"SpaceX IPO: S1 Breakdown" — CJ Gustafson (Mostly Metrics)

Why this is in the vault

CJ's SaaS-CFO lens on the SpaceX S-1, filed the evening of 2026-05-20 and processed by him in time for a 2026-05-21 07:22 ET send. Direct corroboration / extension of yesterday's vault filing [[2026-05-20-spacex-s1-ipo-filing-with-xai-consolidated]] and the locked [[../01-projects/investing/theses/2026-05-20-elon-verse-v2]] thesis. Surfaces multiple disclosures yesterday's EDGAR-direct read either missed or under-weighted — most load-bearing: (1) Musk CEO-comp plan is gated on a million-person Mars colony, (2) a separate award is gated on 100 TW of non-Earth-based data-center compute, (3) Valor equipment-lease guarantee totals $20.2B with SpaceX as guarantor, (4) Cursor termination liability is $10B if SpaceX walks. Triggers a thesis-update sweep.

⚠️ Sponsorship

None disclosed in this issue. Single-thread editorial; no top-of-issue sponsor block, no embedded "presented by." CJ's standard "Disclosures: None of this is investment advice. Do your own homework" closer is present. Treat as un-sponsored.

Issue contents

Single-thread S-1 breakdown. CJ walks the segments, key stats, Starlink unit economics, the consolidated-entity accounting, the financials, the red flags, the cap table and IPO mechanics, the valuation comp ladder, and a misc. section. No "Mostly Multiples" benchmarking sidebar this issue — full attention on SpaceX.

Verbatim section headings (from the body)

1. SpaceX IPO: S1 Breakdown
2. What Does SpaceX Do?
3. Key Stats
4. The Starlink Section (it's late, not a creative title, IK)
5. About that consolidated entity
6. Financials
7. Potential red flags
8. Cap table and IPO structure
9. Valuation
10. Misc. stuff of note

Core thesis

CJ's TL;DR on the consolidated entity: "Starlink is the part that makes money. xAI is the part that burns it. The launch business sits between them, mostly serving the other two." The $18.7B 2025 revenue and $2.6B operating loss only exist as one number because of common-control accounting that retroactively combined SpaceX + xAI + ex-Twitter back to 2023 as if they had always been one company. There is no clean SpaceX-only comparable in this filing. The IPO is being priced at 60-70x forward revenue (vs Cerebras at 50x, NVIDIA at 18x, Tesla at 14x, median tech at 3x) — a multiple that requires you to be pricing a 2030 outcome where Starlink is doing $40B revenue, Starship is reusable at commercial scale, and at least one of orbital AI compute or Mars optionality has started to print real numbers.

Key disclosures not previously covered in [[2026-05-20-spacex-s1-ipo-filing-with-xai-consolidated]]

These are the load-bearing additions from CJ's read that warrant a thesis-update sweep.

1. Musk CEO comp plan — Mars colony and orbital compute as vesting gates

• 1 billion restricted Class B shares granted to Musk in January 2026.
• Vests in 15 tranches. Each tranche requires BOTH (a) a market-cap milestone (from $500B up to $7.5T) AND (b) the establishment of a self-sustaining human colony on Mars with at least one million people.
• A separate 302 million share award vests at market caps from $1T to $6.5T PLUS the construction of non-Earth-based data centers delivering 100 terawatts of compute.
• Both awards approved in writing by the board.
• Why it matters for elon-verse v2: shifts how to read Musk's incentive structure post-IPO. The headline "trillion-dollar pay package" is only fully unlockable on outcomes (Mars colony of 1M people, 100 TW of off-Earth compute) that are functionally lifetime objectives. The intermediate tranches still pay out on market-cap milestones alone (per the text — though CJ's exact wording is "each tranche requires both" — worth a primary-source verification on whether early tranches actually require both gates or only the cap milestone). This is a material governance and incentive disclosure that yesterday's note treated only as ticker/structure mechanics.

2. Valor Equity Partners equipment-lease guarantee — $20.2B aggregate

• Antonio Gracias sits on the SpaceX board, runs Valor Equity Partners ($55B AUM), and Valor entities collectively hold ~7.3% of Class A.
• xAI subsidiaries have signed three separate equipment lease agreements with Valor for aggregate cash payments of $6.986B + $6.633B + $6.587B = $20.206B total.
• SpaceX is the guarantor on all three. The SpaceX board (which Gracias sits on) approved the SpaceX guarantee.
• S-1 doesn't specify the equipment but timing and dollar size strongly suggest the COLOSSUS GPU clusters.
• Why it matters: this is a $20B contingent liability on SpaceX's balance sheet structured through a director-affiliated entity. Related-party section runs 9 pages. CJ's framing: "Cerebras's related-party section looked like the cap table of a Targaryen wedding. SpaceX makes Cerebras look like a tame holiday party at a local VFW." Yesterday's vault note flagged controlled-company status but did NOT enumerate the Valor lease/guarantee structure.

