SpaceX

Space Exploration Technologies Corp. (SpaceX)
Type	Private
Industry	Aerospace; Satellite communications; Artificial intelligence
Founded	6 March 2002
Founder	Elon Musk
Headquarters	Starbase, Texas, United States
Key people	Elon Musk (Founder, chief executive officer and chief technology officer); Gwynne Shotwell (President and chief operating officer)
Products	Falcon 9, Falcon Heavy, Starship, Dragon, Starlink, Starshield
Employees	Approximately 15,000
Subsidiaries	xAI
Website	spacex.com

Space Exploration Technologies Corp., trading as SpaceX, is an American aerospace company headquartered at Starbase, Texas. It is the largest launch provider in the world by a very wide margin, operates the largest satellite constellation ever built, and is among the most valuable private companies in existence.

The company was founded on 6 March 2002 by Elon Musk with the stated objectives of reducing the cost of access to space, improving its reliability and eventually establishing a self-sustaining human settlement on Mars. Musk has described the last as the company's purpose and the rest as the means.

SpaceX nearly failed. Its Falcon 1 rocket failed on its first three flights between 2006 and 2008, exhausting the funding Musk had provided from the sale of PayPal, and the fourth flight in September 2008 succeeded with the company effectively out of money. A NASA contract for International Space Station resupply awarded in December 2008 saved it.

Since then it has achieved a sequence of firsts: the first privately developed liquid-fuelled rocket to reach orbit, in 2008; the first commercial spacecraft to visit the International Space Station, in 2012; the first propulsive landing and reuse of an orbital-class booster, in 2015 and 2017; the first private company to carry humans to orbit, in 2020; and the first catch of a returning booster by the launch tower, in October 2024.

Its Starlink constellation, of which the first operational satellites came online in 2019, provides broadband from low Earth orbit and now comprises many thousands of satellites, producing the majority of the company's revenue. A military variant, Starshield, serves United States government customers.

In February 2026 SpaceX acquired xAI, the artificial intelligence company Musk founded in 2023, in an all-stock transaction making it a wholly owned subsidiary, an arrangement that combines a launch and satellite company with a frontier model developer and a very large data centre operation.

History

Founding and Falcon 1, 2002 to 2008

Elon Musk had sold his interest in PayPal in 2002 and considered a project he called Mars Oasis, which would have placed a small greenhouse on Mars to grow plants and to generate public interest in the planet.

He travelled to Russia in 2001 and 2002 to buy refurbished intercontinental ballistic missiles as launch vehicles, was quoted prices he considered extortionate, and concluded on the return flight that the constraint was not the availability of rockets but their cost, and that the cost was not determined by physics. He calculated that the raw materials in a rocket amount to a small percentage of its price, and formed the view that a company building vertically and iterating quickly could reduce the price by an order of magnitude.

SpaceX was incorporated on 6 March 2002. Musk committed approximately US$100 million of his own money and recruited from established aerospace companies, including Tom Mueller from TRW, who led propulsion, and Gwynne Shotwell from Microcosm, who joined as vice president of business development in 2002 and became president and chief operating officer in 2008.

The Falcon 1 was a small two-stage rocket capable of placing a few hundred kilograms in orbit. Its first flight on 24 March 2006 from Omelek Island in the Marshall Islands failed 33 seconds after launch when a corroded aluminium nut caused a fuel leak and fire. The second, in March 2007, failed during second-stage burn. The third, in August 2008, failed when the first and second stages collided after separation because residual thrust in the new Merlin 1C engine pushed the first stage into the second.

Musk had funded three launches and had committed to a fourth. By his account the company had money for one more attempt, and Tesla was simultaneously close to failure. The fourth Falcon 1 flew successfully on 28 September 2008, making SpaceX the first privately developed liquid-fuelled rocket to reach orbit.

On 23 December 2008 NASA awarded SpaceX a Commercial Resupply Services contract worth approximately US$1.6 billion for twelve missions to the International Space Station. Musk has said he was told of the award by telephone and could not immediately speak.

Falcon 9 and reuse, 2010 to 2020

The Falcon 9 first flew on 4 June 2010. It is a two-stage rocket with nine Merlin engines in the first stage, an arrangement chosen for engine-out capability and for the manufacturing economics of producing many identical engines rather than a few large ones.

Dragon became the first commercial spacecraft to be recovered from orbit in December 2010 and the first to berth with the International Space Station in May 2012.

The decisive programme was reuse. Every orbital rocket before Falcon 9 was expended, which is equivalent to discarding an aircraft after each flight. The engineering problem is that a returning first stage arrives at several thousand kilometres an hour and must decelerate, orient itself, re-enter, and land precisely, using propellant that would otherwise have carried payload.

SpaceX's approach was propulsive landing: restart the engines, use grid fins for aerodynamic control, and land vertically on legs. The Grasshopper test vehicle demonstrated the manoeuvres from low altitude from 2012. Several attempts to land on a floating platform failed publicly and were published by the company, which released compilations of the failures, an unusual practice that both generated goodwill and normalised the iterative approach.

The first successful landing of an orbital-class booster occurred on 21 December 2015 at Cape Canaveral, and the first landing on a droneship in April 2016. The first reflight of a recovered booster took place on 30 March 2017.

Reuse transformed the economics. A Falcon 9 first stage has now flown more than twenty times in some cases, turnaround has fallen to weeks, and the marginal cost of a launch is the propellant, the second stage, the refurbishment and the operations rather than the whole vehicle. SpaceX's published price of approximately US$67 million for a Falcon 9 launch is below what competitors charge and far below their cost, and the company's internal cost is substantially lower still.

Falcon Heavy, three Falcon 9 first stages combined, first flew on 6 February 2018 carrying Musk's own Tesla Roadster on a heliocentric trajectory, a demonstration payload chosen because the flight was considered likely to fail.

Crew Dragon carried Bob Behnken and Doug Hurley to the International Space Station on 30 May 2020, the first crewed orbital launch from American soil since the Space Shuttle's retirement in 2011 and the first ever conducted by a private company. The Commercial Crew programme had been awarded to SpaceX and Boeing in 2014, with Boeing receiving approximately US$4.2 billion and SpaceX approximately US$2.6 billion; SpaceX flew four years before Boeing's Starliner carried crew, and Starliner's 2024 flight left its astronauts on the station until a Crew Dragon returned them.

