Note: This article is updated continuously as we work through the full S-1 filing. Last updated: May 27, 2026.
This is a running analysis of the SpaceX S-1 registration statement filed with the SEC on May 20, 2026. We are reading the filing in full and documenting what stands out — what is underreported, what connects to what, and what the numbers actually say when read without the press release framing.
Source: SpaceX S-1 Registration Statement, SEC EDGAR, May 20, 2026.
Read the full filing →THE NUMBERS THAT MATTER
Revenue: $18.7B (2025)
Net loss: $(4.9B) (2025)
Operating cash flow: $6.8B (2025)
Total assets: $102B (March 2026)
Total debt: $29.1B (March 2026)
Accumulated deficit: $41.3B
Starlink subscribers: 10.3M
Starlink ARPU: $66/month (declining)
AI capex Q1 2026: $7.7B in one quarter
Bridge loan: $20B, matures Sept 2027
Elon voting power: 85.1%
Valor leases: $20.2B to board member
Anthropic contract: $1.25B/month
Backlog: $28.4B contracted
THE BUSINESS MODEL (Pages 74-104)
The rocket business loses money. Starlink pays for everything.
Three segments, three very different financial profiles:
Connectivity (Starlink): $11.4 billion revenue in 2025, $4.4 billion operating income, growing 49.8% year over year. 10.3 million subscribers as of March 2026. This is the only profitable segment and the cash engine of the entire company.
Space (rockets): $4.1 billion revenue in 2025, $657 million operating loss. The rocket business exists primarily to support Connectivity and AI — not to generate profit independently. $3 billion of Space segment spend in 2025 was Starship R&D alone.
AI (xAI/Grok/X): $3.2 billion revenue in 2025, $6.4 billion operating loss. $7.7 billion in capital expenditure in Q1 2026 alone. The AI segment is burning cash at a rate that would consume most companies entirely.
The loop: Starlink profits fund Starship development. Starship enables cheap orbital launches. Cheap orbital launches make orbital AI compute economically viable. Orbital AI compute is the actual long-term business. Every dollar Starlink earns is being reinvested into the next layer of the stack.
Total debt as of March 31, 2026: $29.1 billion.
THE THINGS NOBODY IS REPORTING (Pages 145-153)
Macrohard: Built jointly with Tesla, Macrohard is an agentic AI platform designed to fully emulate digital workflows — from coding to management to entire business processes — using autonomous agents. The S-1 says it will "fundamentally transform how companies are structured." One mention in the prospectus summary. Almost zero coverage anywhere.
Terafab: SpaceX plus Tesla plus Intel building the world's largest chip manufacturing facility. Target: one terawatt of compute output per year. Goal: break the Nvidia dependency. Barely reported.
Orbital AI compute satellites: Deployment starting as early as 2028. Powered by solar energy in space. Connected via the existing Starlink inter-satellite laser network. The S-1 describes a goal of deploying 100 gigawatts of compute to orbit annually — requiring thousands of Starship launches per year and one million metric tons to orbit annually. Current global launch capacity is approximately 2,000 metric tons per year. They are describing a 500x increase.
Lunar mass driver: Defined in the S-1 glossary in a single sentence — "a launch system on the Moon's surface that uses electromagnetic acceleration to propel payloads into space without the use of rockets." Never elaborated on. Factories on the Moon manufacturing satellites, launched electromagnetically into orbit. No rocket fuel required.
The Anthropic contract: $1.25 billion per month through May 2029 for access to Colossus compute. $15 billion per year. Claude runs on SpaceX hardware.
THE HONEST RISK FACTORS (Pages 26-46)
Single point of failure: Everything depends on Starship. V3 satellites, orbital AI compute, lunar operations, Mars — all contingent on Starship achieving full reusability at scale. The S-1 states directly: "Our ability to execute our growth strategy is highly dependent on Starship."
One million satellites: Orbital AI compute plans require operating potentially up to one million satellites. Current Starlink is 9,600. Regulatory coordination alone for this scale is explicitly listed as a material risk.
Orbital AI compute is untested: Direct quote from the S-1: "We have not, and no one else has, previously operated or attempted to operate orbital AI compute, and the conditions of space on such AI infrastructure have not been tested."
Active investigations: SpaceX is under active investigation by the Irish Data Protection Commission regarding Grok's processing of EU citizen data including children. The FTC is separately investigating Grok's chatbot safety around minors.
