SpaceX / SPCX IPO + cancelable AI-compute deals (Google, Anthropic)

Built from research/fin-spacex-spcx.json (web-verified: TechCrunch, Yahoo Finance, Reuters-sourced aggregators, VC Corner S-1 teardown). S-1 figures are from the prospectus — treat valuation/price as the offering range until priced. Supersedes the earlier sec-filings.json "not yet filed" status (the S-1 became public 2026-05-20).

The key structural point. SpaceX is the separable node in the bubble graph: it connects to the AI core through cancelable compute deals, so it enters or leaves the circular core (Tarjan SCC) depending on whether those edges are counted — the dual-SCC test the project runs. SpaceX has its own exogenous revenue (Starlink subscribers, launch, Starshield), which is why it is not trapped in the self-funding loop the way the AI-core firms are.

1. The S-1 facts

• S-1 filed 2026-05-20, S-1/A 2026-06-01; Nasdaq / Nasdaq Texas ticker SPCX (CIK 1181412).
• ~$135/share, 555.6M shares (all primary); $1.75T–$2T target valuation; ~$75B raise.
• 2025 revenue $18.67B — Starlink/connectivity ~$10–11.4B, launch ~$6.4B, Starshield ~$1.8B.
• 10.3M Starlink subscribers, 9,600+ satellites, ~50% YoY growth.

2. The cancelable AI-compute edges

• Google → SpaceX: a ~$100B equity stake (2015-invested, 2026-marked, held) plus a reported $920M/month compute arrangement (TechCrunch, Jun 2026).
• Anthropic / others: additional compute-and-capacity deals.
• These are flagged cancelable in the graph (cancelable: true), which is exactly why the dual-SCC analysis in build_graph.py shows SpaceX dropping out of the circular core when cancelable edges are excluded — the formal demonstration that SpaceX is attached to, not trapped in, the AI loop.

2b. Operational fusion in public CT (overlay, 2026-08-21)

The cap-table merger is already an edge. Certificate Transparency on the two apexes shows Grok's admin/auth living on spacex.com (admin.grok.spacex.com, auth.grok.spacex.com) and an xAI API shard named for Palantir (us-south-1-pltr.api.x.ai). That is operational naming, not a new dollar flow; it is the same SpaceX–xAI fusion the S-1/reporting describe, visible without touching either network. Catalog: osint/catalog/entities/spacex-xai.md.

3. Why it matters to the thesis

The project's circularity proof (Z3 T1–T5) is careful to show SpaceX is separable: it has real exogenous cash flows (Starlink/launch/Starshield) and its links to the core are contractual and cancelable. Treating it as part of the self-funding loop would overstate the core — so the model excludes it from the robust SCC. The SPCX IPO (a ~$1.75T+ all-primary raise) is itself a major liquidity event that could either absorb or strain the capital pool the AI core competes for. Cross-ref spec-telecom-satellite, geopolitics-cables-space-layer, data/graph.json (dual SCC).

4. Limits

Valuation/price are the offering range until priced; the Google compute figure is reported, not filed. The "separability" is a structural (graph) result, not a claim about SpaceX's business quality.

Sources: TechCrunch — Google $920M/mo to SpaceX; Yahoo Finance — SpaceX IPO prospectus; The VC Corner — SPCX S-1 teardown.

← Research index · structured data: fin-spacex-spcx.json · fin-spacex-spcx.md