Fleet — Bus supply — SATCOM Intel

The read

Space went commercial, and then one company went vertical. Business-owned spacecraft were under a third of what flew in 2000. They are seven in eight now. Take SpaceX out and the shift is still real, only smaller, which is the first thing worth establishing because every chart after it is a cut of the same population.

The bus industry that should have grown with all this did not. Volume went up more than thirty times. The number of companies supplying that volume did not, because most of the growth was operators building for themselves. The merchant market — hardware somebody actually bought — is a sixth of what flew last year, and barely more than half of it even without SpaceX in the denominator.

Orbit is where the oversupply argument lives. Geostationary has run at roughly thirty spacecraft a year for twenty-five years and is now under a third of one per cent of what launches. Nine companies still build for it. That is the surplus the consolidation dossier has been circling, and it is invisible in any chart that runs every orbit together.

The order book says the structure repeats. Of the satellites under construction for the next decade, seven in ten will be built by the operator that flies them. Take SpaceX out of that and it is still two in three, because Amazon Kuiper occupies the same position one rung down. Captive integration is not a SpaceX fact. It is the shape of the industry.

What this cannot tell you. Every number here counts spacecraft, not money, so a cubesat and a six-tonne GEO bird weigh the same. Nothing in either source names who supplied a communications payload when that was not the prime. Those are the two questions a reader should carry away unanswered.

Written 2026-09-18, due a rewrite by 2027-09-18. The charts below refresh themselves from the mirror; this does not. Edit it at data/census/read.md.

Liveupdated 2 Oct 202687% of everything launched in 2025 was owned by a business, against 31% in 2000 — governments and universities fly roughly what they always didTake SpaceX out and business ownership is 56%, against 31% in 2000 — the commercial shift is real without SpaceX, on 11× the volumeEvery spacecraft launched since 2000, by the class of organisation that owns it. Ownership, not manufacture: this counts the university CubeSats and the one-off builds that the maker charts further down leave out. With SpaceX in, 2025 is 4,511 spacecraft, 479 of them government and 104 academic.Without SpaceX, 2025 is 1,336 spacecraft — 479 government, 104 academic. The lower panel is the same data as a share, so the early years stay readable. Hover a year for the split.

BusinessAcademicCivil governmentDefence

Show data table

Spacecraft launched per year by owner class. SpaceX-owned shown separately so either view can be reconstructed.
Year	Defence	Civil government	Academic	Business	Total	of which SpaceX
2000	29	44	8	37	118	0
2001	27	44	0	15	86	0
2002	14	37	5	34	90	0
2003	27	35	7	19	88	0
2004	17	35	2	18	72	0
2005	15	32	5	19	71	0
2006	27	42	2	22	93	0
2007	33	45	5	27	110	0
2008	22	35	7	37	101	1
2009	32	47	15	29	123	0
2010	41	43	9	27	120	2
2011	32	54	9	32	127	0
2012	29	55	17	33	134	2
2013	57	60	49	42	208	1
2014	41	56	39	118	254	2
2015	48	47	34	103	232	2
2016	23	71	31	90	215	2
2017	42	57	62	282	443	4
2018	64	101	75	220	460	6
2019	62	73	58	285	478	124
2020	51	38	30	1,129	1,248	837
2021	68	75	79	1,591	1,813	996
2022	95	87	94	2,172	2,448	1,695
2023	89	96	83	2,632	2,900	1,990
2024	169	127	82	2,449	2,827	1,970
2025	197	282	104	3,928	4,511	3,175
2026 *	94	105	55	2,320	2,574	1,697

Source: Jonathan McDowell, GCAT main catalogue and organisations table, read at mirror time — data/series/manufacturing-built.jsonl, as of 2026-10-02. Launch attempts, so a failed launch still counts as flown.

Classes: GCAT's organisation Class for the registered owner — A academic, B business, C civil government, D defence. Every spacecraft in the window carries one; nothing falls into an unknown bucket. A state-owned academy such as CAST is civil, not business.

Spacecraft, not payloads: GCAT's "payload" means the whole satellite as against a rocket stage or debris — never the transponder payload as against the bus. Who supplied a communications payload is in no source the desk holds, and this census does not claim it.

SpaceX: 12,506 spacecraft over the window, all business-owned. The without-SpaceX view subtracts them from the business band rather than re-counting.

