar lander programme disclosures β ispace, inc. ↩↩
-
Vega-C, P120C solid rocket motor and propulsion disclosures β Avio S.p.A. ↩↩↩
-
Zhuque-2 and Zhuque-3 vehicle programme disclosures β LandSpace Technology ↩↩↩
-
Ceres-1 and Pallas-1 launch programme disclosures β Galactic Energy ↩
-
Agnibaan SOrTeD and Agnite 3D-printed engine programme β Agnikul Cosmos ↩↩
-
Vikram-S and Vikram-1 launch vehicle programme β Skyroot Aerospace ↩
-
Fireflies hyperspectral constellation and customer disclosures β Pixxel ↩↩
LinkedIn
WhatsApp
Reddit
Hacker News
Email
Kindle
RSS Feed
Spotify
Apple Podcasts
Amazon Music
Audible
YouTube