Everything the company flies today, why each piece exists, and the vocabulary you need to follow a launch thread without looking anything up. Read top to bottom, or jump to what you need.
SpaceX is one flywheel with four teeth. Land and reuse the rocket, so launch gets cheap. Cheap launch makes it affordable to fly your own hardware — thousands of satellites nobody else could justify putting up. That constellation sells internet to consumers and governments, which throws off the cash. The cash pays for a much larger, fully reusable rocket, which makes the constellation ten times bigger and, in the stated plan, puts people on Mars.
So when you read about a Falcon 9 booster flying for the 37th time, that is not a stunt statistic. It is the input to everything else: roughly two-thirds of SpaceX launches carry SpaceX's own satellites, and the marginal cost of those flights is what makes the constellation's economics work at all.
Counts move weekly. Starlink figures are 7 Sep 2026; Falcon totals run through 30 Aug 2026.
Rides to orbit (NASA, the US Space Force, commercial satellite operators), astronaut transport to the ISS, private crewed missions, and — by far the biggest line — Starlink subscriptions plus government variants of the same hardware.
Starship. The vehicle has no paying customers of consequence yet, consumes an enormous fab-and-test operation across Texas and Florida, and carries two commitments: launching the next-generation Starlink satellites, and landing NASA astronauts on the Moon.
Four vehicles in company history, but only three matter today: Falcon 9, its three-core sibling Falcon Heavy, and Starship. Everything is drawn to the same scale below.
Journalists write "Falcon 9" as if it were a fixed object. It has been four substantially different rockets, each stretched, re-engined, and re-plumbed. The generation names show up constantly in launch coverage.
| Generation | Years | Height | LEO payload | What changed |
|---|---|---|---|---|
| Falcon 1 | 2006–09 | 21 m | 180 kg | The company's survival vehicle. Three failures, then two successes; first privately built liquid rocket to reach orbit. |
| Falcon 9 v1.0 | 2010–13 | 53 m | ~9 t | Nine Merlin 1C engines in a square grid. No recovery hardware. |
| Falcon 9 v1.1 | 2013–16 | 68 m | 13.2 t | 60% longer tanks, uprated Merlin 1D, engines rearranged into the "octaweb" ring. First landing attempts. |
| Falcon 9 Full Thrust | 2015–18 | 70 m | 22.8 t | Densified, sub-cooled propellants for more mass in the same tanks. First successful landings, first reflight. |
| Falcon 9 Block 5 | 2018– | 70 m | 22.8 t exp. / 17.5 t reusable | The final form, designed for rapid reflight: titanium grid fins, thermal-protected octaweb, retractable legs, tougher engines. Certified for crew. |
| Falcon Heavy | 2018– | 70 m | 63.8 t exp. | Center core plus two side boosters, 27 Merlins. Rare — around a dozen flights — and reserved for heavy or high-energy missions. |
| Starship V3 | 2026– | 124.4 m | 100 t target | Fully reusable, stainless steel, methane-fueled. Both stages return. See section 08. |
"Expendable" means the booster is flown to depletion and not recovered, buying roughly 30% more payload. LEO = low Earth orbit; GTO = geostationary transfer orbit, a far more demanding target where Falcon 9 delivers about 8.3 t.
Two families. Merlin burns kerosene and flies Falcon. Raptor burns methane and flies Starship. The switch of fuel is a deliberate long-range bet, not an optimization.
Simple, cheap, gas-generator cycle: a small side burner spins the pumps and dumps its exhaust overboard. Less efficient than the alternatives, but forgiving and quick to build. Nine sea-level Merlins power a Falcon 9 booster; one vacuum-optimized Merlin with a giant bell pushes the second stage.
Kerosene leaves soot behind, which is a nuisance for reuse and a hard stop for refueling anywhere off Earth.
Full-flow staged combustion: both propellants pass through their own preburners and every gram ends up in the main chamber, so nothing is wasted and the turbines run cooler. It is the hardest rocket cycle to build, and Raptor is the first flying at scale.
Methane burns clean, which suits fast reuse, and it can in principle be manufactured on Mars from atmospheric CO2 and subsurface water. That is the whole reason for the fuel change.
| Engine | Propellant | Thrust, sea level | Flies on | Note |
|---|---|---|---|---|
| Merlin 1D | RP-1 / LOX | ~845 kN | Falcon 9 & Heavy stage 1 | Nine per booster, throttleable, restartable for landing burns. |
| Merlin Vacuum | RP-1 / LOX | ~981 kN (vac) | Falcon stage 2 | Large expansion nozzle; only works above the atmosphere. |
| Draco / SuperDraco | Hypergolic | 0.4 kN / 71 kN | Dragon | Draco for maneuvering; eight SuperDracos form the crew launch-escape system. |
| Raptor 1 | CH4 / LOX | ~185 tf | Starship V1 | First flight-capable full-flow staged-combustion engine. |
| Raptor 2 | CH4 / LOX | ~230 tf | Starship V1–V2 | Simplified plumbing, higher chamber pressure, far cheaper to build. |
| Raptor 3 | CH4 / LOX | 280 tf | Starship V3 | Under 1,525 kg dry, ~350 s specific impulse. External shielding deleted — cooling is integrated into the engine's own structure, so it looks bare compared with Raptor 2. |
tf = tonne-force. 33 Raptor 3 engines give Super Heavy roughly 9,200 tf at liftoff — about twice a Saturn V.
