Satellite production economics: the merchant market vs vertical integration
Novaspace forecasts 43,000 satellites and a $665B build-and-launch market over 2025–2034 — but five megaconstellations are 66% of the units and only 11% of the value, ~70% of manufacturing value is nationally captive, and only about 7% is genuinely open to any manufacturer. The merchant-addressable pie is roughly $2–3B a year. Inside it, the margin sits in components (Rocket Lab blends 34–40% gross), the risk sits in integration (York runs 19.5% gross at $386M of revenue), and the strategics keep proving they would rather buy distressed capacity than build it.
BUS-CLASS UNIT ECONOMICS · $/SAT AND $/KG, SOURCED
Every $/sat figure below is a contract value divided by a unit count — shown so you can audit the division. Contract values include incentives and ops scope where noted, so these are price-to-customer, not bill-of-materials. Payload class dominates unit cost, not the bus: the spread from a microsat bus to a missile-tracking bird is two orders of magnitude.
| Class | Anchor contract | $/satellite (math) | Implied $/kg |
|---|---|---|---|
| Cubesat / microsat bus | Kongsberg NanoAvionics × SpinLaunch: 280 buses for €122.5M (Orbital Today) | €122.5M ÷ 280 = €437k/bus | ~$2–6k/kg class |
| SDA transport (merchant) | York T1TL: 42 sats / $382M; T2TL Alpha: 62 sats / $615M incl. incentives (AEI/SDA) | $382M ÷ 42 = $9.1M; $615M ÷ 62 = $9.9M | ~$30–45k/kg (≈250 kg class) |
| SDA transport (in-house subsystems) | Rocket Lab T2TL-Beta: 18 sats / $515M FFP (SpaceNews) | $515M ÷ 18 = $28.6M | vertical integration captures the subsystem margin in-house |
| Missile tracking | Sierra Space SDA T2 Tracking: 18 sats / up to $740M (company/SDA releases) | $740M ÷ 18 = ~$41M incl. payload + ops | ~$50–80k/kg (payload-dominated) |
| MicroGEO | Astranis: >$1B backlog across roughly a dozen-sat manifest (SpaceNews) | $50–100M class per dedicated GEO mission | undercuts $250M+ heritage GEO birds |
| Heritage GEO | Lanteris (ex-Maxar/SSL) 1300-class | $150–300M (industry-standard range) | ~$50–80k/kg |
| Mass production (captive) | SpaceX Starlink, Redmond: ~70 sats/week per the 2026 S-1 (GeekWire) | build cost not disclosed; analyst consensus ~$0.5–1M/sat (a circulated "$50k/V2 mini" figure is not credible) | ~$2–4k/kg est. — the cost asymptote merchants compete against |
Starlink: ~8,000+ on orbit ÷ ~5 yr ≈ 1,600–2,000 sats/yr replacement floor — consistent with the disclosed ~70/week (~3,640/yr incl. growth).
SDA PWSA: ~2-yr tranche cadence (T0 2022 → T1 2024–25 → T2 2026–27) × ~5-yr design life ⇒ every tranche is also a future recompete — if tranches continue (see below).