The Brazil precedent: Brazil's Supreme Court froze Starlink's Brazilian financial assets in 2024 due to actions taken by X — before SpaceX even owned X. SpaceX warns this could happen again in other countries based on Elon's affiliations.
The nuclear wildcard: If breakthrough nuclear energy reduces terrestrial energy costs significantly, the entire orbital AI compute business case weakens. Explicitly acknowledged.
In-orbit refueling is untested: Required for lunar and Mars missions. Direct quote: "In-orbit refueling is complex, and we have not yet demonstrated or attempted it."
20% of revenue is US government: Political risk is real and explicitly disclosed. A change in administration or congressional priorities could materially impact the business.
Chinese competition: SpaceX explicitly names Chinese competitors benefiting from government support as a material competitive risk to the orbital compute strategy.
THE NUMBERS THAT STAND OUT (Pages 1-14, 74-104)
75% — SpaceX satellites represent approximately 75% of all active maneuverable satellites in orbit.
500x — The increase in global launch capacity required to achieve the orbital AI compute vision.
$26.5 trillion — SpaceX's estimate of the total addressable market for AI.
Kardashev Type II — The explicit stated long-term goal in the S-1. A civilization that harnesses all energy output of its star. This is in a legally binding SEC document.
99.8% — The percentage of the solar system's energy contained in the Sun. SpaceX cites this as the justification for orbital compute.
350 million — Daily posts on X that Grok has exclusive real-time access to. This is treated as a competitive moat for AI training.
WHAT CONNECTS TO WHAT
Starlink → funds Starship R&D
Starship → enables cheap orbital launches
Cheap launches → orbital AI compute viable
Orbital AI → powered by solar energy
Solar energy → eliminates terrestrial power constraints
No power constraints → unlimited AI compute scale
X real-time data → trains Grok
Grok → powers Macrohard
Macrohard → replaces entire business departments
Terafab → breaks Nvidia dependency
Lunar factories → manufacture satellites
Lunar mass driver → launches them without rockets
Every piece connects. Nothing is random.
THE FULL FINANCIAL PICTURE (Pages 21-25)
Revenue growth: $10.4B (2023) → $14.0B (2024) → $18.7B (2025) → $4.7B Q1 2026.
Net loss: $(4.6B) 2023, $791M profit 2024 — the only profitable year — $(4.9B) 2025, $(4.3B) Q1 2026.
Operating cash flow is positive and growing: $4.5B → $5.8B → $6.8B. The business generates real cash. The losses are from investment spend, not operational failure.
Total assets: $57B (2024) → $92B (2025) → $102B (March 2026).
Cash dropped $9B in Q1 2026 alone — that is the AI capex burn in action.
Starlink subscribers: 2.3M (2023) → 4.4M (2024) → 8.9M (2025) → 10.3M (March 2026). ARPU declining: $99 → $91 → $81 → $66 as international expansion continues. More subscribers, lower revenue per subscriber.
THE TAM BREAKDOWN (Pages 11-14)
$28.5 trillion total addressable market.
Space: $370 billion
Connectivity: $1.6 trillion
AI infrastructure: $2.4 trillion
Consumer AI subscriptions: $760 billion
Digital advertising: $600 billion
Enterprise AI applications: $22.7 trillion
Enterprise applications alone represent 80% of the total TAM. That is Macrohard territory. The rockets, satellites, and data centers are infrastructure for a software business worth $22.7 trillion.
China and Russia are explicitly excluded from all TAM estimates.
THE GOVERNANCE DETAILS WORTH NOTING (Pages 14-20)
Mandatory arbitration: Shareholders are required to use mandatory arbitration rather than courts. An unusually aggressive governance choice for a company going public.
All share counts are blank: Every voting percentage, share count, and IPO price is blank. Preliminary filing.
Use of proceeds priority order: AI compute infrastructure first, then launch infrastructure, then satellite constellations.
DIGITAL HUMAN AUGMENTATION
Appears multiple times in the filing listed as both a growth strategy and a key challenge involving unproven technologies or technologies that do not exist. No definition provided anywhere in sections analyzed so far.
THE FINANCIAL SUMMARY IN FULL (Pages 21-25)
Revenue: $10.4B (2023) → $14.0B (2024) → $18.7B (2025) → $4.7B Q1 2026.
Net loss: $(4.6B) 2023, $791M profit 2024 — the only profitable year in company history — $(4.9B) 2025, $(4.3B) Q1 2026.
Operating cash flow positive and growing: $4.5B → $5.8B → $6.8B. The losses are from investment spend, not operational failure.