A Falcon 9 booster is about 60% of the vehicle's cost and it separates roughly 70 km up, traveling somewhere near 2 km/s. Getting it back intact takes three precisely timed engine burns and a target it can hit within a few meters.
Three drone ships do the offshore catching: Just Read the Instructions and A Shortfall of Gravitas in the Atlantic, Of Course I Still Love You in the Pacific. They are converted barges with station-keeping thrusters, holding position in open ocean while a 40 m rocket lands on a deck the size of a small parking lot.
Landing legs are dead mass. Starship's booster instead flies back to the launch tower and is caught in mid-air by two arms — the "chopsticks" — that pin it under its grid fins. The ship is meant to be caught the same way. Booster catches have been routine since 2024; the ship catch is the next big milestone.
Six launch pads across three states, plus the factories that feed them. Geography is not sentimental here — each site exists because of the orbits it can reach and the empty space it can drop hardware into.
A launch site can only reach orbits at least as tilted as its own latitude, and the range safety corridor narrows it further. Cape Canaveral at 28.5° N throws east over the Atlantic and covers most commercial and ISS-class orbits. Vandenberg fires south over open Pacific and owns polar and sun-synchronous work — Earth-observation, reconnaissance, and a large slice of Starlink's high-inclination shells. Starbase, at 26° N in a corner of Texas, exists mostly because SpaceX owns the land and can iterate on Starship without asking anyone for a pad.
| Site | Vehicle | Status | What it is for |
|---|---|---|---|
| SLC-40, Cape Canaveral SFS | Falcon 9 | Active | The highest-cadence pad on Earth. Now crew-capable, and hosts LZ-40, the Florida return-to-launch-site landing pad since February 2026. |
| LC-39A, Kennedy Space Center | Falcon 9 / Heavy | Active, converting | The Apollo and Shuttle pad. Falcon Heavy and crew launches; Dragon work is moving to SLC-40 as the pad is rebuilt for Starship. The FAA has cleared up to 44 Starship flights a year from here. |
| SLC-4E, Vandenberg SFB | Falcon 9 | Active | West Coast polar launches, with landings at LZ-4 on the neighboring pad. |
| SLC-6, Vandenberg SFB | Falcon 9 / Heavy | In construction | The old Delta IV and never-used Shuttle West pad, being stripped and rebuilt to roughly double West Coast capacity. |
| Starbase Pad 1 & Pad 2, Texas | Starship | Active | Pad 1 flew the early tests; Pad 2, with a flame trench and a stronger tower, has flown the V3 vehicles. |
| SLC-37, Cape Canaveral SFS | Starship | In construction | The fourth Starship pad. Between it, LC-39A and Starbase, approved capacity points toward ~146 Starship launches a year. |
A rocket gets you to orbit. A spacecraft is what stays there and comes back. SpaceX has exactly one operational one — Dragon — and one in development that is also its own rocket.
Five Crew Dragons exist and each is flown repeatedly. Roughly eighteen crewed flights since 2020 — most for NASA's ISS rotations, the rest private: Inspiration4, the Axiom missions, Polaris Dawn and its commercial spacewalk. Dragon is currently the only American vehicle that carries people to orbit.
The near-term job and the only one with a customer already waiting. A wide door in the side ejects V3 satellites like cards from a shoe.
To go anywhere past low orbit, a Starship must be refueled in orbit by other Starships. Transferring hundreds of tonnes of cryogenic propellant between two ships has never been done and is the program's biggest open technical question.
A stripped, legged, airless-landing version contracted by NASA to put astronauts on the Moon. No heat shield, no flaps — it never returns to Earth.
The stated destination. Nothing has been built; there is no life-support system, no cabin, no flown crew hardware.
Suborbital hops between cities. Frequently mentioned, never funded.
NASA's Artemis plan was restructured in February 2026: Artemis III became a crewed low-orbit docking demonstration with Starship and Blue Origin hardware, targeting 2027, with the actual landing pushed to Artemis IV.
Around 11,100 working satellites, roughly two-thirds of everything operating in orbit, serving about 12 million subscribers. It is the largest satellite constellation ever built by a wide margin, and it was assembled in seven years.
