Track: openai — sibling case under the AI-mega-IPO-supercycle frame. No dedicated OpenAI thread yet; this is the first OpenAI-track artifact, held by the parent lane ([[2026-06-02-ai-mega-ipo-fast-entry-liquidity-regime]]) until it earns one. Template: [[2026-06-03-spacex-net-flow-model-passive-demand-vs-lockup-supply]] + [[2026-06-17-spacex-post-ipo-study-4-float-melt-up-vs-the-dated-staggered-lockup-overhang]]; sibling scaffold: [[2026-06-17-anthropic-ipo-net-flow-scaffold-parametric-pre-s1]].

Question: OpenAI's draft is confidential (~2026-05-22, secondary-sourced). Same fill-in-the-blanks goal as the Anthropic scaffold — but OpenAI carries one risk the others don't: it may not be a clean common-stock IPO at all.

Verdict (prelim): The net-flow framework would transfer the same way as SpaceX/Anthropic if OpenAI lists as ordinary equity. The dominant open question is upstream of float/lockup: OpenAI's capital structure (nonprofit parent / capped-profit / PBC-conversion / Microsoft profit-participation units) may make "float," "shares," and "lockup" behave unlike a normal IPO — or gate what can be sold to the public at all. Hold OpenAI as a separate case, not a SpaceX clone, until the structure is public.

Status (S-1 watch, 2026-06-17)

S-1 watch (task TASKS-RESEARCH.md → "Anthropic + OpenAI S-1 filing watch") — trigger NOT fired: still confidential, crosswalk playbook does not run yet. Confirmed via web (now primary-sourced, upgrading the earlier secondary-sourced ~05-22):

• Confidential draft S-1 filed 2026-05-22; OpenAI publicly acknowledged it 2026-06-08 (and expected it to leak). Not on EDGAR; full statement public ≥15 days before roadshow. Target listing window: Labor Day → Thanksgiving 2026 (Q4) → financials likely visible Aug–Sep 2026.
• Valuation $852B–$1T (per the confidential filing, secondary-reported). Underwriters: Goldman Sachs, Morgan Stanley, JPMorgan.
• Unit economics now quantified (and brutal): $2B/month revenue ($24B annualized run-rate) but loses ~$1.22 for every $1 earned — confirms the "burn + ARR-quality, not clean P/S" framing below; a negative-and-widening operating loss is a first-order S-1 read.
• Federal-revenue line live: same $800M Pentagon agentic-AI award (xAI/OpenAI/Google/Anthropic) → OpenAI carries a govt/defense revenue line into its S-1 (read item added below). (Cross-ref the gov-compute integrators gap.)

Trigger stays armed; the dominant unknown (capital structure) is still unresolved until the public S-1.

Why OpenAI is the wildcard (the prelim note that matters most)

Where Anthropic is a comparatively conventional PBC with VC + hyperscaler equity (so the SpaceX shape is a strong prior), OpenAI's structure is the gating unknown:

• Capped-profit / nonprofit-control overhang. Profit participation is capped and the nonprofit retains control — so what the public buys, how much can float, and who can sell may not map onto "free float % + lockup ladder" at all. The melt-up mechanic needs a thin tradeable float; if the structure constrains or reshapes that, the mechanic could be amplified (even thinner economic float) or muted (atypical instrument). Unknown until filed.
• Microsoft as the strategic holder + proxy. MSFT is the public read-against for OpenAI the way AMZN/GOOGL are for Anthropic — its OpenAI economics + any mark sensitivity are the closest tradeable surface. No clean OpenAI proxy otherwise.
• Same fundamentals froth as the labs broadly: net loss, enormous compute burn, SPV-stacked pre-IPO marks (the Anthropic risk #1 applies here too), and no profit path proven — so the "crazy multiple" shows up as burn + ARR-quality + structure risk, not a clean P/S.

Fill-in-the-blanks (gated on the public S-1)

What this does NOT decide

No trade, no valuation, no structural ruling. Scaffold + prelim notes only. The lane's standing Trading Posture applies unchanged: good company ≠ good trade; the pop is a trap for the real book; wait for the post-lockup reset; double-patience (no dated calendar until the S-1) — and for OpenAI, triple-caveated on the structure question above.

When the S-1 surfaces (the actual work this unblocks)

1. Resolve the capital structure first — is the public instrument ordinary equity? Only then does the net-flow table apply.
2. If equity-like: plug float/lockup into the net-flow equation; recompute the dated overhang.
3. Grade MSFT mark sensitivity + any SPV-stack down-round risk vs the IPO print.
4. Read the federal/defense revenue line (Pentagon agentic-AI program + any IC contracts) → revenue-concentration / political-risk grade, same as the Anthropic read item.
5. Decide whether OpenAI graduates from "parent-lane child case" to its own thread/holding-pen (the Anthropic precedent — stand one up once there's a real S-1 to read).