Total assets: $57B (2024) → $92B (2025) → $102B (March 2026).
Cash dropped $9B in Q1 2026 alone.
Accumulated deficit since founding: $41.3 billion.
Starlink subscribers: 2.3M (2023) → 4.4M (2024) → 8.9M (2025) → 10.3M (March 2026). ARPU declining from $99 to $66 as international expansion continues.
THE DEBT STRUCTURE (Pages 68-71)
Total debt: $29.1 billion.
The most important single number: $20 billion bridge loan maturing September 2, 2027. This is a hard deadline. SpaceX needs to retire or refinance $20 billion within roughly 15 months of the filing date. The IPO proceeds are the most likely source of repayment. Investors buying SPCX are in part funding the retirement of this bridge loan.
Additional debt includes $9.1 billion in other financings related to AI infrastructure assets recorded as failed sale-leaseback transactions.
THE ELON DEPENDENCY (Pages 57-60)
SpaceX does not maintain key-person life insurance on Elon Musk. The filing dedicates an entire section to the risk of losing him — death, disability, or unwillingness to continue. He currently splits time between SpaceX, Tesla, Neuralink, The Boring Company, and previously served as Senior Advisor to the President of the United States.
The charter explicitly states Elon can direct business opportunities to his other companies without any obligation to offer them to SpaceX first. Macrohard and Terafab — two of the most strategically important initiatives in the filing — involve Tesla, and their financial terms have not been finalized. Elon negotiates both sides of those deals.
THE SHAREHOLDER RIGHTS STRUCTURE (Pages 61-63)
Class A shareholders have essentially no ability to challenge management:
No jury trials — waived entirely in the bylaws.
No class action lawsuits — prohibited entirely.
3% ownership threshold to bring any derivative claim, held continuously for six months.
67% of voting power required to submit a shareholder proposal.
All disputes go to Texas Business Court, then mandatory arbitration if that court declines jurisdiction.
Company can consolidate cases. Shareholders cannot.
This is the most aggressive anti-shareholder rights structure in any major US IPO filing reviewed in recent memory.
THE LAUNCH MIX NOBODY IS TALKING ABOUT (Pages 86-87)
In 2025 SpaceX completed 165 Falcon launches. Only 43 were customer launches. 122 were internal — launching their own Starlink satellites. 74% of all SpaceX launches in 2025 were for their own constellation, not paying customers.
In Q1 2026 it got more extreme — 40 launches, only 7 were customer launches, 33 were internal. 82.5% of all launches going to their own payloads.
The rocket business as a customer service is actually quite small relative to total launch activity. SpaceX is primarily a satellite deployment operation that also sells some excess capacity to outside customers.
AGI STATED AS BELIEF IN SEC FILING (Page 83)
Buried in the AI segment description is this statement: "We believe AI is rapidly converging toward AGI, where human cognitive capabilities can be replicated and scaled at machine speeds, profoundly augmenting human productivity. Once an AGI system exists, its true value derives from the ability to create limitless duplicates of human-like intelligence."
That is SpaceX's lawyers putting AGI convergence belief in a legally binding SEC document. Not a tweet. Not a keynote. An SEC filing.
THE ADDITIONAL FINANCIAL OBLIGATIONS (Pages 94-100)
EchoStar spectrum acquisition: $19.6 billion in equity and cash consideration. Closing expected November 2027. This is on top of the existing $29.1 billion in debt and comes due at roughly the same time as the $20 billion bridge loan maturity.
The paid subscriber conversion rate: 6.3 million paid subscribers out of 550 million MAUs. That is a 1.1% conversion rate. Enormous growth runway or structural limitation depending on how you read it.
Terminal production: 200,000 Starlink terminals per week — over 10 million per year. More than current subscriber count, building inventory for growth.
THE DETAILED FINANCIAL TABLES (Pages 102-110)
GPU depreciation alone: $908 million in Q1 2026. In one quarter. That is how fast the hardware is being written down. The AI segment is not just spending on infrastructure — it is also taking enormous depreciation charges as that infrastructure ages.
The "One Big Beautiful Bill" impact: SpaceX reversed $659 million in deferred tax assets in 2025 due to the enactment of the One Big Beautiful Bill Act. This directly worsened the 2025 net loss. A legislative change had a measurable nine-figure impact on SpaceX's financials.
