Invest1 publisher3 min readPublished
Anthropic's IPO asks public markets to double a $965bn private mark
David Sacks wants the listing paused until a departing researcher's safety claims are checked, which turns a governance argument into a calendar problem for a deal that has to price before the November midterms.
The Investor · Invest desk
What happened
- Jacob Coxon resigned from Anthropic and said publicly that after three years of pretraining research at OpenAI and Anthropic, neither company is acting responsibly and both are gambling with our lives.
- David Sacks responded on Thursday that Anthropic's initial public offering must be paused until the whistleblower's claims can be investigated.
- Lawmakers picked it up from both flanks, with Representative Ted Lieu tying the warning to a bipartisan AI Kill Switch Bill and Representative Lori Trahan saying Congress can no longer remain passive.
- Anthropic filed a draft S-1 confidentially with the SEC on June 1 and is expected to open IPO marketing in mid-October in order to list before the November midterm elections.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint The scarce asset is the pre-midterm window, so an investigation thorough enough to answer Sacks consumes the thing the deal was scheduled around, and the fallback date sits inside a legislative session that now has kill-switch sponsors.
- decision Buyers have to decide how much governance discount to apply before the disclosures that would size it exist, which means the safety argument gets priced by instinct while margins and compute costs get priced by arithmetic later.
- exposure Morgan Stanley, Goldman Sachs, JPMorgan and Citi are the parties who must now judge whether a departing pretraining researcher's warning belongs in the risk factors, and they carry it either way they rule.
- contradiction Coxon says neither lab is behaving responsibly while Anthropic points to published safety research and asks for a lawful, verifiable industry coordination mechanism, and that request reads either as evidence of care or as an admission that unilateral restraint is not commercially available.
Sixty-five billion dollars of annualized revenue at the end of July is the denominator everything else hangs off [13], and a public mark near $2 trillion puts that line at about 31 times [11][1], where the last private round priced it at roughly 15 [10][2]. Nothing in the record says the business changed between the June 1 draft filing and this week [8]. What changed is who is being asked to pay the multiple.
Dollars are the clearer frame. The gap between a $965bn private mark and a $2 trillion listing is about $1.03 trillion of value nobody has ever priced [10][11][3], which is roughly 1.2 times OpenAI's entire latest private valuation of $852bn [10][3]. Cryptopolitan puts Anthropic's ask at about a 107% step-up and OpenAI's reported public target at about 17% [12], a sixfold gap between two companies in the same sector filing in the same month [5].
What David Sacks asked for is an investigation before pricing [1], and the available evidence describes the demand and the political echo around it, not any authority to enforce it. The echo is real enough on its own terms, with Ted Cruz calling AI a catastrophic risk and Bernie Sanders calling for artificial superintelligence to be halted until safety rules exist [6]. Jacob Coxon resigned on September 9 [2], marketing is expected to open in mid-October [8], and five weeks is not long enough to investigate a claim that advanced systems could end the species by the end of the decade [3][4]. Any process deep enough to satisfy the demand eats the pre-midterm window [8] and drops the deal into a Congress where Ted Lieu already has a bipartisan kill-switch bill to point at [7]. That cost is counted in weeks, a different currency from the basis points that will price the deal later.
Which is also why the governance question resists pricing at the moment: the figures that would let a buyer size it, audited margins, compute expenses and customer concentration, are not out [14]. My read, and the evidence supports it in direction rather than in magnitude, is that the safety dispute costs Anthropic timing while the compute line costs it valuation, and that Evan Hubinger's better-than-one-in-ten estimate of everyone dying within a decade [4] will move the book less than a single disclosed gross margin. The more telling point is that a buyer who genuinely believed the extinction claim would not be haggling over the step-up at all, so the presence of haggling tells you what the marginal investor thinks the claim is worth.
That read would be wrong under two conditions. If the deal prices at or above $2 trillion with a covered book, the whistleblower fight was noise and the step-up was underwriting the sector rather than the company. If it slips past November, Sacks got his pause on the calendar without ever winning the argument.
What to watch
- Whether IPO marketing actually opens in mid-October, the first hard test of whether the pause demand has any grip on the timetable.
- Whether the underwriters add researcher departures or safety governance to the risk factors when the filing goes public.
- Whether Ted Lieu's bipartisan AI Kill Switch Bill picks up a hearing date before the listing rather than after it.