2026 is partial — the catalogue runs to 2026-10-02. Shown hatched.

Liveupdated 2 Oct 2026SpaceX built 75% of every spacecraft that flew in 2025; the other 193 makers together built 1,079Without SpaceX the industry built 1,079 spacecraft in 2025, 10× its 2012 output — and 44% of them never reached a merchant marketSpacecraft launched per year by the company that built them, 2012 to 2026 (2026 partial). The stack peaks at 4,343 in 2025, which is SpaceX; the shapes beneath keep their own scale on the other view.On its own scale the rest of the industry peaks at 1,079 in 2025 — an axis 3× shorter than the one SpaceX needs. The lower panel is the share of each year's spacecraft built by their own operator — captive, not merchant, supply, SpaceX included, which is why it sits so high.SpaceX excluded, which is the honest read on the merchant market. Hover a year for the split.

SpaceX186 other makersSwarmSpireSECM (Shanghai)CASTAmazon KuiperOneWebPlanet

186 other makersSwarmSpireSECM (Shanghai)CASTAmazon KuiperOneWebPlanet

Show the numbers

Spacecraft launched per year by maker, 2012 to 2026. Captive = built by the operator that owns them. The last two columns are the without-SpaceX view.
Year	SpaceX	186 other makers	Swarm	Spire	SECM (Shanghai)	CAST	Amazon Kuiper	OneWeb	Planet	All makers	Captive	Ex-SpaceX	Captive ex-SpaceX
2012	2	97	0	0	0	13	0	0	0	112	20%	110	18%
2013	1	146	0	0	1	6	0	0	1	155	29%	154	29%
2014	2	141	0	0	1	10	0	0	61	215	47%	213	46%
2015	2	148	0	4	1	11	0	0	40	206	43%	204	42%
2016	2	117	0	16	8	15	0	0	32	190	41%	188	40%
2017	4	193	0	36	9	9	0	0	139	390	51%	386	50%
2018	6	275	7	28	17	18	0	0	36	387	35%	381	34%
2019	124	210	2	16	5	19	0	0	32	408	60%	284	43%
2020	837	172	36	16	9	13	0	104	35	1,222	80%	385	37%
2021	996	268	76	16	11	19	0	284	48	1,718	72%	722	34%
2022	1,703	427	56	11	4	17	0	110	44	2,372	84%	669	43%
2023	1,994	515	24	25	11	19	2	132	73	2,795	84%	801	44%
2024	2,076	505	0	15	43	28	0	20	36	2,723	82%	647	40%
2025	3,264	609	0	39	84	91	180	0	76	4,343	84%	1,079	44%
2026 *	1,739	354	0	29	99	38	216	0	4	2,479	83%	740	48%

Source: Jonathan McDowell, GCAT main catalogue and organisations table, read at mirror time — data/series/manufacturing-built.jsonl, as of 2026-10-02. Spacecraft with a named maker; launch attempts, so a failed launch still counts as built.

Spacecraft, not payloads: GCAT's "payload" means the whole satellite as against a rocket stage or debris, never the transponder payload as against the bus. The bus line each spacecraft flew is a separate column and a separate metric; who supplied a communications payload is in no source the desk holds.

Makers: GCAT names the plant; the extractor folds sites and subsidiaries into one maker (SpaceX/Seattle → SpaceX, Spire/Glasgow → Spire). "All makers" counts 194 in the window; the named seven are the largest outside SpaceX by total.

Not plotted: 1,028 spacecraft built by makers with fewer than five across the whole window — a university flying one CubeSat is not supply capacity. The cut is by count, never by size: Planet and Spire fly small hardware in volume and are counted in full.

Captive: the spacecraft's owner is its maker or the maker's parent. Starlink, Kuiper and Planet are captive; a bus sold to an operator is merchant supply. This is the history half of the bus-supply census; the forward half is the Novaspace backlog in manufacturing-backlog.jsonl.

2026 is partial — the catalogue runs to 2026-10-02. Shown hatched.

Liveupdated 2 Oct 2026Only 16% of the spacecraft launched in 2025 were built for somebody else — the rest were built by the company that flies themStrip SpaceX out and 56% of what the industry built in 2025 was sold to a third party: a merchant market of 601 spacecraftMerchant supply — a maker's spacecraft minus the ones its own group owns — stacked by who built it, 2012 to 2026 (2026 partial). This is the market a bus is actually sold into, and it is the one the charts above hide: captive builders dominate the totals. SpaceX barely sells buses, so the switch hardly moves the top panel — 690 against 601. What it moves is the denominator underneath.The merchant market is nearly the same size either way; what the switch changes is how much of the industry it represents — 56% here against 16% with SpaceX counted in. The lower panel is that share. Hover a year for the split.

SpaceX129 other makersAirbus D&SNorthrop GrummanDFHSpireCASTSECM (Shanghai)OneWeb Satellites

129 other makersAirbus D&SNorthrop GrummanDFHSpireCASTSECM (Shanghai)OneWeb Satellites

Show the numbers

Merchant spacecraft per year — built, less the ones the maker's own group owns.
Year	Built	Merchant	Share	Merchant ex-SpaceX	Share ex-SpaceX
2012	112	90	80%	90	82%
2013	155	110	71%	110	71%
2014	215	115	53%	115	54%
2015	206	118	57%	118	58%
2016	190	113	59%	113	60%
2017	390	193	49%	193	50%
2018	387	252	65%	251	66%
2019	408	163	40%	163	57%
2020	1,222	243	20%	243	63%
2021	1,718	475	28%	475	66%
2022	2,372	386	16%	378	57%
2023	2,795	449	16%	445	56%
2024	2,723	491	18%	385	60%
2025	4,343	690	16%	601	56%
2026 *	2,479	426	17%	384	52%

Source: Jonathan McDowell, GCAT main catalogue and organisations table — data/series/manufacturing-built.jsonl, as of 2026-10-02. Merchant is derived, not reported: a maker's spacecraft for the year minus the ones whose owner is the maker or its parent.

What this is not: a revenue figure. A unit of merchant supply is one spacecraft, whether it is a 5 kg cubesat or a six-tonne GEO bird, so this counts deals, not money. The mix by orbit is the next chart.

OneWeb Satellites is the largest merchant maker in the window on this definition — the Airbus joint venture carries its own organisation code, so building for OneWeb the operator reads as a sale. A reader who treats the JV as captive should subtract it; the data table gives the number.

Folded: 129 makers beyond the named 7 sit in the grey band. Colour follows the maker, so a company keeps its hue across every chart in this sequence.

2026 is partial — the catalogue runs to 2026-10-02. Shown hatched; the annotation reads the last complete year.

Liveupdated 2 Oct 2026Geostationary never grew: 43 spacecraft in 2000, 25 in 2025 — and Maxar (SS/Loral), CAST, Northrop Grumman are still the ones building themMEO is a navigation programme, not a market: 252 spacecraft in 27 years, almost all of them ISS Reshetnev and Lockheed MartinLEO went from 58 spacecraft in 2000 to 4,451 in 2025 — this is the whole of the industry's growth, and one company is most of itWithout SpaceX, LEO still went from 58 to 1,276 — the growth is real, but it is 3× smaller than the headlineSpacecraft launched per year into each orbit regime, stacked by the company that built them, 2000 to 2026 (2026 partial). The lower panel is the regime's share of everything launched that year. Roughly thirty a year for twenty-five years, through two booms and a bust — the flat market the consolidation dossier is about. The sequence's SpaceX switch does nothing here: SpaceX has never built a geostationary spacecraft.Constellations are replaced in batches, so the count is lumpy. The buyers are governments and the builders barely change. The SpaceX switch is inert here too.Hover a year to see who built the stack.

25 other makersLockheed MartinAirbus D&SThales Alenia SpaceBoeingNorthrop GrummanCASTMaxar (SS/Loral)

12 other makersPolyotThales Alenia SpaceCASTBoeingOHBLockheed MartinISS Reshetnev

192 other makersCASTSpireSECM (Shanghai)Amazon KuiperOneWeb SatellitesPlanetSpaceX