Starlink, 2015 to 2026

Starlink was announced in January 2015 and is the largest satellite constellation ever built.

The proposition is that broadband delivered from low Earth orbit, at around 550 kilometres rather than the 35,786 kilometres of geostationary orbit, has latency low enough for ordinary internet use, around 20 to 40 milliseconds rather than 600, and can reach anywhere on the planet without terrestrial infrastructure.

The difficulty is that low orbits cover a small area each, so that continuous service requires thousands of satellites, which requires launching them very cheaply, which is why only a company that owns a reusable rocket could attempt it. Previous attempts, Iridium and Teledesic in the 1990s, failed on exactly this.

The first operational satellites came online in 2019. The constellation now comprises many thousands of satellites, with regulatory approval for tens of thousands more, and serves several million subscribers across more than a hundred countries. It is used by consumers in rural areas, by aircraft including several major airlines, by ships, by militaries and by humanitarian operations.

Starlink generates the majority of SpaceX's revenue and is the reason the company's valuation reflects a communications business rather than a launch business. SpaceX launches Starlink satellites on its own rockets, which means the marginal cost of deployment is its internal launch cost rather than a market price, and which is an advantage no competitor has.

Starshield, announced in 2022, is the military variant, providing Earth observation, communications and hosted payloads to United States government customers under classified and unclassified contracts, including a very large National Reconnaissance Office programme.

Direct to cell service, allowing ordinary mobile telephones to connect to satellites without special equipment, began with text messaging in 2024 through arrangements with T-Mobile US and other operators. In September 2025 SpaceX agreed to acquire rights to use certain EchoStar spectrum for approximately US$17 billion in cash and stock, to be used as the foundation of the direct-to-cell business.

Starship, 2019 to 2026

Starship is the vehicle intended to make Mars settlement possible and is the largest and most powerful rocket ever built.

It comprises the Super Heavy booster, with 33 Raptor engines burning methane and liquid oxygen, and the Starship upper stage, and both are intended to be fully and rapidly reusable. Fully assembled it stands approximately 120 metres and produces roughly twice the thrust of the Saturn V.

Methane was chosen rather than kerosene or hydrogen because it can in principle be manufactured on Mars from atmospheric carbon dioxide and subsurface water through the Sabatier process, which is the only way a vehicle can return.

Development has been conducted in public at Starbase in south Texas, using an iterative approach in which prototypes are built quickly, tested to destruction and revised. Several early vehicles exploded during landing attempts before SN15 landed successfully in May 2021.

The integrated flight tests began on 20 April 2023, when the vehicle cleared the pad, lost several engines, failed to separate and was destroyed, damaging the launch mount severely. Subsequent flights progressed: the second in November 2023 achieved stage separation; the third in March 2024 reached space and was lost on re-entry; the fourth in June 2024 achieved a controlled splashdown of both stages; and the fifth, on 13 October 2024, caught the returning Super Heavy booster with the launch tower's arms, a manoeuvre widely regarded as the most impressive achieved in the programme.

Difficulties followed. The seventh flight, on 16 January 2025, carried the first Block 2 ship and a Starlink simulator payload, and the ship was lost. The eighth, on 7 March 2025, lost contact minutes into flight and broke up, with debris visible over Florida and the Caribbean and commercial air traffic diverted. A ship was destroyed in a major anomaly during a static fire test at Starbase on 18 June 2025, prompting local emergency response.

The twelfth flight, on 22 May 2026, was the first of the Version 3 vehicle with Raptor 3 engines and the first to carry a payload to orbit, deploying Starlink satellites. The thirteenth, on 24 July 2026, was the second Version 3 flight; during a suborbital profile the ship deployed 20 Starlink V3 satellites and survived atmospheric re-entry.

Starship has been selected by NASA as the Human Landing System for the Artemis programme, under contracts awarded in 2021 and extended, which makes the American return to the lunar surface dependent on it. The programme requires orbital refuelling, transferring propellant between two Starships in orbit, which has never been demonstrated at scale and which is the principal technical risk remaining.

xAI and recent corporate developments

By July 2025 SpaceX had agreed, as part of a US$5 billion equity raise, to invest US$2 billion in xAI, the artificial intelligence company Musk founded in 2023.

On 2 February 2026 SpaceX announced that it had acquired xAI outright in an all-stock transaction, making it a wholly owned subsidiary. The combination places a frontier model developer, the Colossus data centres in Memphis, a launch business and the largest satellite constellation in one company.

In May 2026 Anthropic signed a contract reported at US$1.25 billion a month with xAI to purchase the entire compute capacity of the Colossus 1 data centre in Memphis, and a June securities filing indicated that Google and others had entered comparable arrangements, making xAI a substantial supplier of computing capacity as well as a developer of models.

SpaceX has also acquired the coding tool Cursor, which led OpenAI to terminate Cursor's access to its interfaces in August 2026.

The company's valuation has risen through successive tender offers to figures above US$350 billion, making it among the most valuable private companies ever, and speculation about a Starlink listing has recurred without materialising.

Vehicles and spacecraft

Falcon 1 (2006 to 2009) was a small two-stage rocket, flown five times with two successes, and retired once Falcon 9 was available.

Falcon 9 is the workhorse. In its current Block 5 configuration the first stage is certified for many flights with minimal refurbishment, and individual boosters have exceeded twenty flights. It places approximately 22 tonnes in low Earth orbit when expended and rather less when the booster is recovered. It has flown more than 400 times with one in-flight failure, in June 2015, and one pad explosion, in September 2016, and has become by a very large margin the most-flown and most reliable orbital rocket in operation.

Falcon Heavy combines three Falcon 9 first stages and places approximately 64 tonnes in low Earth orbit when expended. It flies rarely, because Falcon 9's capability covers most requirements, and is used for heavy geostationary payloads and for planetary missions including NASA's Psyche and Europa Clipper.

Dragon in its cargo and crew forms carries supplies and astronauts to the International Space Station. Crew Dragon seats up to seven and has flown private missions including Inspiration4 in 2021 and the Axiom flights, and Polaris Dawn in 2024, on which Jared Isaacman conducted the first commercial spacewalk using a SpaceX-developed suit.

