The framework is designed to see the structure behind the noise — to identify the patterns that drive a system, to forecast its trajectory, and to detect the inflection points where it will change. SpaceX’s pending IPO is a real, consequential, contemporary system. The framework has been applied to it.

This page is the public-facing summary. For the methodology in plain English, see The 3x3 framework, explained →. To see the methodology applied to a system of your own, get in touch →.

Analysis date: 2026-06-03.
Composite confidence score: 0.76.
IPO listing: 2026-06-12 (NASDAQ: SPCX) at $135/share — 555.6M shares, ~$75B proceeds, ~$1.75T valuation.
Roadshow: June 4–11. Anchor sessions June 4–6. Retail orders open June 7 (Robinhood, Fidelity, Schwab). Pricing June 11. First trade June 12.
Status: Demonstration of the methodology. Not investment advice.

The headline finding

The IPO at $135/share prices SpaceX at approximately 91–94× 2025 revenue — well above NVIDIA’s peak multiple of ~40×, and NVIDIA is currently profitable. The $1.75T prices a 10-to-15-year terminal state as a present-day certainty. The framework’s analysis identifies that the price sits on seven sequential unproven technology dependencies, in a structure where 78% of IPO proceeds are pre-committed, and where serial dilutive fundraising is explicitly disclosed in the S-1. The probability-weighted expected value of the consolidated company — the center of mass of the three scenarios below — is approximately $725-925B.

The gap between that expected value and the IPO price is not a flaw in the technology. The technology thesis is sound. The gap is the present value of the seven sequential dependencies, priced as near-certain by the IPO narrative and as improbable by the framework’s structural reading.

What the framework found: three business lines at three different states of maturity

SpaceX is not one business. It is three businesses, at three very different states of maturity, consolidated under one $1.75T valuation. The framework’s analysis treats them separately.

The only currently profitable GAAP segment. The framework’s reading of the segment’s state:

  • Average revenue per user is declining structurally. $99/month in 2023, $66/month in Q1 2026, with management guiding further decline. The long-run equilibrium is estimated at $45-55/month — the consumer reference price of $100/month for unlimited 1 Gbps fiber sets a hard ceiling on premium-tier pricing.
  • There is no viable recovery mechanism before V3. V3 deployment requires Starship, which is currently grounded. The dependency is structural, not tactical.
  • Cash flow is negative despite 63% EBITDA margin. Satellite replacement capex runs $7-10B/year on Falcon 9. The depreciation represents a real, mandatory replacement cost, not an optional deferral. The framework distinguishes EBITDA from FCF, because conflating them is exactly the kind of error that produces structural mispricings.

The second business: Starship

The technology thesis is sound. The framework’s reading of the segment’s state is that execution is behind:

  • Flight 12 (May 22, 2026): Ship 39 succeeded (V3 was validated). Booster 19 was lost in a Raptor 3 cascade failure. The cascade failure affects both vehicle stages — a systemic concern, not an isolated incident.
  • Currently grounded. FAA mishap investigation is active. Earliest Flight 13: July-August 2026.
  • On-orbit propellant transfer — required for HLS, Mars, and orbital compute — has never been demonstrated at any scale.
  • Full reusability with 24-hour turnaround is on a Musk-stated 6-to-7-year timeline.

The third business: AI / SpaceXAI

Four sub-businesses at very different maturity stages, all presented as one addressable market:

Sub-business State of maturity Status
Colossus leasing Mature — but confirmed near-term exit Real revenue. Anthropic ~$1.25B/month for ~325,000 Nvidia GPUs. The S-1/A amendment (June 1) confirmed the exit terms in the company’s own language: revenue “depends on continuation terms rather than being fixed over a longer period.” Either party can terminate after an initial 3-month period with 90 days’ notice. The $45B headline value is not a revenue commitment — it is a capacity lease that either party can exit from July 2026 onward. Anthropic’s own S-1 (filed June 1 at ~$965B valuation, $47B ARR) gives it the capital and scale to build its own data centers. Practical earliest exit: Q1 2027 (Anthropic’s IPO quiet period ends ~October 2026; 90-day notice = Q1 2027). SpaceX collects approximately $4–5B in Colossus revenue before termination — a narrow runway, not a durable floor.
Grok Speculative No established competitive position. Open-source models are releasing monthly at frontier capability.
Orbital compute Concept Concept only. Blocked behind Starship and a thermal engineering gap.
Terafab Physically impossible on stated timeline Intel 14A is not at volume before 2029; ASML EUV supply is globally constrained.

The framework’s finding is that consolidating these four sub-businesses at one valuation erases the structural difference between a real revenue stream (Colossus) and a concept (Terafab) that the underlying physics will not support on the stated timeline.