3. Cursor option — $10B termination liability if SpaceX walks

• April 2026: SpaceX entered a compute agreement and an option to acquire Cursor at $60B implied equity value.
• If SpaceX terminates the option, or Cursor terminates for SpaceX material breach: SpaceX owes Cursor $1.5B in termination fees + $8.5B in deferred services fees = $10B.
• If the acquisition happens, paid in SpaceX Class A stock at VWAP over the seven trading days before close — so a stock pop makes the deal more expensive in shares for SpaceX.
• Why it matters: $10B downside is material relative to $15.9B cash on hand at Mar 31, 2026. This is a directional integration with the AI segment that yesterday's note did not capture and that may need its own vault stub. Also, a Cursor exposure now lives implicitly inside any SPCX position.

4. Tesla related-party footprint — $700M+ in a single year

• xAI bought $506M of goods/services from Tesla in 2025, another $34M in the first two months of 2026.
• SpaceX bought another $144M from Tesla.
• Tesla committed to invest in xAI in January 2026; when SpaceX acquired xAI a month later, that investment converted into a SpaceX equity stake (CJ has Tesla at ~1% of Class A post-IPO).
• A joint chip manufacturing project called Terafab is described at length in the prospectus without specifying any timeline, capex, or milestones.
• Why it matters: directly intertwines TSLA and SPCX cashflow lines. The elon-verse v2 thesis treats them as complementary exposures but did not factor in $700M of annual related-party commerce or the Terafab joint program. Terafab unspecified-timeline-no-milestones is itself a yellow flag — investor-prospectus name-drop without disclosure substance.

5. CFO Bret Johnsen comp — the "sane" plan

• 4 million performance-based options amended Jan 2026.
• Vests 371,125 options per $10B of adjusted EBITDA achieved during FY2025-2029, assessed annually.
• Zero options vested based on 2025 performance (since consolidated entity ran an operating loss).
• CJ: "Bret Johnsen has the only sane compensation plan in this filing, which says something about either Bret Johnsen or this filing."
• Why it matters: the CFO is the only person inside the company whose comp is gated on actual cash earnings rather than market-cap milestones plus interplanetary milestones. Useful internal signal — if Johnsen ever resigns, that's the read.

6. Bitcoin on the balance sheet

• SpaceX holds bitcoin as a treasury asset, marked to fair value.
• Dollar amount not broken out in the summary; referenced in the accounting policies section under "digital assets."
• Puts SpaceX in cohort with Tesla, MicroStrategy, Block.
• Why it matters: secondary disclosure but worth tagging — adds to the BTC institutional-holdings dataset and links Musk-entity BTC exposure across TSLA + SPCX.

7. Nasdaq Texas dual-listing under ticker SPCX

• Dual-listed on Nasdaq AND Nasdaq Texas under SPCX.
• Nasdaq Texas launched 2025 in response to political pressure on companies leaving Delaware. SpaceX redomiciled to Texas in 2024; HQ moved from Hawthorne CA to Starbase TX in early 2026.
• SpaceX is the marquee listing; the largest IPO in US history is also Nasdaq Texas's debut institutional credibility moment.
• Why it matters: yesterday's note had "Nasdaq" only. Dual-listing introduces a venue-arbitrage and operational-routing question for Schwab entry — worth confirming Schwab routes orders to one venue or both, and what the BBO discipline looks like across two markets.

8. 30% retail allocation via directed share program

• A typical IPO allocates ≤10% to retail. SpaceX reserves ~30%.
• Mechanism: directed share program where underwriters set aside Class A shares at the IPO price for "employees and certain other designated individuals." Directed shares have no lockup.
• Institutional and pre-IPO Class A is on the standard 180-day lockup.
• Why it matters for the Schwab entry decision: CJ's framing is that opening-day price discovery happens between "meme-stock buyers and forced index buyers, with very little real selling pressure on either side." This is unusually one-sided liquidity. If true, the entry-day premium could be significant; a phased-buy approach (vs single-shot) may be the better Schwab posture. Flag for the elon-verse v2 execution plan.