Starship and Super Heavy are described above.

Raptor is the full-flow staged combustion methalox engine powering Starship, a cycle previously attempted and never flown, in which both the fuel-rich and oxidiser-rich preburner exhausts are fed into the main chamber so that nothing is discarded. It is among the more difficult engines ever built and its successive versions have been simplified and made cheaper as well as more powerful.

Merlin powers Falcon 9 and has the highest thrust-to-weight ratio of any production liquid rocket engine.

Business and finances

SpaceX is private and does not publish accounts, and what is known comes from disclosures in funding rounds, from reporting and from government contract records.

Revenue is generally estimated in the range of US$13 billion to US$16 billion annually, with Starlink producing the majority and launch services the remainder. The company has stated that Starlink revenue exceeded launch revenue from 2023.

Its customers divide between NASA, principally for crew and cargo transport and for the Artemis lander; the United States Space Force and intelligence agencies, for national security launch and for Starshield; commercial satellite operators; and Starlink's own consumer and enterprise subscribers.

The company launches more mass to orbit than the rest of the world combined, by a considerable margin, and a very large proportion of those launches carry its own satellites.

Its valuation has risen through successive employee tender offers, reaching figures above US$350 billion, which places it among the most valuable private companies in history. Musk holds a substantial stake with supervoting shares, and outside investors include Founders Fund, Sequoia, Fidelity, a16z and sovereign funds.

The company has repeatedly deferred a public listing. Musk has said that Starlink might be listed separately once its cash flow is predictable, and that SpaceX itself will not be listed while Mars remains the objective, on the reasoning that public shareholders would not tolerate the expenditure.

Organisation

SpaceX employs approximately 15,000 people, with principal sites at Hawthorne, California, where Falcon and Dragon are built and where mission control operates; Starbase in Cameron County, Texas, where Starship is built and launched and which was incorporated as a city in 2025; McGregor, Texas, for engine testing; Redmond, Washington, for Starlink satellite production; Cape Canaveral and Kennedy Space Center in Florida; and Vandenberg in California.

Gwynne Shotwell has been president and chief operating officer since 2008 and runs the company operationally. She joined as the seventh employee in 2002, negotiated the contracts that kept the company alive, and is generally credited within the industry with the execution that has turned Musk's objectives into a functioning business. The division of labour, in which Musk sets direction and takes engineering decisions and Shotwell runs the company, is unusually explicit.

The engineering culture is characterised by very long hours, rapid iteration, tolerance of public failure, vertical integration to a degree unusual in aerospace, where SpaceX manufactures approximately 80 per cent of its vehicles' components in house, and by decision-making concentrated in engineers rather than in programme managers.

The company's approach to failure differs from the industry's. Conventional aerospace minimises the chance of failure through analysis before flight, because each flight is expensive and a failure is a disaster. SpaceX flies early, accepts failures, and learns from them, which is possible because its vehicles are cheap enough that losing one is affordable. The approach is the principal reason its development is faster and it is the reason its public failures are more numerous.

Engines and propulsion

SpaceX designs and builds its own engines, which is unusual and which is the principal reason it controls its own costs.

Merlin

Merlin, developed under Tom Mueller from 2002, burns rocket-grade kerosene and liquid oxygen in a gas-generator cycle, the simplest and least efficient of the standard arrangements, chosen deliberately because it is cheap to build and reliable.

Its distinguishing property is thrust-to-weight ratio, the highest of any production liquid rocket engine, achieved through aggressive weight reduction and through a pintle injector, a single moving element derived from the Apollo Lunar Module descent engine, which is simpler than the hundreds of injector elements in a conventional design and which is inherently stable against combustion oscillation.

Nine Merlins power the Falcon 9 first stage. The arrangement was chosen for three reasons: the vehicle can complete its mission after losing an engine, which it has done in flight; producing many identical engines allows manufacturing learning that a small number of large engines does not; and the engines must throttle deeply for landing, which is easier with many small ones.

A vacuum-optimised Merlin with a large nozzle extension powers the second stage and can restart in flight.

Raptor

Raptor powers Starship and burns liquid methane and liquid oxygen in a full-flow staged combustion cycle.

In a gas-generator engine a small portion of the propellant is burned to drive the turbopumps and its exhaust is discarded overboard, which wastes a few per cent of the propellant. In staged combustion the preburner exhaust is fed into the main chamber, which recovers it. In full-flow staged combustion, which Raptor is the first flown engine to use, there are two preburners, one fuel-rich driving the fuel pump and one oxidiser-rich driving the oxidiser pump, and both exhausts enter the main chamber, so that all the propellant passes through the main combustion.

The advantages are efficiency, lower turbine temperatures because the entire flow passes through them, and the ability to run at very high chamber pressures. The difficulty is that an oxidiser-rich preburner produces hot oxygen at pressure, which attacks most metals, and that the plumbing is substantially more complex. The Soviet Union built oxidiser-rich staged combustion engines the West believed impossible; nobody had flown a full-flow engine before Raptor.

Methane was chosen over kerosene because it burns more cleanly, leaving no soot to foul the engine between flights, which matters for rapid reuse; because it is denser than hydrogen and therefore needs smaller tanks; and because it can be synthesised on Mars.

Successive versions have simplified the engine radically, removing shielding, integrating components into the castings and reducing the plumbing that is visible on photographs of the early versions, with each iteration cheaper and more powerful. Raptor 3, flown from May 2026, is the current version.

Dragon

Cargo Dragon delivered supplies to the International Space Station from 2012 and returns cargo to Earth, a capability that only it and, briefly, the Space Shuttle have had; every other resupply vehicle burns up on re-entry.

Crew Dragon, also called Dragon 2, carries up to seven astronauts and is the only American vehicle certified to do so. It differs from earlier spacecraft in several respects: it is largely automated and docks without manual control; its interior uses touchscreens rather than physical switches, which drew comment from astronauts trained on the Shuttle; and its SuperDraco abort engines, embedded in the capsule wall rather than mounted on a tower, can pull it clear of a failing rocket at any point in flight, a capability demonstrated in an in-flight abort test in January 2020.

