Expression memo — SpaceX post-IPO: multiple ways to win

Generated 2026-08-27 (green at 88.0, EARLY flag, −14.9% excess vs SPY since re-birth 2026-06-19 · eval 2029-06-19). Everything below the "spoken core" is DERIVED — OURS — NEVER SCORED. Verdicts and the scoreboard trace only to the spoken plays.

The spoken core (what IS scored)

• SPCX (primary) — IPO'd 2026-06-12 at $135, day-one close $161, peaked ~$200, round-tripped to ~$150. The idea is scored on this and nothing else.
• ARKX (adjacent) — diluted space-economy basket.

The besties' stacked-optionality case: Starlink cash machine ($11.4B rev, +50%, 10M subs → "hundreds of millions" per Jason, E274), EWS (Anthropic paying $1.25B/month for Colossus — $15B/yr, "added a Starlink in revenue," E274), Grok+Cursor (Composer 2.5 Pareto-dominant after 3 weeks on Colossus 2, Gavin E274), Starship rapid reusability (Gavin: "a year or two," E274), and Chamath's underwriting path ($25-30B '26 rev → 40-45 → doubling; terrestrial DCs alone $100-200B rev by 2030-32, E274).

Second-order beneficiaries (derived)

• The Colossus supply chain: every EWS dollar is a buildout dollar — NVDA GPUs (but that's the capex idea's turf), behind-the-meter power workarounds Chamath described in E281 (mobile turbines, Bloom-style onsite natgas → the natgas-powers-ai idea already holds that lane), and Tesla Megapods (Chamath: "we would be enormous buyers... 90-day build cycle," E278) — TSLA exposure to SpaceX's buildout is also the tesla-second-act merger lane (Chamath: merged "by end of 2026," E272/E277).
• Index-flow mechanics: early index inclusion (E280) means passive bid grows with each staged lockup that clears without breaking the stock — Gavin's anti-lockup-short case (E278: everyone could already sell every 6 months; half of employees bought at IPO).
• Orbital compute (Gavin's math, E278: $40B/GW in space vs $60B+ terrestrial and diverging; point prediction H2'28–H1'30): there is no pure-play — SPCX itself is the orbital-compute call option.

Supplier / product-level drivers (derived)

• Watch EWS concentration risk become EWS diversification: Gavin expects "partner to become partners" (E274). A second anchor tenant (Google was named as the likeliest buyer of Elon's compute) would de-risk the $15B/yr better than any Starlink quarter.
• Starlink competition is the soft underbelly the besties themselves flagged (Jason, E277: "the next two or three competitors are coming right behind him").
• Cursor data moat: proprietary coding tokens > public internet (Gavin, E274) — model-quality prints (Composer releases topping leaderboards) are the leading indicator for the AI segment re-rating.

Options angles (derived — size like a degen who wants to stay one)

• The −15% post-IPO drawdown with a 3-year eval window is the textbook long-dated call setup: LEAPs (2028s when listed) express Chamath's "grow into the valuation" path with defined risk, instead of tying up cash in common at 60-70x revenue.
• IPO-price gravity: $135-150 is where retail got filled and where it has held — short puts at/below the IPO price are the "get paid to catch the knife where the demand lives" expression. Assignment = owning SPCX at better than IPO terms; only write what you'd hold.
• Post-IPO options are young: IV is rich around lockup dates — spreads over naked longs.

What kills it (invalidation — watch these, don't rationalize)

1. Anthropic's 90-day cancellation clause on the EWS deal (E274) — the single sharpest revenue cliff in the story.
2. A lockup wave that breaks $135 and stays there (deal-price-break selling is mechanical — Gavin's Cerebras lesson, E278).
3. Starship rapid-reusability slipping past the "year or two" consensus.
4. Starlink sub growth decelerating while competitors ship.
5. The idea's own eval: 2029-06-19. If conviction decays out of green on silence, the memo is stale — it does not refresh itself.