The Colossus cancellation risk has a quantifiable magnitude. An analysis of the six organizations with sufficient scale to replace Anthropic as a tenant — Google DeepMind, Meta AI, Microsoft/OpenAI, Apple, AWS, and government customers — finds that all are building their own infrastructure or operate multi-cloud strategies that make exclusive Colossus tenancy unnecessary. Post-Anthropic Colossus revenue is estimated at $1–3B/year, down from $15B. The AI division’s contribution to the $1.75T valuation rests almost entirely on a single contract that is near-certain to exit within 12–18 months of IPO.

The seven sequential dependencies

The $1.75T valuation requires all seven of the following to succeed in order:

  1. Raptor 3 fix demonstrated on booster.
  2. V3 Starlink deployment at commercial density (Starship operational).
  3. Starship full reusability achieving commercial cadence.
  4. On-orbit propellant transfer demonstrated at operational scale.
  5. Artemis HLS crewed mission (government revenue milestone).
  6. Orbital compute satellites deployed and revenue-generating.
  7. Terafab achieving sub-5nm production at commercial volume.

Each is a gate. Failure of any one blocks all downstream value creation. The composite probability that all seven succeed within a valuation-relevant timeframe is low — the dependencies are multiplicative, not additive.

The three scenarios

The framework does not produce a single forecast. It produces a set of scenarios with probability weights, anchored to whether the Starship operational milestone is crossed successfully and when:

Scenario Outcome Timeline Probability
Positive Milestone crossed, on schedule 2027-2028 30%
Base Milestone crossed, delayed 2029-2030 45%
Negative Milestone fails or significantly delayed 2031+ or never 25%

The shift from prior probabilities to the current values was driven by five inputs between May 28 and June 2, with additional structural findings from the June 3 update:

  • Blue Origin New Glenn explosion (May 28) eliminated the nearest-term heavy-lift competitor. SpaceX has at least 6 months with no competitive pressure on heavy-lift, potentially 18-24 months. This shifted probability from Negative toward Base and Positive.
  • U.S. Space Force contracts ($6.45B, May 28) added a multi-year government revenue floor.
  • Akademikerpension governance exclusion (May 29) is a negative signal for institutional demand. It deepened the negative-scenario tail risk without changing the central probabilities.
  • Anthropic S-1 filed (June 1). Anthropic filed a confidential S-1 at ~$965B valuation with $47B annualized revenue — nearly 5× its end-2025 run rate. At that scale, Anthropic has the capital to build its own data centers. The 90-day Colossus cancellation clause is now an active risk, not a contingency. This did not change the scenario probability distribution but deepens the negative-scenario tail in the AI sub-segment.
  • Alphabet $80B equity raise (June 1) — the largest equity capital raise in U.S. corporate history. Lead underwriters are Goldman Sachs, JPMorgan, and Morgan Stanley — the same three banks leading SpaceX’s IPO, Anthropic’s listing, and OpenAI’s upcoming deal. This creates a confirmed four-way competition for institutional capital allocation in the same window, with roughly $200B+ in combined demand hitting the same institutions from the same banks. Alphabet’s $180–190B in 2026 AI capex and $460B+ Google Cloud backlog also gives Anthropic a direct infrastructure alternative to Colossus, making cancellation near-certain. The scenario probabilities are unchanged — the Alphabet raise is a capital-market externality, not an operational change — but it narrows the bull-case path by reducing net IPO proceeds and accelerating the Colossus lease exit.
  • $20B bridge loan — IPO is a debt refinancing (June 3). SpaceX secured a $20B bridge loan with a 6-month takeout clause requiring repayment from IPO proceeds. Combined with the 78% of proceeds already pre-committed per the S-1, net new capital available for growth drops from the previously estimated $16–17B to approximately $5–10B — less than 5% of the $200B+ the company says it needs. The IPO is effectively a debt refinancing event, not a growth capital raise. A 15% greenshoe could add ~$11B in additional proceeds, but also creates $11B of insider selling capacity at the IPO price.
  • S-1/A amendment — new material disclosures (June 1). Key additions: (1) The Anthropic contract is confirmed as a 90-day exit lease in the company’s own language (see AI section above). (2) Q1 2026 actuals: revenue $4.69B, EPS loss –$1.27 (widened from –$0.18 YoY) — AI segment capex alone was $7.7B in a single quarter, more than double both other segments combined. (3) Mars milestone grant: 1 billion performance-based Class B shares to Musk, vesting across 15 tranches tied to market-cap milestones and establishment of a permanent Mars colony with at least 1 million inhabitants; a further ~302M shares tied to non-Earth data centers delivering 100 terawatts annually. (4) Water scarcity added as a material risk factor for data center expansion. (5) CalPERS, NYC Comptroller, and NYS Comptroller sent a joint governance letter demanding one-share-one-vote, an independent board, and elimination of the self-removal clause.
  • CFIUS / national security risk (June 3). The xAI-SpaceX merger triggers a mandatory CFIUS review under FIRRMA: SpaceX holds classified defense contracts, xAI’s Grok is used inside the Pentagon, and the merger involves Saudi Arabia’s PIF (100% government-controlled) at $3B+. The statutory review timeline (90+ days) cannot complete before June 12. A separate concern: the S-1 does not appear to disclose Chinese investors on the cap table (investor Iqbaljit Kahlon testified under oath in October 2025 that he routed money from China through the Caribbean to buy SpaceX stakes multiple times), the February 2026 Warren-Kim Senate letter demanding a DoD review, or the potential for a FOCI determination that would impose government oversight of corporate decisions — including restrictions on the classified defense contracts that represent ~20% of revenue. These are potential material omissions.
  • EchoStar spectrum acquisition — arm’s-length valuation signal (June 3). SpaceX’s September 2025 agreement to acquire EchoStar’s spectrum assets priced the deal at $42.40/share in SpaceX Class A stock, implying $531.5B total equity value — roughly one-third of the $1.75T IPO price. This is a binding arm’s-length transaction with a motivated seller, not a model output. It provides an independent valuation anchor well below the IPO price. The acquisition also adds $19.6B to SpaceX’s capital obligations (261.8M shares of dilution + $8.5B cash/debt), bringing the total funding gap from >$180B to >$200B.
  • Directed share program: 5% exempt from lock-up (S-1/A, June 1). Up to 5% of IPO shares are reserved for employees and persons selected at the discretion of executive officers, administered by Morgan Stanley. These participants are not subject to standard lock-up restrictions — they can sell after the first post-IPO quarterly earnings report, while regular investors remain locked. This creates a sell-first privilege for insiders at the exact moment the first 20% lock-up release hits the market and the first earnings report reveals the operational reality.
  • 30% retail allocation (June 2). SpaceX is directing an unusually large share of the offering to individual investors via Robinhood, Fidelity, and Charles Schwab. Typical large IPOs allocate 5–10% to retail. The 30% tranche targets Musk’s follower base but produces a float weighted toward investors who are more likely to sell on the first negative headline.