FACTORY BENCHMARKS · CAPACITY VS CAPITAL
| Factory | Footprint | Rate | Capital context |
|---|---|---|---|
| SpaceX Redmond (Starlink) | ~1M sq ft campus | ~70 sats/week (~3,640/yr) — S-1 disclosure | internal; more output than the rest of the Western industry combined |
| Apex Factory One, LA | >100k sq ft after Sept 2025 adjacent lease (company) | 12 → 18 buses/month projected; claimed peak 200+/yr | >$718M raised total; $200M Series D at just over $1B (Sept 2025, Interlagos) then a separate $200M+ round at $2.3B closed Jun 5, 2026, Glade Brook + Washington Harbour (SpaceNews, Bloomberg) |
| K2 Space, Torrance | 180k sq ft (leased) | up to 100 high-power (20 kW+) sats/yr | ~$430M equity; $250M Series C at $3B, Dec 2025, Redpoint lead (Payload); $500M signed contracts claimed |
| AST SpaceMobile, Midland TX + FL | ~500k sq ft total, 5 TX facilities (BusinessWire) | 40 Block 2 BlueBirds in production, materials for 50+ | public (ASTS); operator-owned, zero merchant sales — the vertical-integration extreme |
| MDA Space, Montreal | 185k sq ft expansion | up to 2 sats/day (~400/yr) software-defined line (MDA) | CA$225–275M guided capex for the ramp year — the cleanest public "cost of a high-rate line" datapoint |
| Airbus OneWeb Satellites, Merritt Island FL | 105,500 sq ft, 2 automated lines (2019) + $85M / 50k+ sq ft expansion (Merritt Island RA) | 2 sats/day proven on OneWeb Gen1 (600+ built) | expansion ≈ $1,700/sq ft fully equipped; OneWeb's 2020 Ch.11 is the canonical anchor-customer loss |
| EnduroSat, Sofia | 17,500 m² (~188k sq ft) Space Center, Oct 2025 | stated target: 2 ESPA-class/day at ramp — claims far ahead of demonstrated rate | ~$153M since mid-2025 ($49M Founders Fund-led + $104M) (company) |
| Terran Orbital's unbuilt plan (cautionary) | 660,000 sq ft proposed, 2021 | 1,000 sats/yr proposed | $300M plan (~$455/sq ft) financed via Space Florida — never built; demand never supported it (TechCrunch) |
Practical read: building shell costs $200–500/sq ft; a fully tooled high-rate line lands at $1,000–1,800/sq ft. Startups lease the shell (Apex, K2 and Terran all leased) and put the capital into equipment and test.
PUBLIC-COMP MARGINS · WHAT THE FILINGS ACTUALLY SAY
York Space Systems (NYSE: YSS) — the integration benchmark. FY2025: revenue $386.2M (+52%), gross margin 19.5% (+6.8 pts), net loss $84.5M, adjusted EBITDA −$8.3M, backlog $542.6M (8-K via SEC filing). The IPO priced Jan 28, 2026 at $34 (upsized, 18.5M shares, ~$582.6M net), opened at $38 and traded to ~$26.5 by early February — the market repricing exactly the SDA-recompete risk discussed below. The lesson: even at $386M of revenue, defense smallsat integration runs sub-20% gross and negative EBITDA.
Rocket Lab (NASDAQ: RKLB) — the components counter-example. FY2025: revenue $601.8M (+38%), GAAP gross margin 34.4%, non-GAAP 39.7%, backlog $1.85B (+73%) (investor release, 2026-02-26). Space Systems — buses plus the component franchises (SolAero solar, Sinclair wheels and star trackers, separation systems, flight software) — is now roughly 70%+ of revenue (Q3 2025 segment revenue $114.2M, ~74% of the quarter); a standalone Space Systems margin is not broken out publicly, so the blended figure is the best available. The 19.5%-vs-39.7% gross spread against York is the single most important number in this vertical: components carry the margin, integration carries the risk.
MDA Space (TSX: MDA) — profitability at scale. FY2025: revenue CA$1.63B (+51%), adjusted EBITDA CA$324M (19.8% margin), backlog CA$4.0B, FY2026 guide CA$1.7–1.9B with CA$225–275M capex (Via Satellite). Even profitable constellation manufacturing consumes cash: 2026 free cash flow is guided neutral-to-negative on the Telesat Lightspeed ramp, and S&P rates the credit BB−. That is where debt markets price the best name in the sector.
