SpaceX company dossier 054
SpaceX: Business Structure, Technology and Economics
A segment-first guide to what Space, Connectivity and AI sell, who pays, how the businesses reinforce one another, who controls the group and which plans remain unproven.01
SpaceX in one minute
SpaceX is a public infrastructure company organised into three reportable segments: Space, Connectivity and AI. Space transports payloads and people. Connectivity provides satellite communications. AI operates compute, models and software.
The segments reinforce one another without being the same business. Space deploys SpaceX-owned satellites; Connectivity turns them into network services; AI combines group-scale compute with models and software distribution. Each segment also serves external customers.
Reportable segments3Space, Connectivity, AI
Q2 2026 revenue$7.814bnQuarter ended 30 June
Starlink service lines12.0mAt 30 June 2026
Public marketSPCXNasdaq and Nasdaq Texas
Reportable segment. A major business area that the chief operating decision-maker separately manages and evaluates for financial reporting. It is a reporting view, not a legal subsidiary; products and customers are different categories.
02
The three operating segments
This is the article’s organising map. Each segment has a different role and revenue logic. Government work is important, but it is not a fourth segment.
| Segment | What it does | Main revenue logic | Key assets |
|---|---|---|---|
| Space | Moves payloads and people to and through space. | Missions and development contracts. | Falcon, Dragon and Starship. |
| Connectivity | Operates satellite communications services. | Subscriptions, hardware and contracted services. | Starlink, Starshield and Plaser. |
| AI | Develops compute, models and software. | Subscriptions, usage, advertising and compute services. | COLOSSUS, Grok, X and Cursor. |
Read the segments before the brands. Products and programmes enable a segment; customers pay that segment; legal entities establish ownership. The detailed sections below map those distinctions.
03
How the three segments fit together
- Space → Connectivity.SpaceX launch systems deploy and replenish the Starlink network; internal launches create Connectivity assets rather than Space revenue.
- Connectivity → group economics.Subscriptions and contracts give Connectivity a recurring revenue profile unlike mission-based Space activity.
- Shared infrastructure → AI and future systems.AI currently uses terrestrial compute. Planned orbital compute would combine Space launch, Connectivity links and AI workloads, but it is not operational.
Feedback loop. Shared capital, engineering, procurement, networking and software can strengthen more than one segment, while every segment also has its own external customers.
04
Space: transport and mission systems
The Space segment sells access to orbit and develops spacecraft or mission systems. It serves commercial satellite operators, civil agencies and national-security customers. It also supplies launch capacity used internally to deploy SpaceX-owned satellites.
External missions generate launch or development revenue. Internal Starlink launches create assets for Connectivity rather than Space-segment revenue. Space can therefore be strategically essential even when much of its visible launch activity is absent from reported segment revenue.
| Product or programme | What it does | Who pays | Commercial maturity |
|---|---|---|---|
| Falcon 9 | Reusable launch for dedicated and rideshare payloads. | Satellite operators such as SES, Eutelsat and OneWeb; NASA and other US agencies. | Operational |
| Falcon Heavy | Higher-energy missions using three Falcon-derived cores. | NASA, national-security agencies and commercial mission buyers. | Operational |
| Dragon | Cargo and crew transport to low-Earth orbit. | NASA, Axiom Space and private human-spaceflight customers. | Operational |
| Starship and Super Heavy | Fully reusable heavy transport intended to increase mass and cadence. | SpaceX programmes plus government and commercial development customers. | In development; early payload deployment |
Why Space matters. It sells transport externally and supplies launch infrastructure to SpaceX itself. Government is a customer class, not a fourth segment; Starshield remains within Connectivity.
05
Connectivity: networks and recurring services
The Connectivity segment turns satellites, spectrum, gateways, laser links and terminals into communications services. It serves households, businesses, airlines, shipping groups, land fleets, mobile operators, governments and spacecraft operators.
Consumer service commonly produces monthly subscriptions. Enterprise, mobility and government arrangements can combine terminals, usage and multi-year contracts. This mixture made Connectivity the group’s largest revenue and operating-income contributor in Q2 2026.
| Offering | What it provides | Who pays | Commercial maturity |
|---|---|---|---|
| Starlink Consumer and Business | Fixed-site broadband through Starlink terminals. | Households, small businesses, enterprises and remote sites. | Operational |
| Starlink Aviation | In-flight connectivity for commercial and private aircraft. | Airlines: United, Southwest, American, Qatar Airways and Lufthansa Group; business-aviation operators. | Operational |
| Starlink Maritime | Broadband for passenger, merchant and offshore vessels. | Cruise and shipping: Carnival, Royal Caribbean, MSC Cruises, Norwegian Cruise Line and Maersk. | Operational |
| Starlink Land Mobility | Connectivity for rail, agricultural, emergency and industrial fleets. | Land mobility: John Deere, Brightline, Italo Treno and emergency-response organisations. | Operational |
| Starlink Direct to Cell | Direct satellite links to compatible mobile devices. | Mobile-network operators and their subscribers. | Operational; service scope expanding |
| Starshield | Secure communications, Earth observation and hosted payloads. | Government, defence and national-security organisations. | Operational; multi-year contracts |
| Plaser | Optical terminals and access to the Starlink laser relay. | Satellite manufacturers and spacecraft operators. | Operational |
Why Connectivity matters. It converts launch capacity into recurring communications revenue. Coverage is not capacity: local demand, spectrum, gateways, terminals and service priorities still determine the experience available to each customer.
06
AI: compute, models and software
The AI segment combines physical compute infrastructure with models and distribution products. COLOSSUS supplies computation; Grok supplies models and APIs; X supplies information and distribution; Cursor supplies a professional software-development surface.
Revenue comes from subscriptions, API usage, advertising, data licensing and reserved compute. The segment is also capital-intensive: Q2 2026 revenue was $2.561bn, research and development expense was $2.178bn and operating loss was $1.257bn.
| Layer | Role | Who pays | Commercial maturity |
|---|---|---|---|
| COLOSSUS and COLOSSUS II | AI training, inference and contracted compute. | Anthropic and other compute customers; internal SpaceXAI workloads. | Operational; scaling |
| Grok | Models, applications and APIs. | Consumers, developers, enterprises and public-sector API or subscription users. | Operational; frequent releases |
| X | Real-time information, distribution, subscriptions and advertising. | Advertisers, subscribers and data or service customers. | Operational |
| Cursor | AI coding editor, agents and cloud execution. | Individual developers, software teams and engineering organisations. | Operational; acquired and integrating |
Why AI matters. It adds compute and software businesses, but remains capital-intensive and loss-making. X.AI Holdings is a legal subsidiary; Cursor’s disclosed acquirer was SpaceX—not xAI.
07
Technology and infrastructure stack
This section maps capabilities to the segments that use them. Product detail remains in the relevant segment sections.
| Technology | Function | Primary segment(s) | Status |
|---|---|---|---|
| Falcon and Dragon | Transport payloads and people. | Space | Operational |
| Starship | Heavy transport and deployment platform. | Space; planned cross-segment use | In development |
| Starlink satellites and lasers | Orbital communications network. | Connectivity | Operational |
| Gateways, points of presence and terminals | Ground access and routing. | Connectivity | Operational |
| COLOSSUS and COLOSSUS II | AI training, inference and contracted compute. | AI | Operational; scaling |
| Grok, X and Cursor | Models, distribution and developer software. | AI | Operational |
| AI1 and Starmind | Proposed orbital AI compute. | Space, Connectivity and AI | Announced / planned |
What is outside the group. Tesla, Neuralink and The Boring Company are separate companies, not SpaceX business units. Collaboration does not create ownership.
08
Starmind and orbital AI data centres
SpaceX has disclosed a programme to place modular AI-compute satellites in sun-synchronous orbit, beginning as early as 2028. The programme is called Starmind; AI1 is the first published satellite design. It is intended mainly for high-power AI workloads such as inference, with results returned through laser links and the Starlink network.
This is the clearest planned intersection of all three segments. Space supplies Starship and satellite manufacturing. Connectivity supplies optical links and global network access. AI supplies processors, workloads and customers. No commercial orbital-compute service is operating at this research cut.
AI1 compute payload150 kWPeak; 120 kW average
Deployed geometry20 mHeight; 70 m wingspan
Gigasat targetLate 2027Thousands of satellites
Deployment target2028As early as stated in filing
- Manufacture modular compute satellites.The planned Gigasat Factory in Bastrop is intended to produce AI satellites at high volume. AI1 is described as processor-vendor agnostic so compute modules can change with new chip generations.
- Use Starship for mass to orbit.The architecture depends on moving large satellite mass economically and repeatedly; full and rapid Starship reuse therefore remains a central unproven dependency.
- Generate power in sun-synchronous orbit.Large solar arrays are intended to supply sustained energy without terrestrial grid, land or weather constraints.
- Run selected AI workloads on the satellite.SpaceX emphasises localised AI compute and energy-intensive inference. Training, data placement, radiation tolerance, maintenance and processor replacement remain important workload boundaries.
- Return results through optical links.High-speed lasers would connect AI satellites to one another and to Starlink, which would carry results towards ground networks and users.
| Evidence level | Published position | Analytical boundary |
|---|---|---|
| Published design | AI1: 150 kW peak and 120 kW average compute payload; 20 m deployed height; 70 m wingspan; 70 kW per tonne. | These are company specifications, not independently measured orbital performance. |
| Manufacturing plan | Gigasat Factory intended to support thousands of AI satellites starting as early as late 2027. | Factory completion, production yield and launch integration are not yet demonstrated at that scale. |
| Deployment plan | Orbital AI-compute satellites expected as early as 2028. | A target date is not an operational service commitment. |
| Scale ambition | Potentially millions of satellites and a long-term goal of launching 100 GW of compute annually. | Mass, chips, solar arrays, thermal rejection, networking, orbital safety and economics would all need to scale by orders of magnitude. |
| Efficiency claim | SpaceX says orbital radiative cooling can reduce cooling-power overhead by an order of magnitude. | Vacuum removes convection rather than heat itself; large radiators must reject waste heat, and full-system economics remain unproven. |
Quantum-computing boundary. The reviewed SpaceX roadmap describes conventional AI processors, modular compute hardware and AI workloads. It does not disclose quantum processors or an orbital quantum-computing data-centre programme. Quantum computing should therefore remain outside the stated roadmap unless primary evidence appears.
09
Ownership, control and legal structure
The operating segments explain how management reports the businesses. Legal entities establish ownership. The two relationships are shown separately below.
| Classification | Current position | Why it matters |
|---|---|---|
| Legal entity: parent | Space Exploration Technologies Corp.; Texas corporation; public since June 2026. | Parent of the consolidated group. |
| Reportable segments | Space, Connectivity and AI. | Management and financial-reporting categories, not legal entities. |
| Wholly owned subsidiary | X.AI Holdings LLC; owns X Holdings and X.AI Corp. | Places xAI and X in the legal group; AI remains a reportable segment. |
| Wholly owned subsidiary | Anysphere, Inc.; Cursor is its operating business. | The merger closed on 14 August 2026 with Anysphere surviving as a SpaceX subsidiary. |
| Executive leadership | Elon Musk is chief executive, chief technology officer and chair; Gwynne Shotwell is president and chief operating officer. | The chief executive is the decision-maker across all three segments. |
| Voting control | Class A carries one vote per share; Class B carries ten. The prospectus estimated Musk at 83.5% of voting rights after the fully exercised offering option. | Economic ownership and voting power differ; SpaceX is controlled. |
| Largest other disclosed holder | Valor-related entities held 503.4m Class A shares before the offering and 6.7% after the base offering. | A substantial institutional holding without control of Class B. |
| Public investors | The IPO sold 638.9m new Class A shares after the underwriters’ full option exercise. | Public economic ownership coexists with founder control. |
10
Customers and key counterparties
Not every organisation connected to SpaceX has the same role. The relationship type and relevant segment are therefore stated explicitly.
| Organisation or group | Relationship | Segment | What it means |
|---|---|---|---|
| NASA, National Reconnaissance Office and other US agencies | Customer | Space | Buy launch, crew, cargo, development or classified mission services. |
| Axiom Space, SES, Eutelsat and OneWeb | Customer | Space | Buy dedicated or rideshare payload deployment. |
| United, Southwest, American, Qatar Airways and Lufthansa Group | Customer | Connectivity | Buy fleet connectivity and passenger internet. |
| Carnival, Royal Caribbean, MSC Cruises, Norwegian Cruise Line and Maersk | Customer | Connectivity | Buy vessel connectivity for passengers, crews and operations. |
| John Deere, Brightline, Italo Treno, FEMA, NOAA and national governments | Customer | Connectivity | Buy industrial, transport, disaster-response or government communications. |
| Anthropic | Customer | AI | Agreements signed in May 2026 cover access to approximately 325,000 NVIDIA GPUs across COLOSSUS and COLOSSUS II. |
| Consumers, developers, engineering teams, enterprises and advertisers | Customer or user | AI | Buy or use Grok, X, APIs, subscriptions, advertising, compute and Cursor services. |
| Tesla, NVIDIA and infrastructure providers | Supplier or collaborator | Cross-segment | Supply equipment or collaborate on infrastructure; they are not SpaceX subsidiaries. |
| Launch, spectrum, communications and environmental authorities | Regulator | Cross-segment | Authorise or constrain launches, frequencies, networks, sites and services. |
| Elon Musk, Valor-related entities and public Class A shareholders | Owner or investor | Corporate | Hold economic or voting interests; that does not make them customers. |
Context can change the label. A government agency may be a customer in one transaction and a regulator in another. A Musk-related company may supply or collaborate with SpaceX without being owned by it.
11
How the economics differ
The useful comparison is not merely what each segment sells, but how revenue arrives, what must be funded first and what limits scale.
| Q2 2026, $m | Space | Connectivity | AI | Group |
|---|---|---|---|---|
| Revenue | 962 | 4,291 | 2,561 | 7,814 |
| Operating income / (loss) | (542) | 1,656 | (1,257) | (143) before group interest and other items |
| Dimension | Space | Connectivity | AI |
|---|---|---|---|
| Revenue character | Mission and project revenue. | Recurring services plus terminals and contracts. | Subscriptions, usage, advertising and reserved compute. |
| Timing | Launch revenue usually follows deployment; development revenue is recognised over time. | Consumer service is commonly monthly; enterprise and government terms vary. | Software and compute revenue follows subscription, usage or contracted capacity. |
| Capital base | Vehicles, engines, launch sites and development programmes. | Satellites, launches, spectrum, gateways and terminals. | Processors, power, cooling, storage, networks and model research. |
| Customer concentration | Material exposure to large government and commercial missions. | Broad consumer base plus negotiated fleet, enterprise and government contracts. | Broad software audience, but infrastructure revenue depends on relatively few large customers. |
| Principal scaling constraint | Reliability, launch cadence, regulation and reusable-vehicle economics. | Regional capacity, spectrum, terminals, gateways and replenishment launches. | Chip supply, power, cooling, networking, model demand and capital intensity. |
| Current economic position | Strategically enabling but loss-making in Q2 2026. | Largest revenue and operating-income contributor. | Large and growing, but loss-making and research-intensive. |
Internal infrastructure is not booked revenue. Space’s reported revenue excludes internal Starlink launches. Connectivity benefits from launch capacity without an inter-segment sale appearing as Space revenue. The same principle would apply when deploying SpaceX-owned AI satellites.
12
Corporate developments: xAI, X and Cursor
Cursor is part of the SpaceX group, but the disclosed acquirer is SpaceX—not xAI. Cursor announced on 14 August 2026 that SpaceX had completed the acquisition. The process began with an April partnership involving SpaceXAI, which explains why operational descriptions can differ from the legal ownership chain.
| Date | Transaction or event | Result |
|---|---|---|
| October 2022 | Elon Musk acquired Twitter. | Twitter later became X Corp. within X Holdings. |
| 28 March 2025 | xAI acquired X Holdings Corp. and X.AI Corp. through a share exchange. | X and its operating entities became wholly owned subsidiaries of xAI. |
| 2 February 2026 | SpaceX acquired X.AI Holdings Corp. | xAI became a wholly owned SpaceX subsidiary and the foundation of the AI segment. |
| 19 April 2026 | SpaceXAI and Cursor signed compute and collaboration agreements plus an acquisition option. | SpaceX supplied GPU capacity; the parties collaborated on models, including Grok. |
| 16 June 2026 | SpaceX signed the definitive Cursor merger agreement. | The all-stock transaction used SpaceX Class A shares and a $60bn implied Cursor equity value. |
| 14 August 2026 | Cursor announced that SpaceX had completed the acquisition. | Anysphere survived as a wholly owned SpaceX subsidiary; Cursor works with the SpaceXAI team. |
Status at 23 August 2026, 19:23 ICT. “SpaceX bought Cursor; Cursor joined SpaceXAI” is accurate. “xAI bought Cursor” is legally incorrect based on the merger filing and closing announcement.
