No catastrophic blip: a 65/35 call that the flagship AI IPOs price through without breaking the boom
Asked (summary):
Given Keith Rabois's argument that a weak Anthropic or OpenAI IPO would be catastrophic for venture-funded AI — make your best prediction on what will actually happen. Pick one side.
This call rests on 36 evidence rows (30 distinct articles; syndicated URL variants counted once) from FT, Reuters, CNBC, Bloomberg, TechCrunch, PitchBook and Crunchbase, published 17 August to 10 September 2026. The prediction is judgment, not fact, with an evidence cutoff of 10 September 2026.
Two scenarios, one pick
Probability split is a forecast (judgment). Bars below it are rough revenue multiples: equity valuations divided by annualized-pace or estimated revenue — not audited IPO metrics. Hover any bar or scenario for detail.
Anthropic's annualized revenue pace hit $65bn in July 2026, and the FT reports a late-September or early-October New York debut with Morgan Stanley and Goldman Sachs in top roles — giftarticle.ft.com
Anthropic reached its first adjusted operating profit in Q2 2026 even as investors contemplate $2tn or more for the listing — giftarticle.ft.com
OpenAI's CFO told employees it "will be a public company in 2027" or sooner — timing flexibility that lets it wait out a weak window rather than force a fragile trillion-dollar offer — cnbc.com
Contagion risk is real: over 40% of venture-debt deal value sits in just five deals, and around 70% of AI venture dollars are spent on compute — pitchbook.com, pitchbook.com
What the evidence says about each side
Company evidence, deduplicated across syndicated variants
Anthropic — the likelier near-term listing
$65bn annualised revenue pace in July, up from a $965bn Series H valuation set in late May (techcrunch.com)
Confidential IPO prospectus filed with the SEC in June; $15bn pre-IPO credit facility being finalised (cnbc.com, bloomberg.com)
Reuters sources cite a 2028 revenue figure of as much as $200bn behind the $2tn contemplated valuation (reuters.com)
Risk: a $45bn Nscale compute commitment and reports its best model struggles to attract users against cheaper tools (bloomberg.com, giftarticle.ft.com)
OpenAI — can wait rather than risk a bad print
Valued at $852bn; ads business hit a $1bn annualized run rate and Amazon advertisers are now buying into ChatGPT (cnbc.com, cnbc.com)
$6.7bn revenue in the quarter ended June 2026, with an IPO "expected this year or next" (techcrunch.com)
CFO Sarah Friar told employees it will be public in 2027 or sooner — the delay option is the core of this forecast (cnbc.com)
The brief's context adds HSBC estimates via Reuters Breakingviews of $34bn 2026 revenue, $64bn in 2027, and roughly $25bn cash burn in 2026 — the reason a forced listing would be risky
The nuance: no singular catastrophe does not mean no reset. Public disclosure of compute costs and margins will probably split the market — frontier labs and a few infrastructure winners keep access to capital, while undifferentiated application startups face lower multiples, tougher rounds, and more M&A. PitchBook already reports M&A beating IPOs for AI exits and IPO activity soft outside a handful of megacap names. Selective repricing, not systemic collapse.
Source table
30 distinct articles behind this call, newest first
Date
Company
Title
Key fact
Source
Method: 36 evidence rows collected across FT, Reuters, CNBC, Bloomberg, TechCrunch, PitchBook and Crunchbase, published 2026-08-17 to 2026-09-10; syndicated and AMP URL variants of the same article were deduplicated to 30 distinct pieces for the table. Revenue figures are annualized run-rate paces or analyst estimates, valuations are private equity marks or contemplated IPO values; multiples shown are the ratio of the two and are rough, not audited IPO metrics. The 65/35 probability split is the author's judgment as of the evidence cutoff, 2026-09-10. Some article detail fields were trimmed for space.