X advertising revenue declined in Q1 2026. The S-1 explains it was due to an overhaul of the advertising platform that disrupted ad sales during the rebuild. The platform SpaceX is counting on for AI distribution and revenue is still rebuilding its core business.
Government connectivity revenue declined $175 million in Q1 2026. This is a meaningful decline — roughly 20% of the enterprise and government connectivity bucket. Worth watching as a trend.
Falcon 9 reusability saved $240 million in cost of revenue in 2025. That is a real measurable financial benefit of booster reuse showing up directly in the income statement.
Connectivity margin expansion: Connectivity revenue grew 49.8% in 2025. Cost of revenue only grew 24.2%. Operating income more than doubled on 50% revenue growth. The economics of the satellite business are genuinely improving at scale.
Starship cost accounting shift: Right now every Starship test flight cost is expensed immediately to R&D. Once Starship is commercialized, costs shift to capitalized property depreciated over time. Reported losses will shrink dramatically at that point even if total spending does not change.
THE DEBT STRUCTURE IN FULL (Pages 122-130)
The most important sentence in the entire debt section — the Company is required to use an amount equal to the net proceeds of a qualified initial public offering, including this offering, to repay amounts outstanding under the SpaceX Bridge Loan within six months following receipt of those proceeds.
This is not optional. IPO investors are legally funding the retirement of $20 billion in bridge debt.
The bridge loan can be extended twice at 0.25% fee per extension. Final extended maturity is March 2028 — not September 2027 as the base maturity suggests.
The bridge loan contains a prohibition on disposing of Starlink assets outside the consolidated group. Starlink is explicitly ring-fenced. SpaceX cannot sell or spin off Starlink without triggering default.
The EchoStar deal structure: $11.1 billion in equity at a fixed price of $42.40 per share — issuing 261.8 million new Class A shares. Up to $8.5 billion in cash. This fixed share price of $42.40 implies a valuation floor for the IPO price.
Bridge loan effective interest rate: 4.58% as of March 31, 2026. On $20 billion that is approximately $916 million in annual interest expense from the bridge loan alone.
Satellite useful life accounting sensitivity: if useful life estimates change by one year it would have a $480 million impact on annual operating income. The financials are highly sensitive to accounting assumptions about satellites.
THE CURSOR DETAILS (Pages 145-147)
The Cursor acquisition walk-away cost is $10 billion total — $1.5 billion termination fee plus $8.5 billion deferred services fee. SpaceX is effectively locked into either acquiring Cursor at $60 billion or paying $10 billion to exit.
Cursor's balance sheet as of January 31, 2026: $3.1 billion total assets, $2.7 billion of which is cash. $0.55 billion in liabilities. SpaceX would primarily be acquiring cash and intellectual property.
TERAFAB CHIP TYPES CONFIRMED (Pages 145-155)
The S-1 now confirms two distinct chip types being designed at Terafab:
Type 1: Optimized for terrestrial edge and inference — used primarily in Tesla's Optimus robots and vehicles.
Type 2: Optimized for the space environment — used in orbital AI compute infrastructure.
This is the first explicit confirmation that Terafab chips are designed for Tesla Optimus robots as well as SpaceX satellites.
RECURSIVE SELF-IMPROVING AI (Page 149)
The S-1 references "recursive self-improving learning that minimizes human intervention" as an emerging workload that is "highly token consumptive" and that SpaceX expects to support infrastructure for. This is a reference to AI self-improvement capability stated in a legally binding SEC document.
THE FALCON 9 REUSE RECORD (Page 139)
As of March 31, 2026, a Falcon 9 first stage has been reflown 34 times. The same booster has launched to space and returned 34 times. This is a remarkable engineering achievement that directly explains the cost reduction trajectory in the Space segment financials.
Dragon has flown 78 crewmembers from 20 countries since 2020.
HUMAN EXTINCTION IN THE BUSINESS SECTION (Page 136)
The business section — not the risk factors, not a press release — contains this statement: "Geological and astronomical records indicate a non-zero probability of extinction-level events occurring over periods measurable in millions of years. Reliance on a single planetary home constitutes a single point of failure and carries existential risk with a probability of one that must be solved."
SpaceX's lawyers wrote about human extinction probability in the business justification section of an SEC filing. This is the stated reason the company exists.
THE INDUSTRY NUMBERS THAT MATTER (Pages 155-160)
$7 trillion: Global data center investment needed through 2030 to meet projected AI demand. Generative AI expected to account for 70% of data center power demand by end of decade.