13
Roadmap and dependencies
Four labels are used consistently: Operational, In development, Announced / planned and Long-term objective. They are not equivalent evidence levels.
| Segment | Operational | Next-stage programme | Long-term objective | Principal dependency |
|---|---|---|---|---|
| Space | Falcon and Dragon launch, cargo and crew. | In development. Starship commercialisation after early payload deployment. | Lunar transport and Mars missions. | Recovery, orbital refuelling, rapid reuse, regulation and cadence. |
| Connectivity | More than 10,200 satellites and 12.0m service lines at 30 June 2026. | In development. Scaled V3, wider Direct to Cell and transport integrations. | More network capacity and optical service for spacecraft. | Launch, production, spectrum, gateways and terminal economics. |
| AI | COLOSSUS, Grok, X and external compute services. | In development. COLOSSUS II, Grok-5 training and Cursor integration. | Larger model, compute and software businesses. | Processors, power, cooling, networking, demand and capital. |
| Orbital AI | No commercial service. | Announced / planned. Gigasat production from late 2027 and deployment from 2028. | Potentially millions of satellites and up to 100GW launched annually. | Starship, production, chips, power, cooling, lasers, safety and economics. |
Roadmap discipline. The 100GW figure and million-satellite scale are long-term objectives. SpaceX describes orbital compute as difficult, early-stage and potentially not commercially viable.
14
Key risks and unresolved questions
- Founder control. Super-voting Class B shares give Musk decisive voting power and majority-board election rights.
- Starship dependency. Important Starlink V3, lunar and orbital-compute plans depend on payload, cadence and rapid-reuse capabilities that are not yet operating at commercial scale.
- Orbital-compute execution. AI1 specifications and timing are company plans. Radiation tolerance, service life, maintenance, heat rejection, optical-network performance, debris management and replacement economics remain unproven at scale.
- Capital intensity. Rockets, constellations and AI data centres all require substantial continuing capital expenditure.
- AI customer concentration. Infrastructure revenue depends on a small number of large customers and contracts that can become terminable after initial periods.
- Power and hardware supply. GPU availability, grid access, gas generation, cooling, water, permitting and high-speed networking constrain AI expansion.
- Regulatory exposure. Launch licences, environmental approvals, spectrum rights, communications rules, AI regulation and X content obligations span many jurisdictions.
- Integration complexity. xAI, X and Cursor bring different products, cultures, liabilities, data regimes and business models into one controlled group.
- Cross-company relationships. Transactions with Tesla and other Musk-related entities require careful separation from SpaceX ownership and economics.
- Fast-changing disclosure. Satellite counts, model versions, compute capacity and legal ownership can change between quarterly filings.
The central analytical question is whether shared infrastructure produces economies of scope greater than the execution cost of operating three different industries. Connectivity supplies the strongest current operating profit. Space supplies the physical deployment advantage. AI is the least mature segment economically and the largest incremental capital commitment.
15
Primary sources
- SpaceX Form 10-Q for the quarter ended 30 June 2026Latest consolidated segment structure, results, operating metrics, revenue mechanics, risks and Cursor merger status before closing.
- SpaceX EU prospectus, approved 5 June 2026Ownership, governance, products, customers, technology, legal history, strategy and evidence boundaries.
- SpaceX Starmind and AI1Published AI1 specifications, modular compute architecture, solar power, radiative cooling, optical links, Gigasat Factory and programme dependencies.
- SpaceX Form S-1 registration statementOrbital AI-compute strategy, sun-synchronous deployment, inference workloads, Starlink integration and expected deployment as early as 2028.
- SpaceX IPO closing announcementFinal share count, proceeds, trading date and SPCX listing.
- xAI joins SpaceXOfficial 2 February 2026 announcement of SpaceX’s acquisition of xAI.
- SpaceX Form 8-K: Cursor merger agreementLegal acquirer, merger structure and $60bn implied equity value.
- Cursor is now part of SpaceXOfficial 14 August 2026 confirmation that the acquisition closed and Cursor joined the SpaceX group.
- Starship Flight 13Official mission record for the first deployment of Starlink V3 satellites.
- SpaceX StarshipCurrent vehicle architecture, payload design and development context.
- Starlink updatesOfficial satellite, safety and next-generation network updates.
Research cut. Facts are current to 23 August 2026, 17:56 ICT. Company forecasts are labelled as targets or plans. Financial figures are historical and are not investment advice.