9. Index inclusion forces buying within ~15 days

• S&P 500 inclusion happens within 15 days for IPOs meeting the size threshold (SpaceX clears it by ~150x).
• At $1.5T cap, the index weight is ~3% — large enough that index-fund managers have to sell other holdings to fund the buy.
• Same dynamic applies to 401(k) target-date funds, large pension funds, corporate retirement plans.
• Why it matters: a structural buy-side flow on a ~2-week horizon post-IPO. Combined with the 30%-retail / no-lockup-on-directed-shares dynamic, this materially shapes the first-month tape. The elon-verse v2 entry plan needs to factor this — entering BEFORE index inclusion may be a meaningfully different trade than entering AFTER it.

10. Cerebras comp anchor — 60-70x forward revenue

• 2026E revenue ≈ $25B (Connectivity +50%, Space +8%, AI +22%).
• At $1.5T cap → 60x forward revenue. At $1.75T → 70x.
• Public-market comp ladder per CJ: median tech 3x, top-10 tech (within his 142-name index) 14x, Tesla 14x, NVIDIA 18x, Cloudflare 25x, Palantir 36x, Cerebras 50x.
• "SpaceX is asking the market for somewhere between 1.2x and 1.4x the Cerebras multiple, on a company that loses $2.6B at the operating line."
• Why it matters: yesterday's note had the $1.5-1.75T valuation bracket but did not anchor it to the Cerebras-multiple frame. CJ's frame is sharper for advisory communication — when explaining the trade to the founder, "1.3x Cerebras at scale and on an operating loss" is the calibrated frame.

Operating metrics the SaaS-CFO lens surfaced that we didn't

These are the unit-economic framings CJ brings that an EDGAR-direct read tends to under-weight. Useful for thesis-update granularity.

• Starlink ARPU compression as deliberate strategy: $99 (2023) → $66 (Q1 2026). One-third decline in three years, most concentrated in the last 12 months. The S-1 attributes it to international expansion (sub-$99 markets) plus Starlink Mini residential tier. CJ's read: the trade works as long as cost-per-incremental-sub drops faster than ARPU, and the V3 Starlink satellites (~20x bandwidth per launch) provide the cost lever. Useful sensitivity frame for the Starlink scenario ladder.
• Starlink scenario ladder (CJ's): 10M subs @ $66 → $7B segment EBITDA today; 30M @ $50 → ~$18B; 50M @ $40 → north of $25B. Bear case: ground ops + support scale linearly rather than asymptotically, margin compresses from 63% toward regulated-utility levels, and valuation compresses with it.
• Enterprise Starlink zero-churn anomaly: since 2023, no enterprise customer paying >$750K/yr has voluntarily cancelled. CJ had to re-read the disclosure three times. Customers: United, Carnival, Maersk, John Deere. This is an exceptional churn number for any consumer-adjacent infra business and undergirds the enterprise-side EBITDA narrative.
• R&D-spend-ratio comp vs frontier-lab universe: SpaceX R&D = 46% of revenue ($8.6B). Cerebras 48%, NVIDIA 12%. Most of SpaceX's R&D is xAI compute depreciation ($12B GPU buys losing value fast), not Starship — ~$3B is Starship-specific R&D inside the Space segment. Useful frame for thinking about how much of the consolidated R&D number is structurally GPU-depreciation vs genuinely cash-burning future-bet R&D.
• Capex segmentation: 2025 capex $20.7B split as AI $12.7B (COLOSSUS + COLOSSUS II) / Connectivity $4.2B (sats + launch capitalization) / Space $3.8B (Starship). Q1 2026 capex alone was $10.1B. That run-rate implies $40B 2026 capex against ~$25B revenue — the funding-gap problem is accelerating, not closing.
• Backlog $28.4B with $12.1B already as deferred revenue, about a third recognized in next 12 months. Useful Starlink subscription-billing-upfront cash dynamic.
• Launch business as internal subsidy mechanism: launch services ran $4.1B revenue +8% Y/Y, $653M segment EBITDA. CJ's frame: looks like slow growth, but the launch business is increasingly being used to put up Starlink and AI satellites that don't show as Space revenue. The launch business is more accurately a captive infrastructure provider for the connectivity and AI segments than a standalone P&L.

Direct cross-references

• [[2026-05-20-spacex-s1-ipo-filing-with-xai-consolidated]] — yesterday's EDGAR-direct vault note. CJ's piece extends it on Mars-colony-CEO-comp, Valor $20.2B guarantee, Cursor $10B liability, Tesla $700M+ related-party, Bret Johnsen CFO comp, Bitcoin treasury, Nasdaq Texas, 30% retail directed-share, Cerebras multiple anchor.
• [[../01-projects/investing/theses/2026-05-20-elon-verse-v2]] — locked thesis. Founder approved 5% = $40k SPCX entry via Schwab and TSLA continuation. CJ's disclosures suggest the entry-execution plan should factor (a) the index-inclusion 15-day forced-buy window, (b) the 30%-retail / no-lockup-on-directed-shares one-sided opening-month liquidity, and (c) the Valor + Cursor contingent-liability stack as ongoing risk lines.
• [[2026-05-20-innermost-loop-may-20-singularity-calendar-futures-encyclical]] — Singularity-cluster framing context.
• [[2026-04-11-moonshots-ep246-spacex-ipo-claude-mythos]] — Apr-11 directional-call match on IPO timing and ~$2T bracket.