585 exabytes: Data generated per day globally in 2025. Up from 10.8 exabytes per day in 2015. A 54x increase in a decade.
80%: The percentage of Earth's land mass with no terrestrial network coverage. That is Starlink's actual addressable market.
$13.7 billion: National Security Space Launch Phase 3 contracts already awarded to SpaceX through 2032. 54 missions. Committed government revenue.
THE LUNAR ENDGAME IS BIGGER THAN REPORTED (Pages 165-166)
The S-1 describes a petawatt-scale AI constellation as the long-term goal enabled by lunar operations. Petawatt is 1,000 terawatts — orders of magnitude beyond the 100 gigawatt orbital compute goal that headlines have focused on.
The Moon contains rare materials estimated at over one million tons with applications in nuclear energy and quantum computing. SpaceX explicitly plans to mine and export these to Earth using Starship.
The lunar mass driver is confirmed as the launch mechanism for this manufacturing economy — high frequency, low cost launches of satellites from the lunar surface using electromagnetic acceleration.
MACROHARD IS GROK PLUS OPTIMUS ROBOTS (Pages 170-171)
The most complete description of Macrohard in the filing:
"Macrohard aims to combine our frontier AI model with Tesla's physical AI prowess to achieve the goal of augmenting the operational functions of entire companies."
Tesla's physical AI means Optimus robots. Macrohard is not just software automation. It is the combination of Grok AI models and Tesla Optimus robots running entire companies. This has not been reported anywhere.
THE TALENT AND CULTURE NUMBERS (Page 164)
SpaceX accepted under 2% of engineering applicants in 2025.
Average tenure across SpaceX leadership team: 12 years.
80% of Fortune 500 companies were using AI active agents as of February 2026 — the enterprise market Macrohard is targeting.
The human effort needed to sustain Moore's Law has increased 18-fold since the early 1970s. SpaceX explicitly argues AI augmentation could reverse this trend — this is the stated justification for digital human augmentation as a business.
THE PLASER PROGRAM (Pages 175-176)
SpaceX is already monetizing its 23,000 inter-satellite laser network through a program called Plaser. Third-party satellite operators can purchase Starlink laser hardware and connect their satellites to the Starlink mesh network, offloading data to ground stations anywhere on Earth without building their own relay architecture.
This is a new revenue stream that has not appeared in previous sections and has received almost no coverage. As satellite constellations grow globally, demand for this in-orbit data transport layer is expected to increase significantly.
THE STARSHIP PROPELLANT TRANSFER MILESTONE (Page 183)
The risk factors listed in-orbit refueling as untested. That was technically accurate — full-scale transfer has not been attempted. But the S-1 business section reveals SpaceX has already successfully transferred approximately 5 metric tons of cryogenic propellant between tanks while in space — described as a first of its kind operation.
The small-scale test has been done. Full-scale in-orbit refueling remains the remaining challenge.
THE FALCON 9 ACCOUNTING DETAIL (Pages 181-182)
Falcon 9 boosters are qualified for up to 40 flights based on engineering data. SpaceX uses a maximum accounting useful life of only 25 flights. Two reasons:
1. Strategic transition to Starship is expected to reduce future Falcon 9 demand before many boosters reach 40 flights.
2. Certain government contracts prohibit the use of boosters flown more than 5 times on their missions.
The financial depreciation assumes earlier retirement than the hardware actually supports — which means Falcon 9 cost of revenue is higher than it needs to be on a purely engineering basis.
THE V3 SATELLITE CAPACITY NUMBERS (Page 167)
Current Starlink constellation: over 700 Tbps cumulative downlink capacity across 9,600 satellites.
V3 satellites: 1 Tbps per satellite. 60 V3 satellites per Starship launch. One Starship launch = 60 Tbps added. One Falcon 9 launch = 3 Tbps added.
The capacity per launch increases 20-fold with Starship and V3. This is the step-change in Starlink capacity economics that the S-1 references repeatedly.
FALCON HEAVY PLANETARY MISSIONS (Page 182)
Falcon Heavy has been selected by NASA to launch Europa Clipper (Jupiter moon) and Dragonfly (Saturn moon Titan) — two of NASA's flagship planetary science missions. Confirmed government contracts for deep space exploration.
DRAGON IS THE ONLY RETURN VEHICLE (Pages 187-188)
Dragon cargo spacecraft is the only spacecraft in operation capable of returning significant cargo from the International Space Station to Earth. Return payload capacity: 3,000 kg. This is a genuine monopoly capability with no competitor currently in operation.