Crew Dragon was originally intended to land propulsively on legs, which NASA declined to certify, so it descends under parachutes to a water landing.

Private missions have included Inspiration4 in September 2021, the first orbital flight with no professional astronauts, commanded and funded by Jared Isaacman; the Axiom missions carrying private crews to the International Space Station; and Polaris Dawn in September 2024, which reached the highest Earth orbit since Apollo and on which Isaacman and Sarah Gillis conducted the first commercial spacewalk, testing a SpaceX-developed extravehicular suit in a capsule with no airlock, requiring the entire cabin to be depressurised.

Crew Dragon also returned the Boeing Starliner crew, Butch Wilmore and Suni Williams, in March 2025 after Starliner's thruster problems led NASA to return it uncrewed, extending an eight-day mission to more than nine months.

SpaceX has been contracted to build the United States Deorbit Vehicle, a modified Dragon that will guide the International Space Station to a controlled destructive re-entry at the end of its life, currently planned for 2030 or later.

Starship technical description

Super Heavy, the first stage, is approximately 71 metres tall and 9 metres in diameter, with 33 Raptor engines producing approximately 74 meganewtons of thrust at sea level, roughly twice the Saturn V's. It is built from stainless steel.

Starship, the second stage, is approximately 50 metres tall and carries six Raptors, three optimised for sea level and three for vacuum. It is intended to serve as a launch vehicle upper stage, an orbital tanker, a lunar lander and a Mars vehicle in different configurations.

The choice of stainless steel rather than carbon fibre or aluminium-lithium was made in 2018 and reversed several years of development. Steel is heavier per unit strength at room temperature and better at both cryogenic and high temperatures, which means the tanks are stronger when cold and the structure survives re-entry heating that would destroy aluminium. It is also approximately fifty times cheaper than carbon fibre, can be welded in the open air rather than in a clean room, and is repairable. Musk has described the decision as the best he has made on the programme.

Thermal protection uses hexagonal ceramic tiles on the windward side, which have been the programme's most persistent difficulty, with tiles lost on several flights.

The bellyflop is the ship's re-entry profile: it falls broadside, using its body and four flaps as an airbrake to dissipate energy aerodynamically rather than with propellant, then rotates to vertical in the final seconds for landing. The manoeuvre was demonstrated repeatedly by prototypes from 2020 and has no precedent.

Catching by the launch tower's arms, nicknamed chopsticks, replaces landing legs. The stage's grid fins rest on the arms. This removes the legs' mass, allows the stage to be reconnected to the pad immediately, and requires precision that observers regarded as implausible until the October 2024 demonstration.

Corporate structure and ownership

SpaceX is private and has not published financial statements. Its share structure gives Elon Musk supervoting shares sufficient for control regardless of his economic stake, which was reported at roughly 42 per cent of equity and a substantially higher proportion of votes.

Outside investors have included Founders Fund, Draper Fisher Jurvetson, Valor, Sequoia, Fidelity, Baillie Gifford, a16z, Google, which invested approximately US$900 million in 2015 alongside Fidelity, and sovereign wealth funds.

Liquidity is provided through periodic tender offers in which employees and early investors sell to new ones at a price the company sets, which has risen in most rounds and which produced valuations above US$350 billion by 2025.

The February 2026 acquisition of xAI was conducted in stock, which means SpaceX shareholders now hold an interest in a frontier artificial intelligence developer, a data centre operator and a launch and satellite company, and it consolidated within one private company assets that would ordinarily be held separately.

Musk's other companies, Tesla, The Boring Company and Neuralink, are separate, though personnel, suppliers and in some cases financing have moved between them, and the arrangements have attracted governance criticism.

Starlink's business

Starlink's subscriber base has grown from nothing in 2020 to several million, with service available in more than a hundred countries.

Consumer service is sold with a user terminal at a subsidised price and a monthly fee varying by market, with regional pricing that has fallen substantially in developing markets.

Aviation agreements include United Airlines, Qatar Airways, Air France, Hawaiian and others, replacing geostationary systems whose latency made them unsatisfactory.

Maritime service covers commercial shipping, cruise lines and yachts, a market previously served at very high prices by geostationary operators.

Enterprise and government service includes remote industrial sites, disaster response and military applications.

Direct to cell connects ordinary mobile telephones to satellites for messaging and, progressively, data, through arrangements with mobile operators including T-Mobile US, and the September 2025 EchoStar spectrum acquisition for approximately US$17 billion is intended to give SpaceX its own spectrum rather than depending on partners'.

The business's principal constraints are capacity in dense areas, which requires more satellites and more spectrum; regulatory approval in individual countries, which several including India have granted slowly and which China and others have refused; and the cost of the user terminal, which has been subsidised heavily.

How reuse works

The engineering behind SpaceX's most consequential achievement is worth setting out, because the difficulty is not obvious.

The problem

A rocket's first stage separates at roughly 70 kilometres altitude travelling at between 1.5 and 2.5 kilometres a second, depending on the mission. To return it must reverse that velocity, survive re-entry heating, control its descent through the atmosphere and land precisely enough to be recovered.

The propellant required for those manoeuvres is propellant not used to accelerate the payload, so a reusable stage carries less to orbit than an expendable one, typically 30 to 40 per cent less for a return to the launch site and rather less for a droneship landing downrange. The economic question is whether the saved vehicle is worth more than the lost payload, and the answer depends entirely on how cheaply the stage can be refurbished.

The Space Shuttle is the cautionary case. It was reusable and its refurbishment, particularly of the thermal protection tiles and the main engines, cost so much and took so long that each flight cost more than an expendable rocket would have.

The SpaceX solution

Boostback and entry burns. The stage restarts three of its nine engines to reverse direction, then one or three again before re-entry to slow it sufficiently that aerodynamic heating remains survivable without a heat shield. This is the largest consumer of the reserved propellant.

Grid fins. Four titanium lattice fins deploy after separation and provide aerodynamic control during descent. They are made from a single cast piece of titanium, machined, and require no refurbishment between flights, having replaced aluminium fins that burned.