Traditional satellite internet uses a handful of satellites in geostationary orbit, 35,786 km up, where each one hangs over a fixed point on Earth. That geometry costs you a signal round trip of 70,000 km — half a second of latency, enough to make a video call miserable. Starlink instead flies at 550 km, cutting the round trip by a factor of sixty. The price is that at that altitude a satellite crosses the sky in minutes, so you need thousands of them, arranged in overlapping shells, and every user antenna must electronically re-aim itself from one to the next without moving.
Each generation is bigger than the last, and each is sized to whatever rocket can lift it. That is the constraint the whole design lives under: the satellite has to fold into a Falcon 9 fairing, and dozens have to stack flat inside it. Starship removes that constraint entirely.
| Generation | From | Mass | Defining feature |
|---|---|---|---|
| v0.9 | 2019 | 227 kg | Sixty test articles, no inter-satellite links, most already deorbited. |
| v1.0 | 2019–21 | 260 kg | The first operational build. Flat-panel bus, single solar wing, krypton ion thruster. |
| v1.5 | 2021–23 | ~295 kg | Added optical inter-satellite links, so traffic can hop satellite to satellite instead of needing a ground station in view. |
| V2 Mini | 2023– | ~800 kg | Four times the user capacity, argon thrusters, much larger arrays. The fleet workhorse. A "Direct to Cell" variant carries an LTE base station and talks to ordinary phones. |
| V2 Mini Optimized | 2025– | ~525 kg | Same capability at lower mass, so more fit on each Falcon 9. |
| V3 | 2026– | ~1,900 kg | 1 Tbps down and 160 Gbps up per satellite, gigabit service to users. Too large for any fairing — it exists only because Starship does. |
Neither the satellite nor your dish physically points. Both steer their beams electronically, thousands of times a second, hopping between spacecraft crossing overhead.
Over 24,000 optical terminals now fly, letting the network carry traffic across oceans and over Antarctica with no ground station underneath.
Satellites acting as cell towers in orbit. Texting first, then data, on unmodified phones, through partner carriers such as T-Mobile.
Ground stations that bridge the constellation to the terrestrial internet, sited near major network exchange points.
The government line: the same bus and manufacturing, hardened and classified, sold to the US national security community — including a large reconnaissance constellation.
Thousands of bright objects in low orbit interfere with optical and radio astronomy. SpaceX has added sunshades and darker coatings; the objection has not gone away.
Thirteen test flights since April 2023. The booster is close to routine. The ship is not, and nothing has reached orbit yet — every flight so far has been deliberately suborbital, aimed at the Indian Ocean, so a failure comes down in the water rather than staying up.
| Flights | When | Vehicle | What was learned |
|---|---|---|---|
| 1–2 | 2023 | V1 | Flight 1 destroyed the pad and tumbled out of control. Flight 2 proved hot staging and got both stages flying properly before both were lost. |
| 3–4 | 2024 | V1 | Reached space, survived reentry with a badly eroded flap, then achieved controlled splashdowns of both stages. |
| 5–6 | late 2024 | V1 | The first booster catch by the tower arms — arguably the program's single most important result. |
| 7–9 | 2025 | V2 | A rough season. The stretched V2 ship failed repeatedly, twice breaking up during ascent, while boosters kept performing. |
| 10–11 | 2025 | V2 | Clean flights: Starlink simulators deployed through the payload door, engine relit in space, controlled reentry. V2 retired on a high note. |
| 12 | 22 May 2026 | V3 | First flight of the taller, Raptor 3 vehicle. The booster's boostback burn failed to light properly and it crashed into the sea; the ship lost a vacuum engine on ascent but flew the rest of the mission. |
| 13 | 24 Jul 2026 | V3 | Deployed twenty V3-class Starlink satellites, relit an engine in space for 14 seconds, and splashed down intact — the first ship to survive in one piece for inspection. |
| 14 | expected now | V3 | The step everyone is waiting on: actual orbital insertion, and the first attempt to catch the ship itself. |
The near-term commercial case does not need most of that. Launching V3 Starlink satellites needs orbit, a working payload door, and a booster that comes home — which is roughly one good flight away.
Enough of it to read a launch thread without stopping.
The first Starship flight meant to reach orbit, with a ship catch attempt behind it. If both work, Starship stops being an experiment.
LC-39A is months from operational and the Gigabay at Roberts Road is closing out. An East Coast Starship debut turns one launch site into a network.
Once Starship is flying operational Starlink loads, network capacity climbs in steps of 60 Tbps rather than 3.
A single Falcon 9 first stage is two flights from passing the Space Shuttle orbiter's all-time reuse record.
The demonstration that decides whether Starship is a very large Earth-orbit truck or a vehicle that goes places.
NASA now needs Starship for a crewed orbital docking demo in 2027 and a lunar landing after that. Schedule pressure runs both ways.