The June 2 update widened the confidence interval on Blue Origin’s return-to-flight (CEO Limp claims end-2026; Ars Technica’s Berger assesses “almost certainly not 2026, H1 2027 heroic”; NASA Isaacman says 2028 possible), but did not change the central estimate of SpaceX’s competitive relief.

The valuation assessment

Component Comparable Fair value range
Starlink Rural broadband utility (Comcast/Charter) $150-300B
Space / Starship Option value; currently negative FCF $50-200B
AI (all sub-businesses) Compute leasing + option on orbital $50-300B
Sum of parts $250-800B

Gap from the $1.75T IPO price: $950B-$1.5T.

The gap is entirely the present value of the seven sequential unproven dependencies — priced as near-certain by the IPO narrative. Probability-weighted expected value: approximately $725-925B.

The $1.75T is defensible as a long-duration option price if all seven gates resolve favorably over 10-15 years. At that point SpaceX could credibly be worth $3-5T. Discounting back at appropriate rates and applying realistic probability weights yields a present value closer to $500-900B.

Two independent analysts have now published valuations: Morningstar at $780B (48% below the IPO price, using probability-weighted sum-of-parts) and NYU’s Aswath Damodaran at $1.22T (the “Dean of Valuation” used 8% cost of capital with 2036 revenue projections, and stated he “would not be interested in buying at the rumored IPO pricing of $1.75 trillion” — while also cautioning that SpaceX is “an even more dangerous company to sell short” than Tesla). Both are independent of the framework’s analysis and both land well below the IPO price.

The strategic implications

For management: the strategic priority is Raptor 3 investigation above all other technical programs. It is the single-point blocker on the entire sequential dependency chain. Every week of Raptor 3 delay is a week of Kuiper subscriber growth in SpaceX’s core market, a week of open-source AI advancement eroding the Colossus moat, and a week of depreciation tsunami building without V3 revenue to offset it.

For investors: the fundamental question is not whether SpaceX will eventually succeed — the technology thesis and the execution track record are both compelling. The question is whether $1.75T today is the right price for that long-duration option, given that all seven sequential dependencies are unproven and the quarterly earnings cadence will systematically test each one against reality beginning in approximately eight weeks.

The SpaceX analysis demonstrates four things the framework is built to do: make maturity differences visible across consolidated business lines rather than smoothing them into a single valuation; price a long sequence of multiplicative dependencies explicitly rather than treating them as narrative; produce a structured scenario set with probability weights rather than a single forecast; and identify the specific inflection point — Starship operational cadence — on which most of the valuation sits, and date it. The finding is not that the IPO is wrong or right. It is a structural reading of the gap between the expected value and the price, with explicit confidence bounds, that lets the reader form their own judgment.

This is not investment advice. The framework produces an analytical reading — the decision to invest or not is the reader’s. Analysis date-stamped 2026-06-02; conditions will change.

To see the methodology applied to a question of your own, get in touch →.