THE FULL COMPETITOR LIST (Pages 220-221)
For the first time in the filing SpaceX names every competitor across all three segments.
Space competitors: United Launch Alliance (Boeing and Lockheed Martin joint venture), Arianespace, Northrop Grumman, Blue Origin, Rocket Lab, Firefly Aerospace, Relativity Space.
Connectivity competitors: Verizon, Comcast, AT&T, T-Mobile, Lumen, Charter, Google Fiber, BT, Deutsche Telekom, Liberty Global, EchoStar, SES, Telesat, Viasat, Amazon LEO, Blue Origin TeraWave, Eutelsat OneWeb, Iridium, AST SpaceMobile, Lynk, Globalstar, Skylo.
AI competitors: OpenAI, Anthropic, Google, Meta, Microsoft, open source model providers, Threads, Reddit, TikTok. As SpaceX scales compute infrastructure it also expects to compete with AI cloud providers CoreWeave and Nebius as well as hyperscalers.
Notable: Blue Origin appears in both Space and Connectivity competitor lists — Blue Origin's TeraWave satellite constellation is explicitly named as a Starlink competitor.
THE MANUFACTURING NUMBERS (Pages 205-206)
Redmond satellite factory: 70 satellites per week average from December 2025 to April 2026. Approximately 3,640 satellites per year at full rate from one facility.
Bastrop facility: produces tens of thousands of Starlink Kits per day. In 2026 SpaceX expects to more than double the size and add production of gateway antennas, solar cells, and AI compute satellites. The AI compute satellite manufacturing line is being added to an existing Starlink hardware factory.
Starlink manufacturing cost: 9x reduction in cost per Gbps from V1 to V3 satellites. Terminal manufacturing cost down 59% since 2022.
THE LAUNCH INFRASTRUCTURE EXPANSION (Pages 207-209)
Four operational Starship launch pads planned by end of 2027: Two at Starbase, Texas. Two at Cape Canaveral, Florida.
Currently one operational Starship pad exists. Going from 1 to 4 in roughly 18 months.
Gigabay being built at both Starbase and Kennedy Space Center. Supports vehicles up to 85 meters tall with 24 work cells and cranes capable of lifting up to 400 tons.
McGregor, Texas is described as the most active rocket development and testing facility in the world with 15 specialized test stands.
400+ Starlink ground station sites globally.
SATELLITE USEFUL LIFE AND PLASER (Pages 194-196)
Starlink satellites have estimated useful lives of 3 to 5 years. SpaceX often deorbits them earlier than useful life to maintain autonomous collision avoidance system performance. The actual replacement cycle may be faster than accounting estimates suggest.
The Plaser program: SpaceX is already monetizing its 23,000 inter-satellite laser network by allowing third-party satellite operators to purchase Starlink laser hardware and plug their satellites into the Starlink mesh network — offloading data to ground stations anywhere on Earth without building their own relay architecture. This is an unreported new revenue stream.
GROK COMPETITORS NAMED AND GROKIPEDIA (Pages 195-200)
The S-1 explicitly names Grok's frontier AI competitors: OpenAI, Anthropic, and Google. These are the only three named in the frontier model section.
A new unreported product called Grokipedia is listed as a consumer product powered by Grok. No description beyond the name appears in the filing.
THE BOARD AND MANAGEMENT DETAILS (Pages 226-231)
SpaceX has 22,000+ full-time employees as of March 31, 2026. None subject to collective bargaining agreements.
Board composition: 8 directors total. Elon, Shotwell, Gracias, Harrison, and Nosek are Class B Directors elected by Elon's shares. Ehrenpreis, Glein, and Jurvetson are Common Stock Directors elected by all shareholders.
Google's Donald Harrison — President of Global Partnerships and Corporate Development at Google — sits on SpaceX's board. He has been a director since 2015.
Luke Nosek previously served on DeepMind's board before Google acquired it.
THE LEGAL DISCLOSURES WORTH NOTING (Pages 228-229)
Elon's 2018 SEC settlement: Paid $20 million civil penalty over the Tesla "funding secured" tweet. Did not admit wrongdoing. Required to step down as Tesla Chairman for three years.
Twitter securities fraud verdict: On March 20, 2026 — weeks before the S-1 filing — a jury found Elon Musk violated securities law in connection with statements made during his Twitter acquisition in May 2022. Partial judgment entered April 3, 2026. Motion for judgment as a matter of law pending as of filing date. This is active litigation against the CEO of the company going public.