191 other makersDFHCASTSpireSECM (Shanghai)Amazon KuiperOneWeb SatellitesPlanet

Show the numbers

Spacecraft launched per year by orbit regime. HEO, transfer and beyond-Earth orbits are counted in the denominator but have no panel of their own.
Year	GEO	MEO	LEO	LEO ex-SpaceX	All orbits
2000	43	6	58	58	118
2001	23	5	50	50	86
2002	27	3	54	54	90
2003	27	6	46	46	88
2004	19	7	41	41	72
2005	22	5	40	40	71
2006	29	8	48	48	93
2007	25	9	65	65	110
2008	30	8	59	58	101
2009	31	5	79	79	123
2010	32	7	74	72	120
2011	32	10	78	78	127
2012	38	9	84	82	134
2013	20	6	171	170	208
2014	32	19	193	191	254
2015	37	11	177	175	232
2016	32	12	164	162	215
2017	35	9	395	391	443
2018	31	31	385	380	460
2019	24	16	428	304	478
2020	19	6	1,213	376	1,248
2021	25	5	1,772	776	1,813
2022	43	13	2,382	687	2,448
2023	27	8	2,850	860	2,900
2024	26	11	2,766	796	2,827
2025	25	13	4,451	1,276	4,511
2026 *	8	4	2,555	858	2,574

Source: Jonathan McDowell, GCAT main catalogue and organisations table — data/series/manufacturing-built.jsonl, as of 2026-10-02. Orbit is GCAT's OpOrbit, the operational orbit and not the injection orbit, so a GEO bird delivered through a transfer orbit counts as GEO. Coverage is 100% of spacecraft in the window.

Corporate lineage: a predecessor is folded into the company that holds the product line today — SS/Loral reads Maxar, Alcatel reads Thales Alenia, OSC reads Northrop Grumman, NPO-PM reads ISS Reshetnev. That answers who can supply a bus now, at the cost of renaming history.

Hues are per regime, because the supplier sets barely overlap: the companies that build GEO buses are almost absent from LEO. The legend is redrawn with each regime, so identity never rests on colour alone.

Not shown: HEO, transfer and beyond-Earth orbits — 252 spacecraft over the window. They are in the denominator of the share panel and in the table's last column, but no regime panel of their own.

2026 is partial — the catalogue runs to 2026-10-02. Shown hatched; the annotation reads the last complete year.

Snapshotas of 18 Sept 20268,580 satellites are on order for 2025 to 2034, and 71% of them will be built by the company that flies themTake SpaceX out and the book is 6,739 satellites, still 65% captive — Amazon Kuiper alone accounts for 3,236 of themThe order book: satellites under construction by the year they are due to launch, stacked by maker. Captive concentration is not a SpaceX artefact — take it out and the share barely moves, because another operator-builder replaces it.Amazon Kuiper occupies the place SpaceX held in the history charts, which is the point: the structure repeats with a different name on it. The lower panel is the captive share of each year. The book thins from 2030 because orders that far out have not been placed yet — that is the survey date showing through, not demand falling away.

SpaceX426 other makersYuanxin SpacesailSpireNorthrop GrummanMDAPlanetNot yet assignedAmazon Kuiper

426 other makersYuanxin SpacesailSpireNorthrop GrummanMDAPlanetNot yet assignedAmazon Kuiper

Show the numbers

Satellites on order by launch year. Aggregated counts only — the underlying survey rows are licensed and are not published.
Launch year	On order	Captive	Share	Ex-SpaceX	Share ex-SpaceX
2025	2,912	2,332	80%	1,131	57%
2026	1,406	974	69%	1,346	72%
2027	1,361	941	69%	1,361	69%
2028	1,274	945	74%	1,274	74%
2029	928	710	77%	928	77%
2030 †	224	61	27%	224	27%
2031 †	186	54	29%	186	29%
2032 †	170	36	21%	170	21%
2033 †	85	35	41%	85	41%
2034 †	34	7	21%	34	21%

Source: Novaspace, Satellites to be Built and Launched by 2034 — World Market Survey (2025), commercial and government backlog sheets — data/series/manufacturing-backlog.jsonl, extracted 2026-09-18. Licensed data: only aggregated counts are published here, never a satellite row.

† The book thins from 2030. Satellites are ordered two to four years before launch, so the far end of the survey holds only the contracts already signed. Read the near years as a forecast and the far years as a floor; the taper is the survey's date, not a demand curve.

Against the history: the charts above count what flew, from the GCAT catalogue. This counts what is on order, from a commercial survey. Two sources, one unit — a satellite — and deliberately not one axis: the overlap years would be double counted.

Folded: 426 makers beyond the named 7 sit in the grey band, and "Not yet assigned" is the survey's own bucket for satellites whose builder is not yet contracted. Colour follows the maker across every chart in this sequence.

Who builds the fleet

The read