The underwriting rule of thumb that falls out: gross margin 15–25% for bus integration, EBITDA breakeven only above ~$400–500M of revenue, working capital negative during ramps (milestone billing lags BOM purchases), and ~CA$225–275M of capex for a 400-sat/yr line. Set against private marks — Apex $2.3B, K2 $3B, Sierra $8B, Astranis $2.8B (the last attributed only to "a source close to the deal") on zero-to-minimal disclosed revenue — York's public fade is the cleanest markdown-risk signal in the sector.
THE SDA METRONOME · WHEN THE DEMAND CLOCK SKIPS
SDA's Proliferated Warfighter Space Architecture has been the demand metronome the merchant market priced itself against: a tranche roughly every two years, each with a ~5-year design life implying a future recompete. The award record set the unit prices in the table above — York's $382M/42 and $615M/62, Rocket Lab's $515M/18, Sierra's $740M/18. Rocket Lab's cumulative SDA book now exceeds $1.3B ($515M T2TL-Beta plus the $816M missile-warning prime confirmed Dec 2025).
The metronome skipped in 2025. The FY26 budget request canceled Tranche 3 Transport in favor of $277M sole-sourced to SpaceX's MILNET (Via Satellite); Senate appropriators moved to restore +$500M (Air & Space Forces); and by April 2026 the Space Force had pivoted to a "Space Data Network backbone" (Breaking Defense). The Tracking Layer survives and feeds Golden Dome — Lockheed selected Terran Orbital buses for its 18 T3 Tracking vehicles in Dec 2025, the captive flow-down pattern in action.
Net read: missile-warning/tracking demand is up; competed transport-bus demand is structurally at risk of SpaceX capture. Monopsony drift is risk #1 for every merchant bus maker, ahead of anchor-customer credit (OneWeb's Ch.11; Rivada's unfunded $2.4B Terran order) and fixed-price defense work at immature rate lines — the combination that turned a $1.8B SPAC into Lockheed's $450M take-out.
M&A COMPS · WHAT CAPACITY ACTUALLY CLEARS AT
| Deal | Consideration | Closed | Read |
|---|---|---|---|
| BAE Systems / Ball Aerospace | $5.55B | Feb 2024 | The cycle's largest sat-manufacturing comp — bigger than every other deal here combined. Heritage + classified franchises clear at strategic premiums. |
| Lockheed Martin / Terran Orbital | ~$450M at $0.25/share incl. debt retirement (LM) | Oct 30, 2024 | A $1.8B SPAC clearing at ~75% off. Primes buy distressed bus capacity rather than build greenfield. |
| Intuitive Machines / Lanteris (ex-Maxar Space Systems) | ~$800M — $450M cash + $350M stock (SEC 8-K) | Jan 13, 2026 | The price of the US heritage GEO bus franchise. GEO orders fell from ~20/yr industry-wide to single digits; the buyer is buying prime status, not GEO growth. |
| Rocket Lab / Mynaric | $155.3M aggregate (SEC 8-K) | Apr 14, 2026 | Optical-terminal chokepoint bought out of German StaRUG restructuring at a fraction of peak public valuation. |
| AE Industrial / York (51%) | at a $1.125B valuation (Payload) | Nov 2022 | PE entry that exited via the Jan 2026 IPO — the one institutional round-trip in the sector. |
| Apex / Phase Four (thruster tech) and Quantum Space / Phase Four (remaining assets) | undisclosed (distress) | Sept 2025 | Component assets get absorbed by vertically integrating customers — an exit path and a warning in one (SpaceNews). |
| Kongsberg / NanoAvionics (majority) | not fully disclosed | 2022 | Defense strategics consolidating bus makers as industrial assets — the European pattern. |
THE COMPONENT-SUPPLIER ENTRY PATH · WHERE NEW CAPITAL ACTUALLY FITS
The supply signals are unambiguous: Orbion's CEO says electric-propulsion demand "exceeds industry-wide production capacity" (Space Intel Report); RTX's Blue Canyon announced it is quadrupling reaction-wheel throughput (Apr 2026); Phase Four's collapse stranded customers mid-program. Meanwhile the incumbent price umbrella is wide — heritage-class wheels at $250k–1M+ per unit with 12–18 month leads (Honeywell) versus smallsat-era entrants winning at $50–150k with 3–6 month leads. And the margin spread says the same thing the public comps do: components ~35–40% gross (Rocket Lab's blend) versus 19.5% for integration (York).
