Question: Large IPO winners — the positive base rate: which big IPOs did NOT drop for a decade, and what they shared Verdict: winners-cluster-in-reasonable-multiple-profitable-durable-moat-not-at-peak; NO-canonical-winner-priced->20-40x-sales-at-a-market-high-and-compounded-a-decade; snowflake-decisive-counterexample-100x-sales-still-negative-5yr; great-business-not-equal-great-stock-from-peak-cisco-intel-msft; spacex-~94-107x-sales-net-loss-all-time-high-falls-in-86pct-loser-bucket-on-all-four-factors

What we're asking

The inverse of the issuance-surge / Ritter base rate. Most large high-multiple IPOs underperform (Ritter: 86% of >40× revenue IPOs trail the market over 3y). The more interesting question: which large IPOs did NOT drop — compounded for a decade-plus — and what did they share that the 86% lacked? Then: which bucket does SpaceX fall in? Companion to 2026-06-02-ai-mega-ipo-issuance-surge-base-rate.

What we found

The winners cluster in one profile

Cautionary set (great business ≠ great stock from the peak price): Cisco & Intel — dominant, profitable, growing — but 2000-peak buyers waited 16–26 years to break even (never in real terms). Snowflake (2020) is the decisive modern counterexample: IPO'd at ~100× sales, best-in-class 174% growth, durable data-cloud moat → revenue up ~600% since, stock still negative 5+ years later because the multiple collapsed ~90%.

The distinguishing profile (ranked by discriminating power)

1. Digestible entry multiple (P/S), not rich. Cleanest separator. No durable winner IPO'd above ~10–20× sales except hyper-durable SaaS. The "expensive" winners (Google P/E 120, Apple P/E 110) were rich on earnings but cheap on sales/growth — the multiple could hold/expand, not de-rate 90%.
2. A moat that grows fundamentals faster than the multiple de-rates. Network duopolies (MA/V), search monopoly (GOOG), software lock-in (MSFT/ADBE). Math: IPO at 3–10× sales and the business only has to grow while the multiple stays flat or falls modestly and you still win; IPO at 90–100× and even 600% growth loses (Snowflake).
3. Not bought at a frothy cycle/sentiment peak. The anti-peak winners prove it — Netflix (2002 dot-com bottom), Visa (March 2008, into the crisis), GOOG/CRM/Domino's (2004, post-bust). The cautionary set is the mirror: great businesses bought at the 2000 peak.
4. Profitability or a fast, cheap path to it — strong, but not strictly necessary if (1) holds (Ritter: positive pre-IPO earnings predict better long-run returns).

The "boring winners" effect (Mastercard, Domino's, payment networks, franchises beating hyped tech) is just factors 1–4 in disguise: boring businesses get priced reasonably because they aren't hyped.

The rich-multiple-at-a-peak stress test

Essentially NO canonical winner priced at a genuinely rich revenue multiple (>20–40× sales) AND at/near a market high and then compounded for a decade from that price. The exceptions are deceptive (Google rich on earnings, cheap on sales). Snowflake is the clean falsifier: everything a winner needs except a digestible price, and it still lost money for IPO buyers. "Rich multiple at a peak" is almost always a loser even among eventual great businesses.

Verdict + reasoning

SpaceX falls in the 86% / loser bucket on all four discriminating factors. At ~$1.75T on 2025 revenue ~$18.67B = ~94–107× sales (richer than Snowflake's ~100×), a $4.94B net loss (post-xAI), debuting into all-time highs / peak underwriting cycle — it fails the entry-multiple, profitability, and not-at-peak tests, and the bundle sells a genuine moat-y compounder (Starlink, ~$11.4B rev) inside a frothy AI wrapper at a frothy price.

To be a rare rich-multiple winner it would need the NVIDIA/Amazon move — outgrow the multiple before it de-rates: ~$175B revenue (≈10× growth) just to grow into today's price at a mature ~10× sales, during which the stock could go nowhere for years (the Cisco pattern), and margins must inflect from a $5B loss to strong profitability simultaneously. The historical record says paying ~90– 100× sales at a peak loses even when the business turns out great.

This sharpens two lane items:

• Regime-Top hypothesis: reinforces that the regime-top read is strongest as a top-tick for the high-multiple AI complex and for the IPO buyers themselves, not the broad S&P. The winner profile is almost the photographic negative of where SpaceX/the AI labs sit.
• SPCX child case (not parent-owned): the base rate is bearish for $SPCX as a stock from the IPO price — distinct from SpaceX-the-business, where Starlink alone could be a Visa-class compounder if it ever traded at a reasonable multiple. Route this to the SpaceX perspective, not the lane.

Caveat: unprecedented deal; the genuinely secular end-markets (Starlink, launch monopoly) are the strongest counter-tell. The claim is about entry price and timing, not business quality.

Generalized into a reusable instrument

The four-factor profile generalizes beyond this lane into a standing IPO-Quality Scorecard, validated 2026-06-02 against three recent IPOs: Cerebras 0/4 (loser — fails on substance: ~111× sales, accounting-flattered profit, ~86% UAE customer concentration, frothy peak), Figma 2/4 (mixed "great business, rich stock" — passes profitability + moat, but the +250% day-1 pop took ~26× → ~90× into a hot window), SpaceX ~0/4 (fails on both). The gating finding: factors 1 (entry multiple) and 4 (timing) protect the buyer; factors 2–3 only pay off through the multiple — so a winner-business can still be a loser-stock. Score every IPO at two prices (offer AND day-1 close); the gap is the retail trap.

Sources

Price, RSI and trend figures read from the desk's pre-computed scan summaries (summaries). No number in this note was computed in prose.