Mapping against Ray Data Co

• Elon-verse v2 thesis — load-bearing: triggers a thesis-update sweep. The 10 new disclosures above (especially Mars-colony comp gates, Valor $20.2B guarantee, Cursor $10B option, index-inclusion forced-buy dynamic, 30%-retail no-lockup dynamic) should be folded into the thesis as either incremental risks or execution-plan refinements. The founder-locked decisions (✅ approve v2 / ✅ Schwab entry / ✅ 5% = $40k / ✅ TSLA continuation) are not invalidated by CJ's read — none of the new disclosures break the thesis — but the entry-execution playbook benefits from the index-inclusion + directed-share dynamics.
• MAC / harness-engineering / Sanity Check: minimal direct mapping for these surfaces. The CJ-as-SaaS-CFO lens has more relevance for general newsletter-reader calibration than for any specific RDCO operating decision.
• Investing skill set: the Cerebras-multiple anchor frame (60-70x = 1.2-1.4x Cerebras on an operating loss) is a transferable communication frame — file under "calibrated valuation framing patterns" for future investing-thesis communication.

Sharp lines worth saving

• "Starlink is the part that makes money. xAI is the part that burns it. The launch business sits between them, mostly serving the other two." — the cleanest one-line summary of the consolidated entity's economic shape.
• "I have sat in audit committee meetings over a single $200,000 related-party transaction. The Tesla footprint here is north of $700 million in a single year."
• "Bret Johnsen has the only sane compensation plan in this filing, which says something about either Bret Johnsen or this filing."
• "SpaceX is asking the market for somewhere between 1.2x and 1.4x the Cerebras multiple, on a company that loses $2.6 billion at the operating line."
• "If 30% of the float goes to retail without lockups, and the rest gets vacuumed into S&P 500 trackers within two weeks, the price discovery in the opening month is happening between meme-stock buyers and forced index buyers, with very little real selling pressure on either side."

Quality / bias flags

• CJ's voice: SaaS-CFO operating-perspective, dry, willing to flag governance issues directly (the "LOL" at the comp-plan board sign-off, the Targaryen-wedding line). No detectable directional positioning on the stock — he's neither pumping the IPO nor shorting it editorially. Treats the disclosures as a CFO reads filings: what would I have flagged in audit committee, what would I want to know before I bought.
• Verification needed on Mars-colony comp tranche structure: CJ writes "Each tranche requires both a market cap milestone (from $500 billion up to $7.5 trillion) and the establishment of a self-sustaining human colony on Mars with at least one million people." If literally true, EVERY tranche is gated on the Mars colony — which would mean $0 of the trillion-dollar package vests until a million-person Mars colony exists. That seems extreme enough to warrant primary-source verification before relying on this framing in the thesis. May be that early tranches require only the cap milestone and later ones add the Mars gate. Surface as TODO.
• Common-control accounting framing: CJ correctly identifies that historical SpaceX-only comparables no longer exist in this filing. Worth noting that any analyst piece (incl. yesterday's vault note) that quotes "2024 SpaceX revenue" is now quoting a recast number that includes a full year of Twitter ads. Backward comparisons to pre-merger SpaceX leaks are not apples-to-apples.

TODOs surfaced

1. Thesis-update sweep on [[../01-projects/investing/theses/2026-05-20-elon-verse-v2]] — fold in the 10 new disclosures above, particularly the entry-execution refinements (index-inclusion window, 30%-retail no-lockup dynamic) and the contingent-liability stack (Valor $20.2B guarantee, Cursor $10B termination).
2. Primary-source verification on Mars-colony comp tranche structure — pull the actual award schedule from the S-1 (page 235 vicinity per CJ) to confirm whether every tranche or only the terminal tranches require the colony milestone.
3. Cursor sub-stub — the $60B-implied-value Cursor option with $10B termination liability deserves its own vault stub linking back to the SpaceX S-1 note and forward into the AI-IDE-tooling landscape tracking.
4. Terafab disclosure-substance check — joint TSLA/SPCX chip manufacturing program described at length without timeline/capex/milestones. If there's any external reporting on Terafab (Information / WSJ / FT), pull it; otherwise flag as an opaque program in the elon-verse v2 risk register.