The financing template is CesiumAstro — with the arithmetic stated correctly: its February 2026 "Series C" totals $470M, of which $270M is equity (Trousdale-led) and $200M is EXIM Bank debt arranged with J.P. Morgan under the "Make More in America" initiative — the $200M is inside the $470M, not on top of it (Via Satellite, company release). Stack a Texas Space Commission SEARF grant of up to $10M on top and you have the model: equity for product, export-linked federal debt for the factory, state money for the site. Contract-collateralized debt is now a repeating pattern — Muon's $45M of credit inside its Series B, Astranis's $155M Trinity delayed-draw facility.
What entry actually costs: a credible ADCS, green-propulsion or harness/PCBA line is $5–25M of capex in 10–30k sq ft, against $50–150M+ for full bus integration at credible rate (Apex, K2 and Muon spent $150–500M+ getting there). The compliance sequence gates the timeline more than the tooling: ITAR/DDTC registration (month 1; $2,500–4,000/yr in fees, $50–150k/yr in real program cost), AS9100D (months 3–9; $20–45k all-in for a small shop), CMMC Level 2 (months 6–18; $70–250k+) — roughly $150–400k of compliance budget before the first defense PO. The grant ladder substitutes for early dilution: SBIR Phase II (~$750k–1.9M) → TACFI ($375k–2M, matched) → STRATFI ($3–15M, matched) — Gravitics' up-to-$60M STRATFI is the purest "STRATFI-as-Series-B" precedent (BusinessWire).
Two caveats from the cohort's own history. First, your customer can become your competitor: Apex absorbed Phase Four's thruster line rather than keep buying it. Second, the "ITAR-free" merchant niche (South Africa's NewSpace Systems) cuts both ways — US primes increasingly require domestic, CMMC-compliant sources, which is precisely the moat a new US wheels-and-sensors line would occupy. Single-source items we are deliberately not pricing: Orbion's reported March 2026 deal (paywalled, unverified), Enpulsion's undisclosed March 2026 round, Benchmark's undisclosed Series B, and Sidus Space's $120M Lonestar total-contract-value claim.
Want the underlying dataset?
This deep dive is distilled from a 31-company research pass with per-company funding, award, facility and credit notes, cross-checked against USAspending and SEC filings. For the full roster or a facility-entry pro forma, get in touch.
Request a deeper analysis Join the intelligence waitlist →METHODOLOGY & DISCLOSURES
Compiled 2026-06-10 from SEC filings (York 424B4/8-K, SpaceX S-1, Intuitive Machines 8-K, Rocket Lab and MDA investor releases), company announcements, SDA award records, and trade press (SpaceNews, Via Satellite, Payload, Breaking Defense, Aviation Week); DoD/NASA award figures for Apex, K2, York, Millennium, Blue Canyon, Astranis, Gravitics, Benchmark, CesiumAstro and Sierra were corroborated against the AstraVeris USAspending-derived grants table. Sixteen of the highest-stakes claims were adversarially re-verified; the two contradictions found — Apex's conflated rounds (corrected to >$718M total with the June 5, 2026 $2.3B round separate from the Sept 2025 Series D) and CesiumAstro's double-counted EXIM debt (corrected to $470M inclusive) — are fixed on this page. Novaspace market-structure figures (43,000 sats; $665B; 66%/11%; 7% open; 70% captive) were corroborated verbatim across multiple outlets. Single-source claims are named and excluded from pricing use. Test-equipment cost ranges are vendor-quote estimates, marked as such in the source file. Nothing on this page is investment advice. See the methodology page, the Rocket Lab case study, or the research hub.