ELON'S ACTUAL COMPENSATION (Pages 233-239)
Elon Musk's base salary: $54,080 per year. Unchanged since 2019. No bonus. No equity grants in 2025. The CEO of a company filing a $200+ billion IPO earns $54,080 in cash annually — roughly California minimum wage for an exempt employee.
His actual compensation is two performance equity awards:
Award 1: 1 billion restricted Class B shares granted January 2026. Vests in 15 tranches tied to market cap milestones from $500 billion to $7.5 trillion. But there is a second condition that must also be met for any tranche to vest: establishment of a permanent human colony on Mars with at least one million inhabitants. Both conditions required. No Mars colony — no shares vest regardless of market cap.
Award 2: 302 million restricted Class B shares. Vests upon market cap milestones from $1.065 trillion to $6.565 trillion AND completion of non-Earth-based data centers capable of delivering 100 terawatts of compute per year. Again both conditions required simultaneously.
Elon's compensation is legally tied to a Mars colony and terawatt-scale orbital compute. These are not aspirational goals. They are compensation conditions in a legally binding SEC document.
THE OTHER EXECUTIVES
Gwynne Shotwell 2025: $1.08M salary, $82.9M in option awards, total $85.8M.
Bret Johnsen 2025: $825K salary, $9M in option awards, total $9.8M.
CFO performance milestones: Johnsen has 4 million stock options tied to achieving $10 billion in adjusted EBITDA annually. None vested in 2025. The target was not met.
Internal share price signals: Options granted at $37.00 per share in May 2025 and $42.40 per share in October 2025. The EchoStar deal is also priced at $42.40 per share. These are the clearest available signals of management's view of fair value heading into the IPO.
THE VALOR EQUIPMENT LEASES (Pages 244-245)
Antonio Gracias sits on SpaceX's board. His firm Valor Equity Partners has three separate GPU equipment lease agreements with xAI subsidiaries totaling $20.2 billion in aggregate payments:
Lease 1: $6.986 billion
Lease 2: $6.633 billion
Lease 3: $6.587 billion
SpaceX guarantees all three leases. Valor collected $885 million in 2025 and $857 million in just the first two months of 2026.
A board member's firm is receiving over $20 billion from the company he helps govern. This is disclosed in the related person transactions section. It has received essentially no coverage.
THE UNDERWRITER CONFLICTS (Pages 264-268)
Goldman Sachs, Morgan Stanley, BofA, Citi, and JPMorgan are the five lead underwriters for the SPCX IPO. All five are also lenders or administrative agents under the $20 billion SpaceX Bridge Loan. The banks selling IPO shares to the public are the same banks that lent SpaceX the money that IPO proceeds are legally required to repay.
Barclays, Deutsche Bank, RBC, UBS, and Wells Fargo are also bridge loan lenders and junior underwriters.
Morgan Stanley specifically advised SpaceX on the acquisition of xAI and is now a lead underwriter for the IPO. One bank on both sides of the deal.
Retail platforms: Charles Schwab, Fidelity, Robinhood, SoFi, and E*TRADE are designated selling group members for retail investors.
THE LOCK-UP STRUCTURE (Pages 266-267)
Elon Musk: 366-day lock-up. No early release provisions. He cannot sell a single share for one full year regardless of stock performance.
Other early investors: 180-day base lock-up with tiered early release:
- 20% released after Q2 2026 earnings
- Additional 10% if stock is 30% above IPO price at that point
- 7% every 15 days thereafter
- 28% after Q3 2026 earnings
- Remainder at 180 days
These tiered releases create predictable selling pressure windows that retail investors should understand before buying.
THE FINANCIAL STATEMENTS — KEY NUMBERS (Pages F-1 to F-96)
The balance sheet as of March 31, 2026:
Total assets: $102.1 billion
Cash and marketable securities: $23.7 billion combined
Property, plant, and equipment: $53.9 billion net
Servers and networking equipment alone: $23.9 billion — up from $6.9 billion a year earlier. That is a $17 billion increase in GPU hardware in twelve months.
Construction in progress: $14 billion at March 31, 2026 — up from $4.6 billion at year end. The build rate is accelerating.
Total debt: $29.1 billion as of March 31, 2026, essentially all in the $20 billion bridge loan plus $9.1 billion in Valor equipment leases.
Accumulated deficit: $41.3 billion since founding.