Landing burn. A single engine restarts in the last seconds and throttles to bring the stage to zero velocity at zero altitude, a manoeuvre called a suicide burn because the engine cannot hover; Merlin's minimum thrust exceeds the empty stage's weight, so the burn must be timed so that the stage arrives at rest exactly at the surface.

Landing legs deploy in the final seconds.

Droneships. Missions with insufficient margin to return to the launch site land on autonomous barges positioned several hundred kilometres downrange, named Just Read the Instructions, Of Course I Still Love You and A Shortfall of Gravitas after ships in Iain M. Banks's Culture novels.

Fairing recovery. The payload fairing, two halves costing several million dollars, descends under parafoil and is recovered from the water.

The result

Block 5 Falcon 9 boosters were designed for ten flights without major refurbishment and have exceeded twenty. Turnaround has fallen from months to as little as three weeks for a given booster and to a few days for the pad.

Starship extends the approach further: the booster is caught by the launch tower rather than landing on legs, which removes the legs' mass and allows the vehicle to be reconnected to the pad directly, and the ship is intended to be caught in the same way. The October 2024 catch demonstrated the booster half of this.

Starlink technology

Each Starlink satellite weighs a few hundred kilograms, is manufactured at Redmond, Washington, at a rate the company has described as several a day, and is flat-packed so that dozens fit in a Falcon 9 fairing without a dispenser, separating by releasing tensioned rods and drifting apart.

Phased array antennas on the satellite and on the user terminal steer beams electronically rather than mechanically, which is what allows a terminal to track satellites crossing overhead at several kilometres a second without moving.

Laser inter-satellite links, added from 2021, connect satellites directly in orbit, which removes the need for a ground station within view of both the user and the network and which is what makes service possible over oceans and in polar regions.

Krypton ion thrusters provide orbit raising, station keeping, collision avoidance and controlled deorbiting at end of life. Krypton was chosen over the conventional xenon because it is substantially cheaper, and later satellites use argon.

Autonomous collision avoidance uses United States Space Force tracking data and manoeuvres without ground intervention.

Ground stations connect the constellation to the internet backbone, and the user terminal, originally costing SpaceX considerably more than its selling price, has been redesigned repeatedly to reduce cost.

The constellation's performance has improved from around 50 megabits a second to several hundred for many users, and capacity remains the constraint in densely subscribed areas, which is the reason for the successive satellite versions and for the EchoStar spectrum acquisition.

The Mars plan

Musk's stated objective, repeated since 2002, is a self-sustaining city on Mars, which he defines as one that would survive if resupply from Earth stopped.

The architecture he has described requires several things in sequence.

Full reuse. Both stages must be reused rapidly, because the vehicle cost dominates otherwise. This has not been demonstrated.

Orbital refuelling. A Starship departing for Mars must be refuelled in Earth orbit by several tanker flights, because the propellant required to leave Earth's gravity well and then to land on Mars exceeds what one vehicle can carry. Transferring cryogenic propellant between vehicles in microgravity has never been done at scale and is the principal technical unknown.

Propellant production on Mars. Returning requires manufacturing methane and oxygen on the surface from atmospheric carbon dioxide and subsurface water ice through the Sabatier reaction, which requires substantial power, presumably solar, and equipment that has not been built.

Launch windows occur approximately every 26 months, which constrains the cadence of any campaign.

The obstacles beyond propulsion are substantial and less discussed by the company: radiation exposure during transit and on a surface without a magnetic field or thick atmosphere; the physiological effects of months in microgravity followed by years at 38 per cent of Earth's gravity, about which almost nothing is known; closed-loop life support, which has never been demonstrated for the durations required; and the economic question of who would pay for and who would populate such a settlement.

Musk's timelines have consistently slipped. He suggested uncrewed landings in 2018 at various points, then 2022, then 2026, and crewed landings correspondingly later. The company has said it intends uncrewed Starship flights to Mars in a forthcoming window.

The strongest defence of the programme is that the intermediate products are independently valuable: a fully reusable heavy-lift vehicle transforms Earth orbit economics whether or not anyone goes to Mars, and Starlink, which funds the effort, exists because Falcon 9 made it possible.

NASA and government customers

SpaceX's relationship with NASA is symbiotic and is the reason the company exists.

Commercial Orbital Transportation Services, awarded in 2006, provided milestone-based funding for cargo delivery and required SpaceX to contribute its own capital. Commercial Resupply Services, awarded in December 2008, provided the contract that saved the company.

Commercial Crew was awarded in September 2014 to SpaceX at approximately US$2.6 billion and to Boeing at approximately US$4.2 billion. SpaceX flew crew in May 2020; Boeing's Starliner flew crew in June 2024 and returned uncrewed after thruster failures, its astronauts returning on a Crew Dragon in March 2025.

Human Landing System for Artemis was awarded to SpaceX in April 2021 at approximately US$2.9 billion, in a decision protested unsuccessfully by Blue Origin, and extended for subsequent missions. A second provider, Blue Origin, was added in 2023.

Science missions launched by SpaceX include DART, which deflected an asteroid in 2022; the Double Asteroid Redirection Test; Psyche; Europa Clipper; and, from 2025, the launch of the Nancy Grace Roman Space Telescope was contracted.

National security launch is conducted under the Space Force's National Security Space Launch programme, in which SpaceX has taken a majority of assignments since certification, and Starshield contracts including a National Reconnaissance Office programme reported at approximately US$1.8 billion.

The relationship has become politically sensitive. NASA's dependence on a single provider for crew transport, for cargo and for the lunar lander has been noted by the agency's own inspector general and by the Government Accountability Office, and Musk's political role since 2024 has made the dependence a subject of congressional attention.

xAI

xAI was founded by Elon Musk in July 2023 and became a wholly owned subsidiary of SpaceX on 2 February 2026.

Origins

Musk had co-founded OpenAI in 2015 and left its board in 2018 after proposing to take control and being refused. He subsequently criticised the organisation's conversion to a for-profit structure and its relationship with Microsoft, and sued it in 2024, losing in May 2026 on limitation grounds.

xAI was established as a competitor with a stated mission of understanding the universe, and recruited researchers from DeepMind, OpenAI, Google Research and elsewhere.

