The Gravity Well: How Reusable Launch Collapsed Orbital Costs, Fractured Space Tech, and Created the Ultimate Capital Cycle Test

1. The Tower Catch at Starbase and the $1,500 Breakthrough

In October 2024, on the south Texas coast at Starbase, a 232-foot stainless-steel booster came back from the edge of space, relit its engines, slowed to a hover beside the launch tower it had left seven minutes earlier, and was caught in mid-air by two mechanical arms that closed around its grid fins and took the weight.¹ The Super Heavy stage never touched the ground. Engineers called the tower "Mechazilla," and the footage travelled the world as spectacle.

The spectacle is the least interesting part. What mattered was an accounting change disguised as an engineering stunt.

For six decades, the first stage of an orbital rocket was a consumable. An Atlas booster, an Ariane 5 core, the first stage of a Long March — each was a machine costing tens of millions of dollars, engineered to the tolerances of a Swiss instrument, designed to operate for roughly eight minutes and then fall into the ocean. The industry's cost structure followed from that single fact. If the vehicle dies on every flight, then every flight must carry the full amortised cost of building a vehicle, and the only way to make the arithmetic tolerable is to fly rarely, charge enormously, and load each flight with a payload valuable enough to justify the ride. Rarity begat expense; expense begat rarity. Satellites, in turn, were designed to that regime: built over five to seven years, gold-plated against failure because there would be no second chance, and parked in geostationary orbit 35,786 kilometres up, where one spacecraft could cover a third of the planet for fifteen years.

The cost curve tells the story more plainly than any narrative. Around 2010, delivering a kilogram of payload to low Earth orbit on a Western expendable vehicle cost somewhere between $10,000 and $20,000, depending on how much of the launch provider's fixed infrastructure you allocated to the flight.³ By 2025, SpaceX's partially reusable Falcon 9 — a vehicle whose first stage returns, is inspected, and flies again — had brought the internal marginal cost of putting a kilogram into low orbit to roughly $1,500.³ That is a reduction of about 90%, achieved not through a new propellant, a new material, or a breakthrough in physics, but by refusing to throw the expensive part away.

The consequences arrived at industrial scale. In calendar 2025, SpaceX flew 165 Falcon 9 and Falcon Heavy missions and placed on the order of 2,400 tonnes of hardware into orbit, a figure that represented the great majority of all mass humanity delivered to space that year.¹² A single company, flying roughly one mission every 53 hours, moved more material off the planet than every space programme on Earth had managed in most previous decades combined.

Here is where the story turns, and where investors have repeatedly misread it.

Cheap launch does not make launch companies rich. It makes launch a utility, and utilities earn utility returns — unless the utility is also a monopoly. Notice the gap between two numbers that are frequently confused. The roughly $1,500 per kilogram figure is an estimate of what it costs SpaceX to fly. The price a third-party customer pays to put a small satellite on a Falcon 9 Transporter rideshare mission has run in the range of $5,500 to $6,000 per kilogram.¹³ The cost collapsed by 90%. The price to outside buyers fell by far less. The difference — several thousand dollars a kilogram, across thousands of tonnes — is the wedge that funds everything else SpaceX does.

That wedge is also the trap laid for everyone else. A rival building a small launch vehicle is not competing against SpaceX's cost. It is competing against SpaceX's price, which is set high enough to harvest margin and low enough to make dedicated small rockets uneconomic, with room to fall further if anyone gets close. Competitors have been fighting a price floor that their opponent can lower at will and still profit.

The damage propagated up the supply chain in ways that are easy to miss. United Launch Alliance, the Boeing–Lockheed joint venture that for years held the United States government's most sensitive launches, buys its GEM-63XL strap-on solid rocket boosters from Northrop Grumman, which is effectively the sole domestic source for large solid motors of that class.⁴ Sole-source status normally means pricing power. It means considerably less when the vehicle those boosters bolt onto flies a handful of times a year against a competitor flying more than a dozen times a month. A monopoly supplier to a low-cadence customer is a monopoly on a small number. The lesson generalises: in this industry, the value of a chokepoint depends entirely on the throughput passing through it.

Which brings us to the problem facing the reader this article is written for.

Take the default decision context: an institutional public-equity investor, global mandate, a three-to-seven-year thematic horizon, no position sizing and no security recommendations, trying to decide whether the collapse in launch cost is an investable theme or merely a true story. That investor faces an immediate structural insult. The company that caused the change is private. SpaceX's launch economics, Starlink's subscriber economics, and Starship's development spending are disclosed only when SpaceX chooses to disclose them, which is selectively and without audited segment detail.¹ Every listed instrument in this theme is, in some sense, an adjacency: a supplier, a challenger, a customer, a displaced incumbent, or a company operating in the shadow of a competitor whose accounts nobody can inspect.

This is the governing tension of the entire industry, and of this article. The forecast can be completely correct — space access has become cheap, orbital mass is compounding, satellites are becoming consumer-electronics-like objects manufactured on lines rather than in laboratories — and the security outcome can still be terrible, because the investor chose the wrong layer of the value chain, the wrong capital structure, or the right company at the wrong price. Between 2021 and 2024, dozens of investors were right about space and lost most of their money anyway.

To understand why, we have to go back further than the tower catch, to the belief that made this industry worth examining in the first place.

Subscribe on LinkedIn

↑ Back to Top

2. The Macro Belief: Triangulating Mass, Spectrum, and Silicon

Satellite manufacturing used to be a craft. A commercial geostationary communications satellite was a bespoke object, assembled over years by technicians who could name every unit on the bus, tested to exhaustion because a defect discovered in orbit was permanent, and priced accordingly. The transition that has actually occurred over the last six years is that a class of satellites stopped being made that way. They are now produced in batches, on lines, with a design lifetime measured in a handful of years and a unit cost that assumes some will fail and be replaced rather than that none may fail at all. That is a different manufacturing philosophy, a different supply chain, and a different set of winners.

The upstream belief that sent us into this industry can be stated as a single falsifiable proposition:

When the unit cost of delivering mass into a new physical domain falls by an order of magnitude, capital cascades into infrastructure and services built in that domain — but durable equity value accrues only to assets with a distribution moat, proprietary spectrum, or a sole-source component bottleneck, because the transport layer itself commoditises.

This proposition has two halves, and both are testable. The first half predicts observable physical and financial behaviour: mass deployed should rise non-linearly, capital allocation by firms outside the space industry should redirect toward orbital infrastructure, and the component supply chain should show demand and margin expansion. The second half predicts something more specific and more uncomfortable: that the layers doing the visible, heroic work — building and flying rockets — should show weaker economics than the layers selling the unglamorous inputs and the layers controlling end-customer distribution.

Three independent evidence streams triangulate the first half.

The first is physical and hard to fake. Mass delivered to orbit is measured by independent trackers, notably the orbital catalogue maintained by the astronomer Jonathan McDowell, whose Jonathan's Space Report has for decades recorded every launch attempt, payload and orbital decay.² Before 2020, global annual mass to orbit ran below roughly 500 tonnes a year. In 2025, it ran several times that.²³ A caution on the arithmetic is warranted here, because the underlying figures are quoted inconsistently across industry sources: the roughly 2,400-tonne figure for 2025 is best understood as the mass attributable to SpaceX's own missions, and if SpaceX accounted for the commonly cited 80%-plus of global orbital mass that year, the global total must have been materially higher — on the order of 3,000 tonnes.¹² We flag the tension rather than averaging it away, and use the direction of travel, which is not in dispute, rather than a false decimal point.

The second stream is behavioural, and it comes from companies that have no reason to be sentimental about space. Apple committed several hundred million dollars to Globalstar to underwrite the satellite capacity behind Emergency SOS messaging on iPhones, a commitment that turned a struggling mobile satellite services operator into a capacity supplier for the world's most valuable consumer hardware franchise.⁹ T-Mobile contracted with SpaceX for direct-to-cell service using Starlink satellites. AT&T, Verizon and Vodafone signed commercial agreements with AST SpaceMobile to carry satellite traffic into their own spectrum and onto ordinary handsets, and several took equity positions alongside those agreements.⁸ On the government side, the US Space Development Agency has awarded contracts across Tranche 0, Tranche 1 and Tranche 2 of its Proliferated Warfighter Space Architecture, buying hundreds of small satellites for missile tracking and battlefield data transport on two-year development cycles rather than the decade-long acquisition rhythms that defined military space procurement for a generation.⁶

The third stream is the corroborating signal from suppliers who sell into space but are not of it. HEICO's Electronic Technologies Group supplies space-qualified microelectronics alongside its broader defence and commercial aerospace electronics; TransDigm's portfolio includes proprietary valves, connectors and actuators qualified for launch and spacecraft applications; Teledyne's e2v operation supplies radiation-hardened image sensors and data converters to the European Space Agency's programmes and to Airbus Defence and Space.¹¹¹²¹⁵ Meanwhile the terrestrial communications industry has been re-architecting its silicon: the 3GPP standards releases that define modern cellular networks incorporated non-terrestrial network provisions, which is what allows an unmodified handset to talk to something moving at 27,000 kilometres an hour rather than to a tower on a hill.

Put those three together and you have a structural claim rather than a cyclical one. Physical throughput up several-fold; buyers outside the industry reallocating capital; the supply chain retooling. A cyclical launch backlog does not produce all three at once.

Here is the base evidence in compact form.

Exhibit 1 — Global space activity, observed series, 2020–2025

Read that table aloud and three things jump out. Launch attempts roughly doubled in five years, but SpaceX's own launches rose more than six-fold, so essentially all of the industry's cadence growth came from one company's flight rate rather than from a broad-based expansion of providers. Satellite count roughly tripled while launch attempts doubled, which tells you satellites got smaller and travelled in batches. And the rideshare price line went up over the period, not down, while the underlying cost of flying fell sharply — the clearest single piece of evidence that the cost collapse has been captured by the provider rather than passed to buyers.

Now the transmission mechanism into industry economics, which is where most thematic analysis of space goes wrong.

Cheaper launch lowers the barrier to deploying a constellation. It does not lower the cost of owning one. A satellite in low Earth orbit at 500 to 600 kilometres flies through the thin upper atmosphere and loses altitude continuously; it carries propellant to fight that drag, and when the propellant is gone the satellite comes down. Operators plan three-to-five-year service lives at those altitudes, against fifteen years or more for a geostationary spacecraft. So the capital intensity does not disappear when launch gets cheap; it migrates. It moves out of the launch line item and into a permanent, recurring satellite manufacturing and replacement line item. Total industry capital expenditure rises even as unit launch cost falls, because falling launch cost is precisely what makes deploying thousands of short-lived satellites rational in the first place.

That migration is the mechanism that determines who makes money. If the recurring spend is on satellites rather than rockets, then the pricing power sits with whoever supplies the parts of a satellite that cannot easily be second-sourced — space-grade solar, radiation-tolerant electronics, precision valves and actuators, optical terminals — and with whoever owns the customer relationship and the spectrum rights at the far end. The layers in between, which assemble buses and fly rockets, get squeezed from both sides.

One example makes the chain concrete. Globalstar contracted MDA Space to build a fleet of new satellite buses for the constellation refresh that underpins its Apple capacity commitment.¹⁰⁹ Globalstar owns the spectrum and the customer contract; MDA builds the hardware on a contracted basis; Apple pays for capacity it resells to no one, folding it into a handset feature. Three companies, three completely different economic positions: a spectrum owner with one enormous customer, a contract manufacturer with programme risk, and a consumer platform monetising the whole thing indirectly through hardware differentiation.

The same belief that sent us here implicates several sibling industries — terrestrial mobile operators contemplating whether to buy remote coverage from orbit rather than build it, defence electronics primes converting from exquisite single spacecraft to proliferated fleets, compound semiconductor and photonics suppliers, and the maritime and aviation connectivity markets replacing steerable dishes with flat electronically steered arrays — and we will touch each only where it changes the economics of space itself.

The proposition has a clock and it has kill conditions. Over a three-to-seven-year horizon, it would be falsified if fully reusable heavy vehicles fail to demonstrate rapid turnaround and sub-$500-per-kilogram economics; if orbital debris density in the crowded 500–600 kilometre shells forces regulatory launch restrictions; or if terrestrial network densification makes satellite direct-to-cell economically pointless outside genuinely remote geography. Each of those is observable before it shows up in anyone's revenue line, and we return to them at the end.

First, though, we need to understand why reusability was hard, why it stayed hard for so long, and why the moat it created cannot be crossed with money alone.

↑ Back to Top

3. The Physics and Economics of Reusability: How SpaceX Rewrote the Cost Curve

At the Stennis Space Center in Mississippi, NASA operates test stands where rocket engines are bolted down and fired for the length of a mission while instruments record every oscillation in the combustion chamber. In 2025 and 2026 those stands were being used to qualify Archimedes, the engine Rocket Lab designed for its medium-lift Neutron vehicle.⁵ Engine qualification is where launch programmes live or die, and it is worth understanding why, because it explains the entire competitive structure of this industry.

Start with the arithmetic every rocket obeys. The rocket equation, formulated by Konstantin Tsiolkovsky in 1903, says that the velocity a vehicle can gain equals its exhaust velocity multiplied by the natural logarithm of the ratio between its starting mass and its final mass. The logarithm is the villain. Reaching low Earth orbit requires roughly 9.3 kilometres per second of velocity change once you include atmospheric drag and gravity losses, and because the relationship is logarithmic, each additional increment of performance demands an exponentially larger mass of propellant. The practical result is that an orbital rocket at liftoff is about 95% propellant and structure and about 3-4% payload. Everything in launch vehicle engineering is a fight over that last few percent.

Reusability makes the fight harder before it makes it cheaper. Landing a first stage means reserving propellant for the return burn, carrying landing hardware, and building a structure that survives a second trip through the atmosphere at high speed. Every kilogram of that comes directly out of payload. A reusable rocket is a worse rocket on any single flight. It only wins if it flies many times.

Which is why the choice of propellant became a strategic decision rather than a chemistry preference. Falcon 9 burns RP-1, a refined kerosene, with liquid oxygen. Kerosene is dense and well understood, but it leaves carbon deposits — soot — in the engine's turbopumps and combustion chamber, and soot means inspection and refurbishment between flights. Methane burns cleanly. That is the reason Starship, Blue Origin's New Glenn, LandSpace's Zhuque series, and ArianeGroup's Prometheus programme all chose methane: it is the molecule that makes an engine plausibly reusable twenty times rather than a few times, and turnaround time is what converts reusability from an engineering achievement into an economic one.¹¹⁶

The standard analogy is aviation, and it is a good one as far as it goes. If a Boeing 747 were dropped into the Atlantic after each crossing, a New York–London ticket would cost something on the order of half a million dollars, and there would be no airline industry. Reusability is what makes commercial aviation, and now commercial space access, an industry rather than an event. But the analogy has a limit worth stating, because it misleads on the crucial point. An airliner flies with enormous structural margins, is refurbished continuously, and completes hundreds of cycles a year, so its capital cost spreads across a large denominator. A rocket flies at the absolute edge of its structural margin, and even the busiest launch vehicle in history flies a few times a week across an entire fleet. Fixed costs — launch pads, range operations, engineering staff, insurance — amortise across a far smaller number of flights. That is why cadence, rather than reusability by itself, is the variable that actually moves cost per kilogram, and why a company that can land boosters but only flies six times a year captures very little of the theoretical benefit.

SpaceX solved the cadence problem with two decisions made long before the first landing. The first was vertical integration: manufacturing the large majority of its engines, avionics, airframes and fairings in-house rather than buying from the tier-one aerospace supply base. That removed a layer of subcontractor margin, but more importantly it removed subcontractor lead times, which is what allows a design change to reach the factory floor in weeks.¹ The second was a development culture that flies hardware early, breaks it, and iterates — a method that is financially rational only if you own the factory and are not being paid on a cost-plus contract to deliver a document package.

Legacy primes could not copy this, and the reason is institutional rather than technical. A cost-plus government contract pays a fee on incurred cost and penalises failure in review boards; a geographically distributed supply base exists partly to satisfy political constituencies; and a workforce optimised for verification is not a workforce optimised for iteration. None of those are stupidities. They are rational responses to the incentives those firms actually face. But they produce an organisation that cannot fly a prototype expecting it to explode.

The flywheel that followed is the part investors most often underrate. SpaceX does not disclose audited margins, so the following is inference rather than fact: with an internal cost near $1,500 per kilogram and an external price near $6,000, commercial missions plausibly generate gross margins far above anything in traditional aerospace.¹³ Those margins funded Starlink. Starlink, in turn, gave SpaceX a captive launch customer with essentially unlimited appetite, which pushed the flight rate up, which spread pad and range fixed costs across more flights, which lowered internal cost per flight, which improved the margin on external missions. Each loop tightened the next. By 2025 the company was simultaneously the largest launch provider, the largest satellite operator, and its own largest customer.

Now look at how the rest of the world's launch industry is positioned against that.

Europe's answer, Ariane 6, made its maiden flight in July 2024 and entered operational service thereafter, restoring sovereign heavy-lift access after Ariane 5's retirement and the loss of Soyuz launches from French Guiana following the invasion of Ukraine.¹⁵ Ariane 6 is a capable vehicle and a genuine strategic asset. It is also expendable, and its strap-on boosters are P120C solid rocket motors built by Avio of Italy, which supplies the same motor as the first stage of the Vega C small launcher.¹⁵¹⁶ Solid motors are simple, storable and powerful. They are also, by their nature, not reusable in any economically meaningful sense — you cannot refill a solid casing at the pad. Europe therefore built its sovereign capability on an architecture that structurally cannot ride the reuse cost curve, a decision taken in the mid-2010s when reuse was still contested. ArianeGroup's response has been to spin out MaiaSpace, a subsidiary developing a small reusable vehicle around the methane-fuelled Prometheus engine — an acknowledgement, in corporate form, that the parent's flagship cannot get there.¹⁶

Rocket Lab took a different route to relevance. Its Electron vehicle uses a carbon-composite airframe and Rutherford engines whose turbopumps are driven by electric motors and batteries rather than by a gas generator, an unusual architecture that trades some performance for manufacturing simplicity, and whose major components are 3D-printed. Electron passed 100 lifetime flights and became the default dedicated ride for operators who need a specific orbit at a specific time rather than a seat on a rideshare bus.⁵ That is a real leadership position, defined precisely: Rocket Lab leads the Western market for dedicated small-satellite launch, measured by flights flown, and it leads because it industrialised engine production early and built its own launch complexes. What it did not have was a vehicle large enough to deploy a constellation, which is where the volume went.

The moat, then, is not a patent. It is an accumulated stack of manufacturing know-how, flight data from hundreds of recoveries, launch infrastructure, and an organisational metabolism — plus a cadence advantage that compounds. Capital alone cannot buy it, which is the single most important observation for anyone evaluating a challenger. Blue Origin has had access to essentially unlimited capital since inception and reached orbit with New Glenn in January 2025, roughly a decade after Falcon 9 first landed.² Money bought the vehicle. It did not buy the decade.

But a moat around launch is only valuable if launch is where the money is. It increasingly is not — and to see why, we need to follow the payloads.

↑ Back to Top

4. The Adoption Waterfall: Megaconstellations, Direct-to-Cell, and the 5-Year De-orbit Clock

A customer in rural Montana, or on a container ship in the mid-Atlantic, opens a box containing a flat panel and a power supply, sets it where it can see the sky, and has broadband within minutes. There is no installer visit, no alignment procedure, no professional survey. The panel is an electronically steered phased array that finds satellites by itself, and the satellites are 550 kilometres overhead rather than 35,786, so the round-trip latency is tens of milliseconds instead of hundreds — the difference between a connection that supports a video call and one that does not.

That user experience is the entire adoption story of the last five years, and it is worth being precise about what it displaced. Legacy satellite broadband meant a fixed dish aimed at a geostationary spacecraft, professionally installed, with latency around half a second, capacity shared across an enormous coverage footprint, and service that degraded in heavy rain. It was a product people bought because they had no alternative. Low Earth orbit broadband is a product people buy because it is good enough to be a first choice in places terrestrial fibre and fixed wireless do not reach.

Starlink's subscriber base moved from roughly 10,000 users in 2020 to approximately 6 million by 2025.¹ Against a plausible full-year revenue run-rate in the multiple billions of dollars, that makes Starlink, on its own, one of the larger communications businesses created in the past decade. A caution on the denominator: there is no reliable measure of the true addressable market for satellite broadband, because it depends on terrestrial network build-out plans, national subsidy programmes and household willingness to pay, none of which are consistently observable. Anyone quoting a precise penetration percentage is inventing the denominator. The honest proxy is growth against a known constraint — capacity per satellite and per coverage cell — and on that measure the binding limit has consistently been how quickly satellites can be manufactured and lifted, rather than how quickly customers can be found.

The adoption waterfall runs through three distinct tiers, and they are at very different stages.

The first tier, fixed and mobile broadband from low orbit, is commercially established. Starlink is scaled and generating cash. Amazon's Kuiper constellation is deploying, funded by a corporate parent that can absorb years of negative cash flow. Legacy operators are the displaced party: EchoStar's Hughes satellite broadband business faces a competitor offering better latency and comparable speeds, and its consumer subscriber economics have deteriorated accordingly.¹⁴

The second tier, direct-to-cell, is early and genuinely uncertain. It began with the narrowest possible product: emergency text messaging from an ordinary iPhone via Globalstar's satellites, a feature Apple underwrote rather than a service consumers buy.⁹ It is now moving toward voice and data, which is a far harder problem. A satellite must close a radio link with a handset whose antenna is a few centimetres long and whose transmit power is a fraction of a watt, from hundreds of kilometres away, while moving fast enough to cross the sky in minutes. AST SpaceMobile's answer is brute-force physics: unfold an enormous phased array in orbit — its BlueBird spacecraft use arrays on the order of 693 square feet — so the satellite's side of the link makes up for the phone's weakness.⁸ SpaceX's answer is the opposite: smaller per-satellite capability, multiplied by a very large number of satellites and by owning the launch that puts them there. Both approaches are being tested commercially at the same time, which is unusual and analytically useful, because within a few years the market will show which physics-versus-scale trade wins.

The third tier is Earth observation and defence. Planet Labs operates a fleet of more than 200 satellites imaging the entire landmass of the planet daily, selling subscriptions to that archive rather than individual pictures.¹³ The Space Development Agency is building proliferated constellations for missile warning, missile tracking and tactical data transport, buying in tranches on two-year cycles.⁶ These are different businesses with a common feature: value depends on persistence and revisit rate, which requires many satellites, which was impossible when launch was expensive.

Now the constraint that governs all three tiers, and that most enthusiastic modelling ignores.

Satellites in the crowded shells between 500 and 600 kilometres do not stay up. Atmospheric drag pulls them down, and operators must expend propellant to hold altitude. When it runs out, the satellite deorbits. Regulators have made this a legal obligation as well as a physical one: the US Federal Communications Commission requires operators to dispose of low-orbit satellites within five years of mission completion, a rule tightened specifically to keep pace with the arrival of megaconstellations.⁷ Design lives of three to five years are therefore the working assumption for this class of spacecraft.

Run the arithmetic. A 5,000-satellite constellation with a five-year service life must build, test, launch and commission 1,000 satellites every single year purely to stand still. At around 7,000 active spacecraft — roughly where Starlink sat in 2025 — the steady-state replacement rate is on the order of 1,400 a year.¹² The unit cost of those satellites is not publicly disclosed by any operator at scale; if one assumes, purely illustratively, $1 million apiece, the 5,000-satellite case implies $1 billion a year of replacement capital expenditure in perpetuity, before any launch cost, ground infrastructure, spectrum payments or growth.

That is the shape of the business. A low-orbit constellation is not a fixed asset that depreciates on an accounting schedule; it is a fixed asset that physically falls out of the sky on a schedule, and the depreciation charge is a cash charge. Free cash flow only exists above that treadmill. This is why the same cost collapse that makes constellations feasible also makes them dangerous investments: it lowers the cost of entering a business whose defining characteristic is a permanent, non-deferrable reinvestment requirement.

The treadmill also explains a piece of supply-chain theatre that would otherwise be baffling. Amazon, building a constellation intended to compete directly with Starlink, contracted SpaceX for Falcon 9 launches to deploy Kuiper satellites.¹ Amazon had bought launch capacity from Blue Origin, from United Launch Alliance and from Arianespace, and still could not get spacecraft into orbit fast enough against its own regulatory deployment deadlines. Rather than miss the schedule, it paid its principal competitor to carry its payloads. That single transaction tells you where the leverage sits: when your business model requires continuous mass to orbit and one supplier controls most of the world's available capacity, competitive positioning yields to arithmetic.

Compare two ways of financing the treadmill and you see the strategic logic of the direct-to-cell approach. EchoStar's route to a next-generation network runs through building terrestrial wireless infrastructure funded with substantial debt, against declining legacy pay-TV and satellite broadband cash flows.¹⁴ AST SpaceMobile's route runs through mobile network operators: the operators own the customer relationship, the billing, the retail presence and, critically, the spectrum in each market, and AST supplies coverage into that spectrum from orbit under revenue-sharing arrangements.⁸ The second model requires no retail distribution, no handset subsidy and no store network. It also concedes a share of the economics to partners and depends entirely on their continued commitment. The exact split has never been disclosed by either side; published estimates range from roughly even revenue shares to per-subscriber add-on fees in the range of $5 to $10 a month, and investors should treat any modelled figure as an assumption rather than a fact.⁸

So adoption is real, accelerating in broadband, embryonic in direct-to-cell, and structurally durable in defence. The question is whether the money that gets spent to serve it ever returns to shareholders. To answer that, we have to open the value chain and look at where the gross profit actually sits.

↑ Back to Top

5. Anatomy of the Value Chain: From Triple-Junction Solar to Ground Terminals

Consider an object a few centimetres across that costs more per square metre than almost anything else on a satellite: a triple-junction solar cell. Terrestrial solar panels are silicon, cheap and abundant. In orbit they would be a poor choice, because they convert a smaller fraction of incident sunlight and because unshielded silicon degrades under proton and electron bombardment in the radiation environment. Space power instead uses stacked layers of compound semiconductors — typically gallium arsenide and indium gallium phosphide — each tuned to absorb a different band of the solar spectrum, delivering roughly double the efficiency of terrestrial silicon in a fraction of the area and surviving years of radiation exposure.

Very few firms in the world make them at qualified space grade. That is the shape of the entire upstream layer of this industry: small markets, extreme reliability requirements, multi-year qualification cycles, and consequently very few suppliers. It is also, as we will see, where a disproportionate share of the industry's gross profit lives.

Walk the chain from the bottom up, in the order the material actually flows.

Layer one: components and specialised materials. Space solar cells, radiation-tolerant microelectronics, precision valves and actuators, carbon composite structures, optical transceivers, reaction wheels, star trackers. Rocket Lab acquired SolAero, one of the small number of Western producers of triple-junction space solar cells, and supplies power hardware into US defence prime programmes and NASA missions.⁵ HEICO's Electronic Technologies Group and Teledyne's e2v supply space-qualified electronics and radiation-hardened image sensors, the latter into European Space Agency programmes and Airbus spacecraft.¹¹¹⁵ TransDigm's proprietary fluid and motion control components appear across launch and spacecraft applications.¹² These businesses share a characteristic that matters more than any of their individual technologies: qualifying a replacement part for a flight programme takes years and costs money that nobody wants to spend to save a few percent on a component representing a rounding error in programme cost. That is textbook switching-cost power, and it is why these margins persist.

Layer two: satellite buses and payloads. MDA Space builds satellite payload subsystems, antennas and space robotics — including the Canadarm3 robotic system for the planned lunar Gateway — and manufactures buses under contract, including for Globalstar's constellation refresh.¹⁰ York Space Systems and, before its acquisition, Terran Orbital built small buses at volume; Airbus Defence and Space, Thales Alenia Space and Germany's OHB SE occupy the European equivalent position, with OHB serving as a principal contractor on the Galileo navigation satellites and a beneficiary of expanded German defence space spending.¹⁵ This layer takes the squeeze from both directions. Its customers are large, concentrated and increasingly buy on firm-fixed-price terms; its inputs come from suppliers with the pricing power described above.

Layer three: launch and propulsion. SpaceX, Blue Origin, Rocket Lab, Arianespace and ArianeGroup, Avio, Mitsubishi Heavy Industries, China's CASC, and the Chinese private launchers. Discussed at length already; the essential point is that reported economics in this layer vary enormously depending on cadence and on whether the provider also owns the payload.

Layer four: ground terminals and optical equipment. Mynaric develops laser communication terminals for satellite-to-satellite links; Kratos Defense supplies ground systems and satellite command-and-control; Gilat and Cobham Satcom supply terminals and modems for maritime, aviation and enterprise use. The migration here is from mechanically steered dishes to flat electronically steered arrays, which changes the cost structure from precision mechanics toward semiconductors and, eventually, toward volume economics. Whether that transition creates durable profits or simply invites consumer-electronics-style price deflation is an open question; the observable evidence so far is that scaling optical terminal production to constellation volumes has been the binding constraint, not designing them.

Layer five: constellation network operators. Starlink, Amazon Kuiper, AST SpaceMobile, EchoStar, Globalstar, and the state-directed Chinese constellations. This is where the recurring revenue lives and where the treadmill capex is paid.

Layer six: downstream data, analytics and defence integration. Planet Labs and Spire Global convert raw sensing into subscriptions; BlackSky and analytics platforms including Palantir sit on top of imagery streams; L3Harris, Lockheed Martin and Northrop Grumman integrate payloads and constellations for government customers.

Now the profit map.

Exhibit 2 — Characteristic gross margin bands by value-chain layer

Units are percentage of revenue; geography is global; period is the 2024–2026 window; status is a mix of company-disclosed segment figures and industry ranges, and the bands should be read as characteristic rather than precise.

Read that aloud and the shape is a barbell. The two ends of the chain — the specialised inputs at the bottom and the customer-facing services and data at the top — carry gross margins in the forties, fifties and sixties. The middle, where the actual spacecraft and rockets get built and flown, runs in the teens, twenties and low thirties. That is the profit-pool map of this industry in one line: the glamour is in the middle and the money is at the ends.

There is a crucial qualification on the right-hand column, and it is the one that separates careful analysis from theme-chasing. A 50–70% service gross margin at a constellation operator is calculated before the cost of replacing satellites that fall out of the sky. Depending on how an operator capitalises and depreciates its fleet, an apparently spectacular gross margin can coexist with permanently negative free cash flow. Comparing a constellation operator's gross margin to a component supplier's is comparing incomparable things unless replacement capital expenditure is brought back into the frame.

Why does the upstream layer keep its margin? Run Porter's framework against observed behaviour rather than assertion. Buyer power should be high — the customers are enormous defence primes and well-funded constellation operators — but it is blunted because the component is a small share of programme cost and requalifying an alternative costs years of schedule. The threat of substitution is low because performance requirements are absolute: there is no cheaper cell that also survives the radiation. New entrants face a barrier that is regulatory and reputational as much as technical, since flight heritage is itself the product being sold. In Hamilton Helmer's vocabulary, this is switching-cost power reinforced by process power — accumulated manufacturing know-how that cannot be bought — and in a few cases a cornered resource, where the only qualified production line in the West sits inside one company.

Two honest caveats belong here. First, high margin can reflect scarcity rather than power, and scarcity ends when demand justifies a second line; the arrival of megaconstellation volumes is exactly the condition that could attract new capacity into space solar and radiation-tolerant electronics. Second, and more importantly for equity investors, the companies best known for these margins do not break out space. HEICO reports a Flight Support Group and an Electronic Technologies Group; TransDigm reports power and control, airframe and non-aviation segments.¹¹¹² Neither discloses space revenue as a segment. Estimates that HEICO derives 15–20% of revenue from space, or TransDigm 8–12%, are analytical constructions, not disclosures, and an investor buying either company on a space thesis is in practice buying a commercial aerospace aftermarket business with a space option attached.

At the constellation-operator end, the source of power is different and much more legible: spectrum. Rights to transmit on specific frequencies in specific countries are granted by national regulators and coordinated internationally, they are scarce, and they are not manufacturable. Globalstar's value rests substantially on its licensed spectrum, including the band designated n53 in the 3GPP framework, which is why a company with modest revenue and a legacy satellite fleet became strategically important to Apple.⁹ AST SpaceMobile's model rests on borrowing its partners' spectrum rather than owning it, which is capital-efficient and simultaneously its principal dependency.⁸

The chokepoints, stated directionally: Rocket Lab supplies space solar into US defence primes and NASA programmes, and there is no abundant Western alternative.⁵ Northrop Grumman supplies large solid rocket motors to United Launch Alliance's Vulcan, with no domestic second source at that class.⁴ Avio supplies the P120C solid motor to ArianeGroup for Ariane 6 and to its own Vega C, making one Italian company a single point of failure for European access to space.¹⁵ L3Harris supplies infrared tracking payloads into the Space Development Agency's tracking layer, where the qualified supplier base is small and the customer is a single government agency.⁶ Teledyne e2v supplies radiation-hardened sensors into European programmes.¹⁵ MDA supplies buses and payload subsystems to Globalstar and to Telesat's Lightspeed constellation.¹⁰ In every case, ask both directions: who supplies this company, and whom does it supply? A firm with a monopoly upstream and a single customer downstream has less power than its margin suggests, because the customer's failure is its failure.

That interdependence is precisely what makes the next part of the story so brutal. When the money stopped, it stopped for everyone at once.

↑ Back to Top

6. The Capital Cycle Trap: Depreciation, Constellation Overbuild, and the TAM Fallacy

In 2021, a particular slide circulated through investment committees. It showed global telecommunications revenue — a figure north of a trillion dollars — multiplied by an assumed satellite penetration rate, producing a space total addressable market large enough to justify almost any valuation for almost any company with a rocket or a satellite in its pitch deck. Special purpose acquisition companies took roughly a dozen space businesses public against projections built on that logic.

By 2023, Virgin Orbit had filed for bankruptcy and been liquidated, its air-launch system sold for parts.³ Astra Space had exited the orbital launch business after repeated failures. Terran Orbital, a satellite bus manufacturer that had won large constellation contracts, ran into liquidity distress under fixed-price manufacturing commitments and was acquired by Lockheed Martin, its largest customer and shareholder.⁴ Most of the SPAC cohort traded down by 80% or more from their listing-era peaks.

Nothing about the underlying forecast was wrong. Launch costs did collapse. Satellite deployment did explode. The mass really did go up five-fold. The investors were right about society and wrong about securities, which is the most expensive way to be right.

The mechanism is the capital cycle, and space runs a textbook version of it.

Exhibit 3 — Where each segment sits in the capital cycle, July 2026

Geography is global; period is as at July 2026; classification is analytical judgment applied to the observable evidence listed, not a disclosed metric.

Read across that table and the picture is of an industry running several different cycles at once. That is unusual and it matters, because it means "space" as a portfolio category bundles a shakeout, a scaling business, a pre-commercial option and a government-funded demonstration project into one label. Any exposure decision that treats them as a single bet is making a category error before it makes a valuation error.

The six-stage rhythm underneath is familiar to anyone who has watched shipping, semiconductors or shale. Venture capital floods a newly legible opportunity. Capacity is built ahead of demand. Prices fall as the capacity arrives, most acutely for the least differentiated product. Weak balance sheets fail; assets are bought by strategic acquirers at a fraction of invested capital. The survivors, holding consolidated capacity, finally earn returns. And then — in this industry uniquely — the replacement cycle begins, and the survivors discover that a meaningful share of the cash they just started generating must go straight back into rebuilding the asset.

That final stage is what distinguishes space from most capital cycles, and it is where the total addressable market fallacy does its damage. The 2021 slide multiplied a large number by a penetration assumption. It did not subtract the reinvestment required to hold the position. Return to the earlier arithmetic: at a purely illustrative $1 million per satellite, a 5,000-satellite fleet with a five-year life burns $1 billion a year forever, plus the cost of launching them, plus ground network expansion, plus spectrum, before a single new subscriber is added. Revenue growth in that structure tells you almost nothing about shareholder value. The relevant questions are the return on incremental invested capital, and whether subscriber cash contribution compounds faster than the fleet decays.

There is a second fallacy layered on top, and it is one of composition. Add up the market shares implied by the business plans of Starlink, Kuiper, the two Chinese national constellations, Europe's IRIS² programme, and the various regional and defence systems, and the total substantially exceeds the plausible market. They cannot all be right. In a capacity business where supply arrives in indivisible constellation-sized increments, and where several of the largest participants are funded by strategic or sovereign motives rather than return requirements, the clearing price does not have to settle anywhere near the level that makes commercial operators' models work. Competing against a state that does not need a return is a genuinely different problem from competing against a company.

Against that backdrop, one company's escape route is instructive. Rocket Lab's leadership recognised early that dedicated small launch had a hard ceiling on both volume and margin, and responded by buying its way up the value chain: SolAero for space solar power, Sinclair Interplanetary for reaction wheels and satellite components, and Advanced Aero Structures for composite manufacturing.⁵ The result is that a company known publicly as a rocket company generates the majority of revenue — around two-thirds — from space systems and components rather than from launch.⁵ The launch business supplies visibility, customer relationships and strategic optionality; the components business supplies the gross margin. It is a deliberate response to the profit-pool map in Exhibit 2, executed years before most investors had drawn that map.

The same lens applied to the industry's failures is unsparing. Virgin Orbit built an air-launch system whose economics required a flight rate it never achieved. Astra pursued extremely low-cost small vehicles whose reliability record destroyed the customer base before the cost curve arrived. Terran Orbital took fixed-price manufacturing risk on constellation contracts in an inflationary supply environment with insufficient balance sheet to absorb overruns. In each case the failure mechanism was the same: a capital-intensive business with negative operating leverage, financed by equity markets that closed. When financing is a factor of production, the cost of capital is an operating input, and the 2022 rate shock hit these companies through their income statements, not merely through their multiples.

Which raises the obvious question for anyone allocating capital today. If the cycle punished the last cohort this thoroughly, what is different about the participants that survived — and about the ones being built, at state expense, outside the United States?

↑ Back to Top

7. The Global Map: Reusability Rivalries in Europe, India, Japan, and China

Two pieces of infrastructure built in the last few years tell you most of what you need to know about how the rest of the world responded to American launch dominance. One is the commercial launch complex at Wenchang on Hainan Island, where China built dedicated pads for commercial operators rather than making them queue behind the state programme. The other is SaxaVord, a spaceport on Unst in the Shetland Islands, at the northern edge of the United Kingdom, built to give European small launchers a high-latitude site of their own.¹⁶ One was constructed as an instrument of national industrial policy. The other was constructed in the hope that a commercial market would materialise. That difference has shaped outcomes.

Europe. The European launch crisis of the early 2020s had three simultaneous causes: Ariane 5 retired, Vega C was grounded after a failure, and Russian Soyuz launches from French Guiana ended with the invasion of Ukraine. For a period, Europe — a continent with a serious space industrial base and its own navigation constellation — could not reliably reach orbit on its own vehicles and bought rides from SpaceX for European institutional payloads. Ariane 6 restored heavy-lift sovereignty from its July 2024 maiden flight onward.¹⁵ Sovereignty, though, was purchased at the price of competitiveness: as discussed, the architecture is expendable and built around Avio's solid boosters, and no amount of production efficiency closes a cost gap against a reusable competitor flying an order of magnitude more often.

Europe's commercial answer has been a cohort of venture-funded launch startups: Isar Aerospace in Germany, which flew the first test flight of its Spectrum vehicle in March 2025; Rocket Factory Augsburg, preparing its RFA One for a maiden flight from SaxaVord; PLD Space in Spain, which demonstrated the suborbital Miura 1 and is developing the orbital Miura 5; and MaiaSpace, ArianeGroup's reusable spin-off.¹⁶ They compete for the same institutional payloads, from the same small set of European government customers, at a moment when the European Space Agency is being pushed to reform the "geographic return" principle under which contracts were allocated in proportion to each member state's contributions rather than by open competition.¹⁵ Geo-return produced a distributed industrial base and predictable politics; it also produced a supply chain optimised for fairness rather than cost. The reform debate is the most consequential institutional story in European space, because it determines whether the continent's startups get anchor demand or merely encouragement.

As at this article's cutoff, no European private company had placed a payload in orbit on its own vehicle.

China. China runs two tracks deliberately. The state track is operated by the China Aerospace Science and Technology Corporation (中国航天科技集团) and its principal launch vehicle developer CALT, which fly the Long March (长征) family and execute the crewed Shenzhou and lunar Chang'e programmes.² These entities sit under state ownership supervision, and their budgets and priorities follow national plans rather than commercial return calculations.

The commercial track was created on purpose. Municipal and provincial investment vehicles — the same "new infrastructure" funding mechanisms used to seed semiconductor and electric vehicle clusters — capitalised private launch companies and built the launch sites, engine test stands and industrial parks they needed around Wenchang, Jiuquan and coastal Shandong. The results have been rapid. LandSpace (蓝箭航天) put its methane-fuelled Zhuque-3 into orbit in December 2025, making it the leading Chinese private developer of a reusable liquid vehicle; the first stage's recovery attempt suffered an anomaly at touchdown, so the reuse half of the demonstration remains unproven in routine operation.² Space Pioneer (天兵科技) is developing the larger Tianlong-3 after the success of Tianlong-2. Orienspace (东方空间) flew Gravity-1 (引力一号), among the largest solid-propellant commercial rockets in service.²

Behind those launchers sit two enormous state-backed constellations: Qianfan (千帆), also known as the G60 or Thousand Sails project and backed substantially by Shanghai municipal capital, and Guowang (国网), the national network. Together their filings envisage well over 13,000 satellites.² Judging their commercial economics from outside is not currently possible: unit costs, subscriber figures and financing terms are not disclosed in any form an outside investor can verify, and there is no listed pure-play expression of either. What can be established is the strategic intent, the launch cadence required, and the fact that Chinese launch demand is now driven primarily by domestic constellation deployment rather than by external customers.

India. India's transformation is institutional before it is technological. In 2020 the government created the Indian National Space Promotion and Authorisation Centre, IN-SPACe, to authorise private space activity and open ISRO's test stands, launch complexes and technical expertise to companies that could never have afforded to build them.¹⁷ That single administrative decision converted India's deep public space capability into an input private startups could rent.

The payoff arrived in July 2026, when Skyroot Aerospace's Vikram-1 reached orbit, the first orbital launch by a privately developed Indian vehicle and the milestone that placed India third among nations with private orbital launch capability.¹⁷ Agnikul Cosmos took a different technical path, flying its Agnibaan SOrTeD demonstrator with a semi-cryogenic engine manufactured as a single 3D-printed piece — an approach that collapses part count and assembly labour, which matters more in a market where vehicle volumes are low. On the payload side, Pixxel operates the Firefly hyperspectral imaging constellation and has won United States defence customers, an unusual direction of travel for an Indian space startup; Dhruva Space supplies satellite buses and the orbital deployers that ride on ISRO's PSLV. India's advantage is a genuinely low engineering cost base attached to a mature state technical apparatus. Its constraint is domestic demand: the anchor customer set is small, and the export market runs into the same Falcon 9 rideshare price floor everyone else faces.

Japan. Japan's position is the most balanced of the non-US space powers, and the most legible to public investors because the relevant companies are listed in Tokyo. Mitsubishi Heavy Industries (7011.T) developed and now operates the H3 launch vehicle, which replaced the long-serving H-IIA and is the backbone of Japanese sovereign access; MHI is also a major defence prime, and its space activity is a modest share of a large industrial conglomerate.¹⁸ ispace (9348.T) is attempting to build a commercial lunar delivery business with its HAKUTO-R landers; neither of its first two missions completed a successful soft landing, which is an honest reminder that lunar surface delivery remains a frontier engineering problem rather than a service.²⁰ Astroscale (186A.T) has become the most credible commercial actor in in-space servicing, having flown the ADRAS-J mission that rendezvoused with and inspected a derelict rocket upper stage in orbit under contract with the Japanese space agency, with additional work funded by the UK Space Agency.¹⁹ iQPS (5595.T) operates small synthetic aperture radar satellites — radar imaging that works through cloud and at night — with the Japanese Ministry of Defence among its customers.¹⁸

Astroscale deserves a precise leadership statement rather than a vague one. On the parameter of demonstrated rendezvous and proximity operations with a non-cooperative piece of orbital debris, by a commercial company, on a government contract, Astroscale leads, with the closest benchmark being Northrop Grumman's Mission Extension Vehicles, which docked with and extended the life of commercial geostationary satellites — a related but easier problem, because those targets were cooperative and stable.¹⁹ The durability of that lead is moderate at best, because the market depends on regulation that does not yet exist. Debris removal today is bought by governments as a demonstration. It becomes an industry only if someone is compelled to pay for cleanup.

Step back and the global map resolves into a single insight for investors. Outside the United States, launch and satellite manufacturing are not primarily commercial markets. They are sovereign capability programmes with commercial wrappers. That has two consequences pulling in opposite directions. It creates durable demand floors — a European, Japanese, Indian or Chinese government will keep buying domestic launch at uncompetitive prices because the alternative is dependence — which protects national champions from the full force of SpaceX's cost curve. It also caps returns, because a business whose customer is a government buying sovereignty is a business whose pricing is negotiated politically. The equity implication is that these names are more defensive and less explosive than their thematic labels suggest.

↑ Back to Top

8. The Public Market Expression Ledger: Moats, False Positives, and Variant Wedges

Screen for "space" on any institutional terminal and the output is a mess: a high-margin aerospace aftermarket compounder, a pre-revenue satellite developer trading on physics, a leveraged pay-television operator, a Canadian robotics contractor, and three defence primes for which space is a fifth of revenue. These businesses share a keyword and almost nothing else. Sorting them requires asking four questions of each name: what proportion of economics is genuinely exposed to the theme, what role it plays in the chain, whether current expectations already embed the thematic outcome, and what would have to be observed for the thesis at that specific company to break.

Exhibit 4 — Exposure and expectations map, listed expressions, as at July 2026

Definitions: space exposure is share of revenue attributable to space activities, disclosed where a company reports it and estimated otherwise; financial figures are latest available run-rates as at July 2026 in each company's reporting currency, not normalised for fiscal year ends; expectations columns are analytical judgment, not consensus data.

Read that aloud and one row should stop you. HEICO and TransDigm are the two companies most frequently cited as high-quality ways to own the space supply chain, and neither discloses space as a segment. Available public disclosure does not establish measurable space exposure for either; the commonly quoted percentages are estimates built from product descriptions. That is a materially different evidentiary position from Rocket Lab or Planet Labs, whose entire revenue base is space, and an investor should not treat the two situations as equivalent because both appear on the same screen.

Now the two names where the dossier evidence supports an actual expectations gap, stated with their strongest counterarguments.

Rocket Lab. The common framing treats Rocket Lab as a small launch company facing structural competition from Falcon 9 rideshare. The variant reading follows from Exhibit 2: roughly two-thirds of revenue comes from space systems and components — solar power, satellite subsystems, separation systems, constellation buses — which occupy the high-margin end of the value chain and sell to everyone building satellites, including customers who will never buy a launch.⁵ On that reading, Rocket Lab is a component and subsystems compounder carrying an embedded option on medium-lift launch. Neutron, if it flies and works, addresses the one segment with genuine capacity scarcity: medium-lift constellation deployment, where Space Development Agency tranches and commercial constellation operators need an alternative to a single dominant provider.⁶

The first thing a sharp investor would say in rejection: Neutron has slipped. Its debut moved into 2026 while Archimedes engine qualification continued.⁵ Launch programmes slip as a rule, and each slip costs cash at a company already running negative free cash flow on development spending. A second objection is subtler and better: the Space Systems margin, in the high twenties to low thirties, is respectable but sits below the component-monopoly band, which suggests Rocket Lab's mix includes a good deal of contract manufacturing rather than pure proprietary product. For this thesis to merit deeper work, an investor would want Neutron to reach orbit and Space Systems gross margin to expand rather than merely grow. What would kill it: Neutron cancellation or slippage beyond 2027, or Space Systems gross margin compressing below 20%, which would signal that constellation customers, not suppliers, hold the pricing power.

AST SpaceMobile. The sceptical case is easy to state and hard to dismiss: a company attempting an unprecedented physics problem, dependent on regulatory approvals it does not control, funded by repeated equity issuance, with revenue that barely exists. The variant reading rests on two assets. The first is technical and legal: a large patent estate covering the very large unfolding phased array and its operation, which AST has stated exceeds 3,400 claims.⁸ The second is distributional: binding commercial agreements with AT&T, Verizon and Vodafone, several accompanied by strategic investment, which give AST access to partner spectrum and to hundreds of millions of existing handsets without acquiring a single retail customer itself.⁸ If satellite-to-standard-handset broadband works at commercial quality, the combination of patents and carrier agreements looks less like a product and more like a toll on an infrastructure layer.

The first rejection a senior investor would offer: patents are not physics, and none of this matters if the regulator does not permit the power flux density required for a usable link, or if a major carrier walks. The dilution history is a second, entirely legitimate objection — shareholders who are right about the technology can still lose because the capital required arrives at the shareholder's expense. What would kill it: a Federal Communications Commission denial of the operational power waivers, or a tier-one carrier cancelling its agreement.⁷

For the remaining names, discipline requires saying plainly that the evidence does not establish an expectations gap. Planet Labs is a genuinely differentiated data business with strong gross margins, a clean balance sheet and no debt, whose growth decelerated as commercial imagery buyers proved harder to convert than the model assumed; it is worth monitoring, and the current evidence supports no more than that.¹³ MDA Space is a well-run contractor whose economics depend on backlog conversion and fixed-price execution rather than on any thematic re-rating.¹⁰ The defence primes offer stable cash flows and diluted thematic sensitivity, which makes them ballast rather than expression. Globalstar's economics are the economics of one contract; the spectrum is real and the customer concentration is extreme.⁹

Two categories deserve explicit false-positive treatment. The first is the displaced incumbent that screens as a beneficiary: EchoStar carries satellite broadband, spectrum holdings and pay-television in one entity, and the theme that is supposedly its tailwind is in fact the competitive force degrading its consumer satellite business, while its balance sheet is levered against a terrestrial wireless build.¹⁴ The second is the suborbital and adjacent-defence group — companies whose space association comes from tourism flights that carry no payload to orbit, or from legacy geostationary military communications lines facing secular decline. Neither converts theme language into cash flow.

Two structural cautions close this section. Comparing these companies requires care: Rocket Lab and Planet report in US dollars, MDA in Canadian dollars, the Japanese names in yen; growth at Rocket Lab includes acquired businesses while Planet's is largely organic; and segment margins at HEICO and TransDigm are not comparable to consolidated margins at pure-plays. Ranking them on a single multiple would manufacture a false hierarchy. And private valuations — the marks at which SpaceX and the venture-backed launch companies raise capital — are structurally different evidence from public prices: they are negotiated, illiquid, often preference-laden, and cannot be used as a comparable for a listed equity without adjustment nobody outside the transaction can make.

↑ Back to Top

9. The Decisive Scenarios and Future Game Changers: Chinese Methane, Kessler Cascades, and Laser Links

Somewhere above every reader of this article, roughly 10,000 working satellites and tens of thousands of tracked debris fragments are moving at about 28,000 kilometres an hour. At those speeds, a fragment the size of a marble carries the kinetic energy of a car crash. Operators receive conjunction warnings routinely and manoeuvre to avoid close approaches, spending propellant — which is to say, spending satellite lifetime — each time they do.² This is the physical environment in which every scenario for this industry has to be evaluated.

Exhibit 5 — Three scenarios for the global space economy to 2030

Units and definitions as labelled; geography global; horizon to 2030; status is scenario projection, not observed data, and each column describes a different causal path rather than a percentage adjustment to a single forecast.

Read that aloud and notice what separates the columns. It is not optimism. The bear case is a physics and engineering case: orbital propellant transfer — moving cryogenic fuel between two spacecraft in freefall, which has never been demonstrated at operational scale — turns out to be the gating item for Starship's full capability, and the promised step change to sub-$300 per kilogram simply does not arrive this decade. The base case is an industrial case: reusability works, cadence rises steadily, and the constraint becomes factory throughput and launch site capacity. The bull case is an operational case: turnaround times fall to a week, which converts rockets into something closer to aircraft and makes mass to orbit effectively unconstrained. Each implies different winners. In the bear world, scarcity keeps launch pricing high and the component suppliers keep their margins but sell into a smaller volume. In the bull world, launch is nearly free, satellite mass constraints relax, and value moves decisively to whoever owns spectrum and customers.

Three developments could move the industry between those columns, and each deserves treatment beyond a headline.

Chinese commercial methane reusability. The mechanism is straightforward: if LandSpace's Zhuque-3 and Space Pioneer's Tianlong-3 achieve routine booster recovery and rapid reflight, China gains a second high-cadence reusable launch capability at a domestic cost base, and the marginal price of launch for any customer willing to fly Chinese falls sharply.² The adoption hurdles are equally clear. Recovery has been attempted, not proven in routine operation — the Zhuque-3 first stage suffered an anomaly at touchdown on the orbital debut. Export access is constrained by US technology transfer rules that make it effectively impossible to launch a satellite containing American components on a Chinese vehicle, which excludes most Western commercial payloads. The realistic effect is therefore regional rather than global in the near term: cheaper deployment of Chinese national constellations, faster build-out of Qianfan and Guowang, and pressure on non-aligned markets in Asia, Africa and Latin America where spacecraft can be sourced without US content. The observable milestones are a recovered booster reflown, and a published turnaround interval. The losers, if it happens, are European and Japanese launchers competing for third-country institutional payloads.

A debris cascade. The Kessler scenario describes a collision generating debris that causes further collisions, progressively rendering an orbital shell unusable. The mechanism that would matter commercially arrives long before any true cascade: a single major collision, or a deliberate anti-satellite test, in the 500–600 kilometre band where the megaconstellations live would generate thousands of fragments too small to track reliably. The immediate consequences would be regulatory rather than physical — launch approvals suspended pending assessment, insurance markets repricing or withdrawing, and mandatory avoidance manoeuvring that shortens satellite lives and therefore accelerates the replacement treadmill described earlier. The beneficiaries would be space situational awareness providers and, in a more speculative way, the in-space servicing companies, because a debris crisis is the one event that converts Astroscale's demonstrated capability into a mandated market.¹⁹ The losers would be every constellation operator simultaneously, which is precisely why this risk cannot be diversified away inside the theme: it is a common factor sitting underneath every name in the sector.

Optical inter-satellite link standardisation. Today, most satellite traffic must come down to a ground station and go back up. Laser cross-links let satellites pass data to one another directly at multi-gigabit rates, which removes the need for ground stations in inconvenient or politically difficult geographies and makes a constellation a mesh network in its own right. The Space Development Agency has pushed hard for a common optical standard so that satellites from different vendors can interoperate.⁶ If that standardisation takes hold commercially as well as in defence, two things follow. Value migrates toward the suppliers of optical terminals and the integrators building the mesh — Mynaric among the specialist terminal developers, with defence services firms such as CACI and SAIC positioned around integration — and away from ground segment infrastructure. And the competitive advantage of owning a global ground station network, which currently favours the largest operators, erodes. The adoption hurdle is manufacturing: laser terminals must be produced in thousands per year at falling cost with flight-proven reliability, and production scale-up has been the industry's persistent difficulty rather than the optics themselves. The milestone to watch is a published, multi-vendor interoperability demonstration in orbit followed by volume orders.

The rotation logic these three developments imply is worth stating explicitly, without turning it into a trade. If launch hardware pricing deflates — whether from Starship, from Chinese methane vehicles, or from both — the value that leaves the launch layer does not vanish. It reappears downstream, in the services that cheap access enables and in the specialised hardware that constellations consume in volume. Investors positioned in the transport layer at the moment transport becomes abundant would be holding the one asset whose economics the abundance destroys.

↑ Back to Top

10. The Crux KPIs and Falsification Dashboard: Watching the Binding Constraints

Everything above compresses into a small number of things worth actually tracking. The test of a useful indicator is not whether it is interesting but whether it moves before revenue, margin and share do, and whether it sits on a constraint that genuinely binds. Four qualify.

One: global reusable launch cadence. Measured as successful flights per year by reusable-class vehicles — Falcon 9 and Falcon Heavy, Starship, New Glenn and, prospectively, Neutron. The latest reading is a run-rate above 175 flights a year as at mid-2026.¹² This is the most upstream observable in the entire industry, because cadence determines available mass, and available mass determines whether constellation deployment schedules are feasible at all. It leads rather than lags because a launch either happens or does not on a date anyone can see, months before the resulting satellites generate revenue. In the indicator hierarchy it is structural. It discriminates the bear and base cases directly: sustained cadence above roughly 220 flights a year supports the industrial scaling path; a fall below about 140 signals the bottleneck world in which deployment schedules slip across the sector at once. Source: launch catalogues and the orbital reporting maintained by Jonathan's Space Report, updated continuously.² The theme-level kill condition is a year-over-year decline in reusable flights, which would indicate that the cost curve has stopped compounding.

Two: Starship's progress from test article to operational payload delivery, and specifically orbital propellant transfer. As at May 2026, Starship had reached Flight 12 with the upgraded engine configuration flying.¹ The mechanism this measures is the step change from roughly $1,500 per kilogram to something below $500. Orbital propellant transfer is the specific gate, because Starship's full capability for high-energy missions depends on refuelling in orbit, a manoeuvre never performed at operational scale with cryogenic propellants. It leads by years: a successful transfer demonstration would precede any commercial pricing change by a long interval, and a failed programme would show up first as schedule slippage in the lunar lander programme that depends on it. Hierarchy: structural. It discriminates the bull case from the base case. Source: SpaceX flight disclosures and NASA programme milestones, on an event-driven cadence.¹ Break condition: propellant transfer qualification fails or slips past 2027, which would push the sub-$300 per kilogram world out beyond this decade and, with it, the bull scenario's entire premise.

Three: operational direct-to-cell satellites and the regulatory permissions attached to them. Measured as the count of commercially operational direct-to-cell spacecraft — AST SpaceMobile's BlueBirds and SpaceX's direct-to-cell Starlinks — together with the status of power flux density waivers before the Federal Communications Commission.⁷⁸ Current reading: early commercial batches operating, with the regulatory framework still being settled. The causal mechanism is that satellite count and permitted transmit power jointly determine whether the service is a novelty for emergencies or a product carriers can sell, and the regulatory decision arrives before any subscriber revenue does. Hierarchy: adoption. It discriminates the single largest disagreement in the sector — whether direct-to-cell is a feature or an industry. Source: FCC filings, which are public and continuous, and company deployment announcements.⁷ Break condition at the security level: denial of the operational power waivers, or a tier-one carrier terminating its agreement. Note the honest limitation: satellite count is a proxy for capacity, and capacity per satellite is not consistently disclosed, so the indicator tracks direction rather than delivered bandwidth.

Four: upstream component gross margins. Measured as reported gross margin in Rocket Lab's Space Systems segment and in HEICO's Electronic Technologies Group.⁵¹¹ Current readings are roughly 30% and above 45% respectively. This is the single cleanest test of the second half of the upstream belief — the claim that value migrates to sole-source components. If constellation operators are squeezing their suppliers, it appears here first, in quarterly gross margin, long before it appears in supplier revenue, because volume can mask price erosion for several quarters. Hierarchy: industry and company. It discriminates the bull scenario's 50–60% upstream margin world from the bear scenario's 15–25% commoditised world. Source: quarterly filings on SEC EDGAR, every three months.⁴ Break condition: Rocket Lab's Space Systems gross margin falling below 20%, which would mean the bottleneck was never a bottleneck.

Around those four sit the broader dashboard levels. Structurally, watch launch pricing: a sustained increase in Falcon 9 pricing above 25% over two consecutive years would signal that scarcity, rather than abundance, is the operative condition. At the industry level, watch annual mass to orbit; a contraction greater than 15% year over year would falsify the structural claim outright. At the company level, watch the Neutron schedule and AST's deployment rate against stated plans, comparing what management says now with what it said at the last inflection — a discipline that matters more in this industry than in most, because launch programmes have a long history of confident dates followed by quiet revisions.⁵⁸ At the market level, watch whether equity capital markets remain open to space issuers at all; several of these businesses require external financing as an operating input, and a closed window is an existential event rather than a valuation event.

Separate the two kinds of kill criteria carefully. The theme dies if mass to orbit contracts, if launch prices rise durably, or if a debris event forces regulatory restriction. The security thesis at an individual company dies for company-specific reasons — a failed maiden flight, a denied waiver, a lost anchor customer, a covenant breach — and those failures can occur while the theme itself is flourishing. Conflating the two is the most common analytical error in thematic investing, and this industry punishes it severely.

Then there are the common factors, which matter more here than the individual names. Nearly every pure-play in this sector is a long-duration asset whose value sits in terminal cash flows, which makes the group a concentrated bet on real interest rates regardless of operational execution. Nearly every non-vertically-integrated operator depends on SpaceX for access to orbit, so a Falcon 9 grounding after an anomaly would simultaneously delay deployment schedules across the entire listed universe — a single point of failure shared by companies that look, on a screen, entirely unrelated.¹ The defence-adjacent names share exposure to a single appropriations process, where a continuing resolution delays contract awards across primes and subcontractors at once.⁶ A portfolio built from a dozen space tickers can look diversified and behave as one position.

So: is the upstream belief holding, strengthening, or fraying?

The first half is holding and strengthening. The physical evidence is unambiguous — cost per kilogram down roughly 90%, cadence up more than six-fold at the leading provider, mass to orbit up several-fold, and buyers from outside the industry, from Apple to the Pentagon to three national governments, reallocating serious capital toward orbital infrastructure.¹²⁶⁹ Nothing in the observable record suggests a reversal.

The second half — that durable equity value accrues only to distribution moats, spectrum, and sole-source bottlenecks — is holding, but with a warning that has grown louder rather than quieter. It is holding in the sense that the capital cycle did exactly what the belief predicted: the transport layer commoditised at its lower end, the small launch cohort was destroyed, and the companies that endured either owned components, owned spectrum, or owned the customer. It is fraying in one specific and important way. The most valuable position created by this entire transformation — vertically integrated launch plus constellation plus customer, compounding on its own flywheel — is not available to public shareholders at any price. Meanwhile, a large and growing share of global orbital infrastructure is being built by states that are not solving for return on capital at all.

That is the sober conclusion an investor should carry out of this story. The revolution is real, the physics has changed permanently, and the mass keeps going up. Whether shareholders get paid depends on a much narrower question: whether the specific layer you own sits on a bottleneck that the abundance does not dissolve. Cheap access to orbit destroys scarcity, and scarcity is what most of these companies were selling.

↑ Back to Top

Glossary

ARPU (average revenue per user). Recurring revenue divided by active subscribers. In satellite broadband it is the number that must, over a fleet's life, exceed the per-subscriber share of continuous satellite replacement cost for the business to create value.

Active debris removal (ADR). Missions that rendezvous with, capture and deorbit defunct satellites or spent rocket stages. Technically demonstrated; commercially dependent on a regulatory obligation to pay for cleanup that does not yet exist in most jurisdictions.

Cadence. Flights per unit time by a launch provider. The variable that determines how far fixed costs — pads, range, engineering staff — are spread, and therefore the variable that converts reusability into low cost per kilogram.

Direct-to-cell (D2C) / non-terrestrial network (NTN). Satellite service that connects to an ordinary, unmodified mobile handset using standard cellular protocols, rather than to specialised satellite hardware. The difficulty is closing a radio link with a device whose antenna and transmit power were designed for a tower a few kilometres away.

GEO (geostationary Earth orbit). A circular orbit 35,786 km above the equator where a satellite's orbital period matches Earth's rotation, so it appears fixed in the sky. One satellite covers roughly a third of the planet; round-trip latency is around half a second.

Kessler syndrome. A self-sustaining cascade in which collision debris causes further collisions, progressively making an orbital shell unusable. The commercially relevant risk arrives well before a true cascade, in the form of regulatory restriction and insurance repricing after a single major collision.

LEO (low Earth orbit). Altitudes roughly 160–2,000 km. Latency of tens of milliseconds, but each satellite covers a small area and experiences atmospheric drag, so global service requires large constellations that must be continuously replaced.

Optical inter-satellite link (OISL). Laser communication between satellites, allowing a constellation to route traffic in orbit rather than through ground stations. Reduces dependence on ground infrastructure in remote or politically restricted geographies.

Power flux density (PFD) limits. Regulatory caps on how much radio energy a satellite may deliver to the Earth's surface in a given band, set to protect other spectrum users. For direct-to-cell operators these limits determine whether the service can carry data or only messages.

PWSA (Proliferated Warfighter Space Architecture). The US Space Development Agency's layered constellation for missile warning, missile tracking and tactical data transport, procured in two-year tranches — a procurement model designed to buy many cheap satellites quickly rather than a few expensive ones slowly.

Rocket equation. Tsiolkovsky's relation showing that velocity gain scales with the logarithm of the mass ratio between a fuelled and empty vehicle. The reason an orbital rocket is roughly 95% propellant and structure, and why every kilogram of landing hardware comes directly out of payload.

SAR (synthetic aperture radar). Radar imaging that synthesises a large antenna from a moving platform, producing high-resolution images through cloud, smoke and darkness — the reason SAR constellations command defence and maritime demand that optical imagery cannot serve.

Triple-junction (TJ) solar cell. A space-grade photovoltaic cell stacking compound semiconductor layers tuned to different parts of the solar spectrum, delivering roughly double terrestrial silicon efficiency and surviving orbital radiation. Produced by very few qualified suppliers worldwide.