THE BACKLOG (Financial Statement Notes)
Contracted backlog as of December 31, 2025: $28.4 billion. This is revenue that is under contract but not yet recognized. 32% expected within one year. 53% in 2027 and 2028. The forward revenue picture is strong even before new contracts are signed.
THE CUSTOMER CONCENTRATION RISK (Note 3 — Revenue)
A single customer — identified only as Customer A — represented 20.9% of 2025 revenue, 24.2% of 2024 revenue, and 25.2% of 2023 revenue. Revenue from this customer spans all three segments. This is almost certainly the US Government. One unnamed customer accounts for over $3.9 billion in 2025 revenue. No other customer exceeded 10%.
THE ELON CEO AWARD VALUATION (Note 14 — Share-based Compensation)
The 1 billion Mars colony shares were valued at $90.40 to $95.92 per share per tranche using a Monte Carlo simulation. The 302 million orbital compute shares were valued at $91.47 to $95.92 per share per tranche. Combined across 1.302 billion total restricted shares that implies approximately $120 billion in total compensation value at grant — contingent entirely on a Mars colony and terawatt-scale orbital compute. No Mars colony, no shares vest.
THE NAACP LAWSUIT AGAINST COLOSSUS II (Note 16 — Commitments)
On April 14, 2026 — five weeks before the S-1 was filed — the NAACP filed suit against xAI and MZX Tech alleging the mobile gas turbines powering the COLOSSUS II data center in Southaven, Mississippi violate the Clean Air Act because they constitute stationary sources without proper permits. On May 6, 2026 the NAACP filed a preliminary injunction motion seeking to shut down the turbines entirely. If granted this could directly disrupt AI training operations at SpaceX's primary compute facility. This is an active litigation risk filed two weeks before Elon Musk signed the S-1.
THE ANTHROPIC DEAL (Note 22 — Subsequent Events)
On May 3, 2026 — seventeen days before the S-1 was filed — SpaceX entered a cloud services agreement with Anthropic PBC for access to compute capacity at COLOSSUS and COLOSSUS II. Anthropic agreed to pay a monthly fee through May 2029, with capacity ramping in May and June 2026. Either party can terminate on 90 days notice. The business section of the S-1 confirms the monthly fee is $1.25 billion — meaning Anthropic has agreed to pay SpaceX up to $45 billion over three years for compute access. This is the largest compute contract in the filing and it was signed seventeen days before IPO.
THE TECHNICAL DEFAULT NOBODY REPORTED (Note 9 — Debt)
When SpaceX acquired xAI on February 2, 2026 it triggered a technical default on the SpaceX Credit Facility due to debt levels assumed at the xAI subsidiary level. SpaceX had to obtain an emergency waiver from its bank syndicate on March 2, 2026 to cure the default and allow the bridge loan refinancing to proceed. This happened six weeks before the S-1 was filed and has received essentially no coverage.
THE OWNERSHIP TABLE (Pages 247-249)
Elon: 849 million Class A shares (12.3%) and 5.569 billion Class B shares (93.6% of all Class B). Combined voting power: 85.1% before the IPO.
Antonio Gracias through Valor entities: 503 million Class A shares — 7.3% of Class A. The second largest individual shareholder after Elon.
Google is listed as an investor with registration rights under the Investors' Rights Agreement alongside Valor, DFJ Growth, and entities affiliated with Elon.
Class B shares permanently restricted: The charter explicitly states additional Class B shares may only be issued in the future to Elon Musk, his family members, and certain permitted entities. Nobody else can ever accumulate Class B shares. The 10-to-1 voting advantage is permanently and exclusively Elon's.
THE RELATED PARTY TRANSACTIONS (Pages 243-246)
The Boring Company tunnels at Bastrop: SpaceX paid The Boring Company $1 million in 2025 for tunnel construction at the Bastrop Starlink terminal factory. Elon's tunnel company is building infrastructure at SpaceX's main manufacturing facility.
SpaceX pays Elon's security company: SpaceX pays a security company owned by Elon Musk $4 million per year for security services related to Elon's work at SpaceX. SpaceX pays the company owned by the person being protected.
Tesla transactions growing: SpaceX obtained goods and services from Tesla of $144 million in 2025 — up from $4 million in 2024 and $11 million in 2023. The Tesla relationship is accelerating rapidly as Macrohard and Terafab develop.
ANALYSIS COMPLETE
All 277 pages of the SpaceX S-1 registration statement filed with the SEC on May 20, 2026 have been analyzed and documented on this page. Source: