Space Exploration Technologies Corporation

SpaceX is no longer a launch company that happens to own a satellite network. As of September 2026 it is a three-legged infrastructure conglomerate in which the rocket business is a strategic loss-leader, the satellite broadband business is the sole profit engine, and a recently absorbed artificial-intelligence business is both the largest consumer of capital and the fastest-growing revenue line. The company controls roughly 80 percent of global mass delivered to orbit, operates the largest low-Earth-orbit constellation ever built with more than 11,000 working satellites, serves 12 million broadband subscribers across roughly 170 markets, and runs what it describes as the largest coherent AI training capacity on Earth at 1.4 gigawatts nameplate. Its unifying thesis is vertical integration: cheap, reusable launch capacity lowers the marginal cost of putting anything in orbit, including satellites, sensors and eventually GPUs. That thesis is unproven at the scale its $2 trillion valuation implies.

Share price and market capitalisation

Sources conflict materially. Contemporaneous reporting placed standalone SpaceX headcount at roughly 13,000 in late 2024. Tracxn reported 20,133 as of June 2026; Investing.com reports 21,000 as of September 2026. The 2026 figures reflect the combined SpaceX + xAI + X entity and, from mid-August, approximately 150 additional Cursor engineers. Cameron County's 2026 economic impact report records more than 3,400 SpaceX employees and contractors at the Starbase site alone, generating over 21,000 indirect jobs locally.

The company's own characterisation

The S-1 and Q2 2026 release describe SpaceX in near-identical terms: founded in 2002, "the only company building the integrated hardware and software infrastructure of the future across space, connectivity, and AI," and, at its core, "builders" who "design, manufacture, launch, and operate products and services built on cutting-edge technologies, including the world's most advanced rockets and spacecraft." The corporate mission as stated in the prospectus is to "build the systems and technologies necessary to make life multiplanetary, to understand the true nature of the universe, and to extend the light of consciousness to the stars."

The S-1's strategic logic is explicit and worth reproducing in substance because it underpins the entire valuation case. Management argues that reusability drives launch unit costs down; declining launch cost expands the set of economically viable orbital applications; SpaceX is uniquely positioned to capture those applications because it owns both the launch capacity and the satellite manufacturing capability. The prospectus states that reusable rockets, scaled satellite manufacturing and operational expertise can enable cost-effective, rapid deployment of massive AI compute satellites, that satellites in sun-synchronous orbit will be able to handle energy-intensive AI workloads such as inference at greater scale and efficiency than terrestrial alternatives, and that the company expects to begin deploying orbital AI compute satellites as early as 2028. Management further states its belief that Starship will reduce the cost to orbit by "99% or more relative to the historical average."

Independent characterisation

SpaceX operates three reportable segments with radically different economic profiles. The Space segment is a scale monopoly in commercial launch that generates modest external revenue because roughly three-quarters of its launch capacity is consumed internally by Starlink, and which is currently loss-making because Starship research and development is expensed rather than capitalised until commercialisation. The Connectivity segment is a subscription telecommunications business with 60-plus percent segment EBITDA margins, near-zero enterprise churn and a subscriber base that doubled year over year. The AI segment is, in practice, two businesses stapled together: a declining advertising and consumer-subscription business inherited from X, and a rapidly scaling wholesale GPU-compute leasing business ("neocloud") that signed more than $14 billion of contracted sales in a single quarter.

The revenue model mix as of Q2 2026 is therefore: recurring subscription (Starlink consumer, X/Grok subscriptions) approximately 37 percent; contracted enterprise and government services (Starlink enterprise, Starshield, AI compute agreements) approximately 51 percent; transactional product/service sales (third-party launch) approximately 8 percent; and advertising approximately 5 percent. This is a decisive shift from the 2023 profile, in which launch and consumer broadband dominated.

Value chain position. SpaceX is unusually vertically integrated. It designs and manufactures its own engines (Raptor, Merlin), airframes, avionics, satellites, user terminals and ground stations; it owns and operates its launch sites; it operates its own network; and since February 2026 it owns its own AI models, its own social distribution platform, and, since August 2026, its own developer tooling. The Q2 release explicitly attributes margin performance to "the power of extreme vertical integration." The counterpoint is that vertical integration converts what would be supplier margin into fixed capital commitment, which is precisely what shows up as $18.4 billion of quarterly capital expenditure.

Customer types and end markets. (i) US Government civil — NASA (crew and cargo to ISS, Artemis Human Landing System, science payloads such as the Roman Space Telescope launched 29 August 2026), NOAA, FEMA; (ii) US Government defence and intelligence — US Space Force, National Reconnaissance Office, National Security Space Launch programme, Starshield; (iii) international governments and space agencies; (iv) commercial satellite operators; (v) consumer broadband households; (vi) enterprise connectivity — airlines (American Airlines, Southwest, Virgin Atlantic, Iberia, Aer Lingus), maritime, energy, mining; (vii) mobile network operators (T-Mobile, SoftBank, NTT Docomo, Spark NZ); (viii) AI developers and hyperscalers purchasing compute (Anthropic, Google); (ix) advertisers and consumer subscribers on X and Grok.

US Government agencies accounted for approximately one-fifth of 2025 consolidated revenue per the S-1 — a concentration that is high but materially lower than most defence-adjacent primes, and diluted by the xAI/X consolidation.

Stated strategy

The S-1's strategic architecture rests on a single causal chain that management repeats in near-identical language across the prospectus, the earnings release and the earnings call: declining unit costs driven by rapid reusability produce "a compounding competitive advantage" that "not only fortifies our core business, but also provides vast new market opportunities uniquely enabled by space." The Q2 release's opening bullet frames it as "the power of extreme vertical integration."

The three-pillar articulation is: (1) achieve full and rapid Starship reusability, reducing cost to orbit by 99 percent or more; (2) use that cost position to scale Starlink capacity faster than demand, converting a broadband subscription business into a global utility; (3) use the same cost position to place AI compute in orbit, where solar power is continuous and cooling is free, at a scale terrestrial data centres cannot match.

The total addressable market claim in the prospectus is $28.5 trillion, described as the largest in human history, allocated as $370 billion to Space (1.35 percent), $1,610 billion to Connectivity (5.65 percent), and $26,510 billion to AI (93.0 percent). Roughly 90 percent of the entire TAM is therefore attributed to a segment SpaceX acquired four months before filing.

Announced initiatives, last 24 months

Sustainability and ESG commitments. No formal emissions target, net-zero commitment or sustainability-linked financing has been disclosed.

Management's medium-term targets

No formal quarterly or annual revenue/EPS guidance range is issued. The $100 billion ARR figure is an exit run-rate, not a fiscal-year revenue forecast, and includes contributions from Cursor. This distinction is routinely elided in secondary commentary and is a source of considerable confusion in published models.

Space segment

Falcon 9. Two-stage, partially reusable medium-lift launch vehicle; the workhorse of the global launch industry and the highest-cadence orbital rocket ever flown. First stage is recovered by propulsive landing on droneship or land pad. Reached its 500th launch during 2025 and demonstrated a 29th reflight of a single booster — an industry record. Published commercial pricing has historically sat in the $62–70 million range per dedicated mission, with rideshare pricing quoted per kilogram. Customers: commercial satellite operators, NASA, US Space Force, NRO, international agencies. Manufactured at Hawthorne, California.

Falcon Heavy. Three-core heavy-lift derivative of Falcon 9, first flown February 2018. Used for high-energy and heavy national-security missions; launched NASA's Roman Space Telescope on 29 August 2026. Lower cadence, premium pricing.

Dragon (Crew and Cargo). Reusable capsule; the only US-crewed orbital vehicle in service. Per S-1 disclosure, Dragon has carried 78 astronauts since 2020. In 2025 SpaceX flew all five US crew and cargo missions to the ISS. Also supports private astronaut missions and free-flyer missions (the Polaris programme).

Starship / Super Heavy. Fully reusable two-stage super-heavy launch system. Vehicle height 124.4 m, diameter 9 m, liftoff thrust approximately 80,800 kN, design payload to LEO on the order of 100 tonnes in the current V3 configuration. Development spend exceeds $15 billion cumulatively per the S-1, including $3,004 million in FY2025 and $930 million in Q1 2026 charged to R&D. Programme status as of 14 September 2026: thirteen flight tests completed. Flight 12 (May 2026) was the first V3 suborbital mission from the new Starbase Pad 2, achieving a precision upper-stage landing and deployment of modified V2 Starlink satellites. Flight 13 (24 July 2026) achieved all objectives — deployment of 20 production V3 satellites, in-space Raptor relight, and the softest splashdown to date with an intact heatshield. Flight 14, targeted for 18 September 2026 with a two-hour window opening 12:15 UTC, is planned as the first genuine orbital mission, carrying approximately 20 operational Starlink V3 satellites and attempting the first tower catch of the upper stage. Management has stated that Flight 14 will generate revenue, making it the programme's first commercially productive flight. Accounting note from the S-1: Starship costs are expensed to R&D until commercialisation and only then capitalised, which is why the Space segment's operating loss widens as the programme accelerates.

Raptor engine. Full-flow staged-combustion methalox engine powering Starship; 33 engines on the Super Heavy booster. Booster 21 completed a full 33-engine static fire on 28 August 2026; Ship 41 completed single-engine and six-engine static fires.

Human Landing System (HLS). Starship-derived crewed lunar lander under NASA Artemis contract. Included within Space segment launch and development revenues. Current schedule status is contingent on Starship orbital demonstration and propellant transfer.

Launch and development services. A distinct revenue line ($314 million in Q2 2026, $603 million H1 2026) covering development contracts, integration services and non-launch government work.