Grok and Colossus

Grok, the company's model family, was released in late 2023 and integrated into X, the platform Musk had acquired in 2022. Its distinguishing features were access to posts on X as a real-time data source and a deliberately less restricted conversational style, which produced both an audience and a sequence of controversies over outputs.

xAI merged with X Corp in March 2025 in an all-stock transaction valuing xAI at approximately US$80 billion and X at approximately US$33 billion, combining the model with its principal distribution channel and data source.

Colossus, the company's data centre in Memphis, Tennessee, was built at a pace without precedent in the industry, with the first cluster of approximately 100,000 NVIDIA accelerators assembled in a converted factory in roughly four months and subsequently expanded substantially. The speed was achieved partly by installing on-site gas turbines for power while grid connection was arranged, which produced objections from local residents and from environmental organisations concerning permits and air quality in a neighbourhood already subject to industrial pollution.

Within SpaceX

The February 2026 acquisition made xAI a SpaceX subsidiary. The stated rationale combined Musk's interest in consolidating his ventures with arguments about applying machine learning to launch operations, satellite constellation management and, eventually, computing in orbit, where solar power is continuous and cooling is a vacuum problem rather than a water problem.

In May 2026 Anthropic contracted with xAI at a reported US$1.25 billion a month for the entire capacity of Colossus 1, and a June securities filing indicated that Google and others had entered similar arrangements, which made xAI a substantial supplier of computing capacity to its own competitors, an arrangement without close precedent.

The consolidation places within one private company a launch provider on which Western governments depend, the largest satellite constellation, a social media platform, a frontier model developer and a very large computing operation, all controlled by one person. The concentration has been noted by regulators, by legislators and by commentators across the political spectrum, and no mechanism currently exists for addressing it.

Regulatory and legal

Launch licensing is conducted by the Federal Aviation Administration's Office of Commercial Space Transportation, which must approve each vehicle and, for new vehicles or changes, conduct environmental review. SpaceX has criticised the pace repeatedly and the agency has been criticised for approving under political pressure; both criticisms intensified after 2024, when Musk's political role made the relationship contentious.

Spectrum and orbital slots are allocated by the Federal Communications Commission and coordinated internationally through the International Telecommunication Union. SpaceX's applications for very large numbers of satellites have been contested by competitors including Amazon, Viasat and Dish, and the FCC has reduced some authorisations and imposed conditions including deorbit requirements.

Environmental litigation concerning Starbase has been brought by the Center for Biological Diversity, the Sierra Club and others, principally over the effects of launches on the adjacent wildlife refuge, over beach closures restricting public access to a state park, and over debris from the April 2023 launch.

Employment litigation includes the National Labor Relations Board complaint over the 2022 dismissals, which SpaceX answered by challenging the board's constitutional structure; discrimination claims; and a Department of Justice action alleging that the company had discriminated against asylees and refugees in hiring, which SpaceX contested on the ground that export control law restricts who may work on rocket technology, a defence that raises a genuine conflict between two federal regimes.

Export control under the International Traffic in Arms Regulations restricts SpaceX's employment of non-citizens and its dealings with foreign entities, and the company has had to structure its international business accordingly.

The economics of launch

The change SpaceX produced is worth quantifying.

Before Falcon 9, placing a kilogram in low Earth orbit cost roughly US$10,000 to US$20,000 on an expendable vehicle and approximately US$54,000 on the Space Shuttle once the programme's total cost is divided by its payload. Falcon 9's published price implies under US$3,000, and its internal cost is lower.

The consequence is that whole categories of activity became possible. A constellation of several thousand satellites was not affordable at US$15,000 a kilogram and is affordable at US$2,000. Earth observation companies operating dozens of small satellites, which is now an industry, could not have existed. Scientific missions that would not have been funded have flown.

Starship is intended to reduce the figure by another order of magnitude, to a few hundred dollars a kilogram, which if achieved would change the calculation again: at that price, assembling large structures in orbit, manufacturing in microgravity and propellant depots become worth considering, none of which is currently sensible.

The counterargument, made by economists who study the sector, is that the demand for launch is not obviously elastic enough to absorb the capacity Starship would provide, and that Starlink exists partly because SpaceX had to create demand for its own vehicles. The company's answer is that this is precisely how the argument about aviation proceeded in the 1930s.

Reception of the Starship programme

Opinion on Starship divides more sharply than on any other part of the company.

The favourable view holds that a vehicle of that scale flying thirteen times in three years, with a tower catch demonstrated and orbital flight with payload achieved, represents development faster than any comparable programme in history, and that the failures are the method rather than a departure from it. Saturn V flew thirteen times in total across six years with an enormous budget and an accepted expectation of perfection.

The unfavourable view holds that the programme is years behind the schedules given, that Artemis now depends on capabilities that have not been demonstrated, including orbital refuelling, that the ship's thermal protection remains unreliable, and that the number of tanker flights required for a single lunar mission, variously estimated between eight and twenty, implies a cadence nobody has achieved.

Both are substantially correct. The programme has progressed faster than any large launch vehicle ever developed and it has slipped very considerably against the dates the company gave, and the two statements are compatible because the dates were never realistic.

The consequential question is Artemis. NASA selected Starship as the lunar lander, which means the American return to the Moon cannot occur until a fully reusable vehicle that has not yet flown a complete mission profile can be refuelled in orbit by a technique that has not been demonstrated and then land on the Moon and return. The agency's inspector general and the Government Accountability Office have both reported on the risk.

Competition

Blue Origin, founded by Jeff Bezos in 2000, two years before SpaceX, has progressed far more slowly. Its New Shepard suborbital vehicle has flown tourists since 2021; its orbital New Glenn first flew in January 2025 and landed its booster on its second flight in 2026. It holds an Artemis lander contract and a very large amount of capital from Bezos, and the contrast between the two companies' progress from comparable starting points is among the most examined in the industry.

United Launch Alliance, the Boeing and Lockheed Martin joint venture, held the American national security launch monopoly until SpaceX's certification in 2015. Its Vulcan Centaur, which replaced Atlas V and Delta IV, first flew in January 2024 after substantial delay, partly because its BE-4 engines come from Blue Origin. The company has been for sale for several years.

Arianespace and the European Space Agency lost the commercial market to Falcon 9 and developed Ariane 6, which first flew in July 2024 and which is not reusable, a decision taken in 2014 that has been widely criticised in Europe since.

Rocket Lab, founded in New Zealand and now American, operates the small Electron vehicle and is developing the medium-lift, partially reusable Neutron. It is the most credible American competitor at scale after Blue Origin.

China conducts a very large and growing launch programme through the state and through a number of commercial companies, several of which are developing reusable vehicles resembling Falcon 9 closely. Chinese cadence has grown rapidly and the Guowang and Qianfan constellations are direct responses to Starlink.

Russia has lost essentially all commercial launch business and its Soyuz, which carried American astronauts between 2011 and 2020 at rising prices, no longer does.

India through ISRO, Japan through Mitsubishi Heavy Industries, and a number of smaller entrants complete the field.

In satellite communications the competitors are Eutelsat OneWeb, which operates a smaller constellation in a higher orbit and which merged after financial difficulty; Amazon's Project Kuiper, funded to approximately US$10 billion and launching from 2024, which has the advantage of Amazon's cloud infrastructure and the disadvantage of buying launches, including several from SpaceX; and the Chinese constellations.

The structural asymmetry is that SpaceX launches its own satellites on its own rockets at internal cost, which no competitor can match, and which is the principal reason Kuiper's economics are difficult.

Facilities

Hawthorne, California is the original headquarters and the principal manufacturing site for Falcon and Dragon, occupying a former Boeing 747 fuselage plant.

Starbase, Texas, at Boca Chica near Brownsville, builds and launches Starship. It was incorporated as a city in May 2025 following a vote by residents, almost all of whom are SpaceX employees, which gave the company's workforce control of the local government including zoning and beach access decisions previously held by Cameron County.

McGregor, Texas tests engines, and the sound of Raptor testing is audible for many miles.

Redmond, Washington manufactures Starlink satellites and user terminals.

Cape Canaveral Space Force Station and Kennedy Space Center in Florida provide the principal launch sites, including Launch Complex 39A, from which Apollo 11 departed and which SpaceX leases, and where a Starship pad is under construction.

Vandenberg Space Force Base in California provides polar launches.

Bastrop, Texas manufactures Starlink terminals and, since 2024, has housed operations for Musk's other companies.

Colossus, in Memphis, Tennessee, is the xAI data centre acquired with that company, built at unusual speed and powered partly by on-site gas turbines, an arrangement that has drawn environmental objection from local residents and regulators concerning permits and air quality.

Notable flights and missions

28 September 2008: Falcon 1 Flight 4 reaches orbit, the first privately developed liquid-fuelled rocket to do so.
8 December 2010: Dragon becomes the first commercial spacecraft recovered from orbit.
25 May 2012: Dragon berths with the International Space Station.
21 December 2015: Falcon 9 lands its first stage at Cape Canaveral.
8 April 2016: First droneship landing.
30 March 2017: First reflight of a recovered booster.
6 February 2018: Falcon Heavy's first flight carries a Tesla Roadster.
30 May 2020: Crew Dragon Demo-2 carries astronauts, the first private crewed orbital flight.
16 September 2021: Inspiration4 flies the first all-civilian orbital crew.
24 November 2021: DART launches, later deflecting the asteroid Dimorphos.
20 April 2023: Starship's first integrated flight test.
13 October 2024: The launch tower catches a returning Super Heavy booster.
10 September 2024: Polaris Dawn conducts the first commercial spacewalk.
14 October 2024: Europa Clipper launches on Falcon Heavy.
22 May 2026: Starship Flight 12, the first V3 vehicle, reaches orbit with a payload.
24 July 2026: Flight 13 deploys 20 Starlink V3 satellites and survives re-entry.

Timeline

2002 Founded on 6 March; Gwynne Shotwell joins.
2006 Falcon 1's first flight fails; NASA awards COTS funding.
2008 The fourth Falcon 1 reaches orbit in September; NASA awards the resupply contract in December.
2010 Falcon 9's first flight in June; Dragon is recovered in December.
2012 Dragon reaches the International Space Station in May.
2014 Commercial Crew is awarded; SpaceX sues the Air Force over launch competition.
2015 Starlink is announced in January; the first booster landing occurs in December.
2017 The first booster reflight occurs in March.
2018 Falcon Heavy flies in February.
2019 The first operational Starlink satellites come online.
2020 Crew Dragon carries astronauts in May; Starship prototypes begin high-altitude tests.
2021 SN15 lands successfully in May; the Artemis lander contract is awarded in April; Inspiration4 flies in September.
2023 Starship's first integrated flight test in April; Starlink revenue exceeds launch revenue.
2024 The booster is caught by the tower in October; Polaris Dawn conducts the first commercial spacewalk in September.
2025 Starbase is incorporated as a city in May; a ship is destroyed in a static fire anomaly in June; SpaceX agrees to invest US$2 billion in xAI in July and to acquire EchoStar spectrum for approximately US$17 billion in September.
2026 xAI is acquired outright on 2 February; Anthropic contracts for Colossus capacity in May; Starship Flight 12 reaches orbit with payload in May and Flight 13 deploys satellites in July.

Reception

SpaceX's standing divides along lines that have shifted substantially since 2024.

Within the aerospace industry its achievements are not disputed, and the argument concerns whether they are replicable by others or depend on conditions, including a founder willing to spend his own fortune and a tolerance of failure that public companies lack, that cannot be reproduced.

Among scientists the assessment is mixed. Planetary scientists and astronomers have benefited enormously from cheap launch, which has made missions possible that were not, and optical astronomers have been harmed by the constellation.

Among policymakers the dependence on a single company has moved from a theoretical concern to a practical one, and the European Union's IRIS² constellation, Britain's and India's programmes and the Chinese constellations are each responses.

Among the public the company was among the more admired institutions in the United States for most of its history, and polling since 2024 has shown a sharp partisan division that tracks attitudes to Musk rather than to the company.

The most durable judgement is probably that SpaceX demonstrated something about institutions rather than about rockets. The physics of reusable launch was understood in 1970. What had not existed was an organisation willing to try it, fail publicly and repeatedly, and continue, and the reason that organisation did not exist is that the industry's structure, cost-plus contracting and risk aversion by customers whose payloads were irreplaceable, made it impossible. SpaceX's most consequential act was to build one.

Effects on the industry

SpaceX has changed its industry more comprehensively than any company has changed any industry in recent decades, and the specific changes are worth listing.

Price. The cost of placing a kilogram in low Earth orbit fell by roughly an order of magnitude. The Space Shuttle cost approximately US$54,000 a kilogram; Falcon 9 costs under US$3,000 and Starship is intended to be far lower. Activities that were not economically possible, including large constellations, became possible.

Reuse. Every serious launch vehicle now under development is intended to be at least partly reusable, including Blue Origin's New Glenn, Rocket Lab's Neutron, China's several programmes and Europe's studies. The industry regarded reuse as impractical until SpaceX demonstrated otherwise, having concluded from the Shuttle's experience that refurbishment costs more than manufacture.

Competitive structure. United Launch Alliance, the Boeing and Lockheed Martin joint venture that held a monopoly on American national security launch, lost it after SpaceX sued the Air Force in 2014 over sole-source awards and was certified in 2015. Arianespace lost commercial share and developed Ariane 6 in response, and Russia's Proton lost the commercial market entirely.

Cadence. SpaceX conducts more than a hundred launches a year, which is more than the rest of the world combined and more than any nation has achieved historically.

Procurement. NASA's Commercial Orbital Transportation Services and Commercial Crew programmes, which paid for results rather than for costs incurred and which required the contractor to contribute capital, were designed partly around SpaceX and have become the model for a considerable part of American civil and military space acquisition.

Venture investment. The demonstration that a private company could build a rocket and be paid to fly it produced a wave of investment in space startups from about 2015, of which Rocket Lab, Relativity, Firefly, Astra and many others are products, and most of which would not have been funded without SpaceX's example.

Criticism and controversy

Dependence. A very large proportion of the Western world's access to orbit now depends on one company controlled by one person. NASA, the Space Force, commercial operators and several allied governments have no comparable alternative, and the American government has repeatedly noted the risk in its own assessments. The situation is the direct consequence of SpaceX's success and has no easy remedy.

Musk's conduct. Musk's political activity, his statements on social media, his role in the second Trump administration's Department of Government Efficiency during 2025 and his public disputes have repeatedly created difficulty for a company with government customers. His suggestion in 2025, during a dispute with the president, that Dragon might be decommissioned, which he retracted within hours, illustrated the exposure.

Starlink and astronomy. The constellation's satellites are visible from the ground and appear in astronomical images as streaks. The International Astronomical Union and observatories including Vera C. Rubin have raised the effect on optical and radio astronomy. SpaceX has responded with darkening coatings, sunshades and orientation changes during orbit raising, which have reduced brightness substantially without eliminating the problem, and has cooperated with astronomers more than was expected. The wider issue is that no regulatory mechanism exists to weigh the value of a constellation against the value of the night sky.

Orbital debris. Several tens of thousands of satellites in low orbit raise the probability of collision and, in the worst case, of a cascade. SpaceX's satellites operate at altitudes low enough that a failed satellite re-enters within a few years, they carry automated collision avoidance, and the company reports very large numbers of avoidance manoeuvres annually, a figure that is itself an indication of how congested the region has become.

Geopolitics. Starlink's role in Ukraine, where it has been the principal communications system for Ukrainian forces, gave one company decisive influence over a conflict. Musk's decision not to enable service near Crimea in 2022, disclosed in his biography, prevented a planned Ukrainian operation, and produced a substantial argument about whether a private individual should hold such a power. The United States government subsequently contracted for Starshield capacity, which places the service under government control.

Environment and Starbase. The Boca Chica site is adjacent to the Lower Rio Grande Valley National Wildlife Refuge and to Boca Chica State Park. Launches and the April 2023 pad destruction have produced debris, road closures restricting public beach access, and litigation from environmental groups. The Fish and Wildlife Service and the Federal Aviation Administration have conducted assessments, and the incorporation of Starbase as a city in 2025 gave the company's employees control of the local government, which critics describe as a company town.

Labour. Employees have alleged unsafe conditions, and a Reuters investigation in 2023 documented a rate of injuries above industry averages at several sites. Eight employees were dismissed in 2022 after circulating a letter criticising Musk's public conduct; the National Labor Relations Board issued a complaint, and SpaceX responded by challenging the constitutionality of the board itself, litigation that has proceeded alongside similar challenges by other companies.

Regulation. SpaceX has repeatedly criticised the Federal Aviation Administration's licensing pace and the environmental review process, and the agency has been criticised in turn for insufficient oversight. The relationship became politically charged after 2024.

Assessment

SpaceX has accomplished what a very large number of well-informed people said was impossible, and the specific thing it accomplished was not technical.

Reusable rockets had been proposed since the 1960s and the physics was understood. What had not been done was to build an organisation willing to lose several vehicles in public, to iterate in weeks rather than years, to manufacture its own components rather than to contract them, and to fund the attempt from a business that paid for itself along the way. The decisive innovations were organisational and financial, and the engineering followed.

The results are not seriously disputed. Launch costs have fallen by an order of magnitude. The United States regained the ability to carry its own astronauts, four years after a competitor paid substantially more failed to do so. A constellation that previous attempts bankrupted has been built and works. And a rocket larger than the Saturn V has flown thirteen times in three years, which no previous vehicle of that class approached.

Three questions remain open.

The first is whether Starship works as intended. Rapid full reuse, orbital refuelling and the cadence required for Mars have each been described and none demonstrated at scale, and the programme's schedule has slipped by years. Artemis depends on it.

The second is concentration. A single company with a single controlling shareholder now holds a position in launch, in satellite communications, in military space and, since February 2026, in artificial intelligence, that no private entity has held in a strategic sector in modern times. The dependence is a problem for the governments that rely on it and there is no obvious means of reducing it that does not involve funding a competitor.

The third is Mars. Musk has described settlement as the purpose and has given timelines that have consistently slipped. Whether it happens is a question about cost, about life support, about radiation, about human physiology in low gravity and about whether anyone will pay, and none of those is answered by having built a large rocket. What the company has done is make the question a matter of engineering and economics rather than of science fiction, which is itself a substantial achievement.