Connectivity segment

Starlink Residential / Consumer. Ku/Ka-band LEO broadband delivered via user terminal. Consumer revenues of $2,485 million in Q2 2026, up 44 percent year over year. Blended ARPU $66/month at Q2 2026, down from $85 a year earlier and from approximately $99 in 2023 — a deliberate trade of price for volume in lower-income geographies. Constellation status: approximately 9,600 satellites deployed as of March 2026; 12,881 cumulatively launched, 11,093 in orbit and 11,078 working as of 2 September 2026. Coverage: 164 countries at Q1 2026, described as 170 markets on the Q2 call.

Starlink Business / Enterprise. Higher-throughput, priority-tier service with SLA. Per the S-1, since 2023 no enterprise customer paying more than $750,000 per year has voluntarily cancelled — effectively zero churn at the top of the book. Enterprise and government revenues reached $1,806 million in Q2 2026, up 108 percent year over year.

Starlink Aviation. In-flight connectivity. Q2 2026 wins: a major agreement with American Airlines; service activation on Southwest, Virgin Atlantic, Iberia and Aer Lingus.

Starlink Maritime. Vessel connectivity for commercial shipping, offshore energy and cruise.

Starlink Mobile (Direct-to-Cell). Satellite-to-standard-handset text, voice and, prospectively, broadband. Partnerships with T-Mobile in the US and, announced in Q2 2026, SoftBank, NTT Docomo and Spark NZ internationally. Capability is being materially upgraded by the EchoStar spectrum acquisition: the FCC approved transfer of licences covering 65 MHz in the US plus certain global Mobile Satellite Service licences during Q2 2026. The S-1 assigns Starlink Mobile a $740 billion share of its claimed $1.61 trillion Connectivity TAM, against $870 billion for fixed broadband.

Starlink V3 satellites. Next-generation spacecraft, each rated at approximately 1 Tbps of downlink capacity, deployable only by Starship. Management states the next generation carries roughly twenty times the bandwidth to orbit per launch relative to the current generation.

Starshield. Secure satellite network purpose-built for US Government and national-security customers, derived from Starlink engineering: target tracking, optical and radio reconnaissance, missile early warning, and secure communications. In Q2 2026 SpaceX was awarded over $6 billion in multi-year US Government contracts, primarily from two major Space Force contracts for LEO-based communications and sensing constellations.

Starlink user terminals. Manufactured at Bastrop, Texas. Hardware is sold at or near cost as a subscriber-acquisition instrument.

AI segment

Colossus supercomputer. Colossus 1 in Memphis, Tennessee comprises approximately 220,000 Nvidia GPUs and roughly 300 MW; the Colossus estate across sites totals approximately one million GPUs. Nameplate compute rose from 0.4 GW (Q2 2025) to 1.0 GW (Q1 2026) to 1.4 GW (Q2 2026), with Colossus II under construction. Target revised upward from 2 GW by end-2026 to 5–10 GW by end-2027.

Cloud Services Agreements (compute leasing). The fastest-growing product in the company. Anthropic contracted on 6 May 2026 for the entire Colossus 1 capacity at $1.25 billion per month. Google contracted on 5 June 2026 for approximately 110,000 GPUs at $920 million per month, at a reduced rate through September 2026 and the full rate to 2029. A fourth agreement announced in September 2026 is worth approximately $1.11 billion per month (~$13.3 billion annualised) from 1 December 2026. Q2 2026 alone closed $14.1 billion of contracted sales, contributing $1.6 billion of incremental AI infrastructure revenue in the quarter. Material caveat disclosed in the filings: the Anthropic and Google contracts carry 90-day termination clauses.

Grok. Frontier large language model family. Grok 4.5 released July 2026, described as the largest and most capable model to date, trained alongside Cursor and incorporating a 1.5-trillion-parameter V9 foundation model, optimised for coding, agentic tasks and knowledge work. Monetised through consumer and enterprise subscriptions and API access.

X (formerly Twitter). Social platform generating advertising revenue ($367 million in Q2 2026, down from $426 million a year earlier), subscriptions and data licensing. X Money payments product launched in 2026. Advertising is the only declining revenue line in the entire company.

Cursor (Anysphere). AI coding assistant acquired for $60 billion in an all-stock transaction closing 14 August 2026 (389,289,254 Class A shares issued; no cash consideration). Pre-acquisition annual recurring revenue of approximately $4 billion; engineering team of roughly 150. Now a wholly owned subsidiary within the AI division alongside Grok and X. Reporting indicates the Cursor brand will be phased out for new products. Stated rationale from Cursor's side: access to the largest GPU fleet in the world. Internal integration targets include C++ and Rust code generation for flight computers, engine controller firmware and actuation software.

Orbital AI compute ("Starmind" concept). Proposed satellite platform integrating GPU hardware for in-orbit inference, referenced on the Q2 2026 earnings call. The S-1 targets first deployment as early as 2028. This product does not exist, generates no revenue, and carries the largest share of the valuation case. Morningstar assigned a 7 percent probability to the combined scenario of full Starship reusability plus commercially viable orbital data centres.

--