David Sacks, Co-host of All-In and Former White House AI Advisor: Conflict Between Doomsday Narrative and Anthropic's IPO
David Sacks, co-host of All-In and former White House AI advisor, stated in the program that Anthropic embraces the doomsday narrative of AI destruction while pushing for an IPO, which is logically inconsistent. Investor Phil Trubey shared this discussion, adding that both Anthropic and OpenAI as public companies will face troubles; while current financial reports look good, their strategic thinking is poor, and they are managed by what he calls "quirky and untrustworthy" individuals.
Sacks characterized a recent public post by a short-term researcher who left as "doomsday psychological warfare," aiming not just at the reputation of a single company but at promoting the establishment of a federal-level AI regulatory body, ultimately targeting open-source. The researcher claimed that frontline builders genuinely believe AI could kill everyone within this decade, and that both leading labs are rushing towards self-improving superintelligence. Anthropic's alignment researchers subsequently publicly agreed, stating that their colleague's judgment was correct and estimating a more than 10% chance of extinction in the next decade, with alignment solutions yet to take shape. Sacks believes that the internal agreement among senior executives turns the "personal views of the departed" into a risk disclosure close to company endorsement, directly impacting the prospectus and roadshow window.
The IPO schedule and valuation have been repeatedly priced by the public market. Anthropic submitted a confidential prospectus to the U.S. Securities and Exchange Commission on June 1, with a previous round of financing at about $65 billion and a post-money valuation of approximately $965 billion; market rumors suggest a target valuation could reach $1.5 trillion to $2 trillion, benchmarked against SpaceX's IPO record of about $1.75 trillion to $1.77 trillion. Underwriters include Morgan Stanley, Goldman Sachs, JPMorgan, and Citigroup. Reuters cited insiders saying the public prospectus has been delayed to late September, with the roadshow starting as early as mid-October, and the IPO possibly occurring a few days before the November midterm elections; the company is also finalizing about $15 billion in revolving credit as a precondition.
Sacks presented a choice: either dismiss the researcher as an entry-level exaggerator or acknowledge their judgment, the latter of which would validate the political argument for "freezing development," making a trillion-dollar issuance unjustifiable. Chamath Palihapitiya compared this to his own experience during Google's IPO quiet period and his television remarks during Slack's IPO, which had to be rewritten in the prospectus, pointing out that once a safety narrative is confirmed internally by the company, it becomes a risk factor under securities law rather than a brand advertisement. The effective altruism network has been described by Sacks as having injected over $1 billion into AI risk advocacy, and a successful IPO would bring a new pool of public capital into the same narrative.
Dario Amodei's earlier judgment that "half of entry-level knowledge jobs will disappear within one to five years," along with rumors that "Anthropic could one day be the only private company in the world," were marked by Sacks as close to SBF-style self-aggrandizement. Gavin Baker, in the same program, stated that Anthropic is still growing rapidly, but marginal shares are simultaneously flowing to OpenAI, open-source models, and Grok; if the $65 billion financing figure is accurate, he suspects someone leaked data to inflate the upcoming roadshow's recurring revenue metrics.
The market mechanism is a struggle for pricing power driven by events. Buyers aim to acquire institutional funds in the public market for pure AI lab beta, as well as strategic shareholders like Amazon and Nvidia that have already priced in their stakes; sellers are existing shareholders in the primary market, employee allocations, and possibly insiders who may reduce their holdings during the issuance. Funds are migrating from private equity valuations to public indices and mutual funds, with the path requiring the prospectus to frame the "extinction probability" as a litigable risk paragraph. The beneficiaries are those who go public first, absorb disclosure risks first, and set industry multiples first, along with open-source and low-cost Chinese models; the pressured parties are the management of Anthropic and OpenAI, who must maintain both a safe brand and a growth story during the quiet period. Current financial figures are still used to support high multiples, while strategic narratives are turning the same set of figures into regulatory moats or regulatory bombs.
Trubey's judgment ties the two companies together: it is not the current break in the revenue curve, but rather that governance and strategic credibility are unsustainable under the public shareholder structure. Sacks condenses the conflict into a structural proposition—companies that issue the loudest warnings about danger are often the ones most likely to gain a moat from dangerous regulations.
Source: Public Information
ABAB AI Insight
David Sacks' attack is not a new front. He has moved from PayPal operations and the sale of Yammer to venture capital and All-In, then into roles as a White House AI and crypto advisor, and upon returning, he continues to frame Anthropic's safety lobbying as regulatory capture: pushing state and federal rules together, the prophecy of halving entry-level positions is shared by politicians but remains uncorrected, and whether the motives are sincere does not change the accusation of "turning the state machinery to one's own agenda." He also identifies effective altruism funding networks like Open Philanthropy as narrative blood pumps. The historical analogy for this path is his past handling of crypto regulation, content moderation, and open-source disputes as the same proposition: who defines risk, defines the license.
The capital path is to turn safety premiums into issuance premiums. $65 billion in private equity, about $965 billion in valuation, rumored $2 trillion IPO, $15 billion in revolving credit; money is first used to lock in computing power and corporate shares, then attempts to sell "responsible labs" as scarce assets on the public side. Amazon cloud distribution, Nvidia's stake, and coding agent revenues pull the growth curve into a narrative range of about three to ten times; at the same time, safety letters, internal extinction probabilities, whistleblower posts, and senior executive agreement rewrite media exposure that could have been used for the roadshow into securities law risks. The strategic motive is to price the industry ahead of OpenAI, at the cost of having to explain in the prospectus: if management truly believes there is a double-digit extinction probability within a decade, why are they still selling perpetual growth to public shareholders?
The analogy is not traditional tech IPOs, but rather two sets of precedents stacked together. Google's quiet period interviews nearly overturned the issuance, and Slack's network effect remarks had to be forced back into the S-1; on the other side, the FTX-style founder narrative collides with risk disclosures at its peak. The industry's position has shifted from expansion to control: whoever goes public first and discloses unit economics first will anchor CoreWeave, open-source models, and the next round of private equity. Baker points out that shares are simultaneously flowing to competitors, indicating that control is not stable; low-cost Chinese models are siphoning workloads at an 80-90% token discount, turning the "safety premium" into a cost item that can be circumvented.
Structural changes occur simultaneously with regulatory changes and the transfer of pricing power. The mechanism is: if the doomsday narrative only stays in the podcast, it is branding; once confirmed in writing by company researchers, it becomes a contingent liability in the IPO documents and political ammunition for demanding a federal approval agency. Public market shareholders cannot privately bet like private equity LPs on "we might destroy civilization but please buy in at 20 times revenue," so governance discounts will be itemized during the roadshow. Capital continues to flow towards the revenue curve, but pricing power shifts from model capability to who can finish writing risk factors without self-contradicting.
ABAB News · Cognitive Law
Those who warn of risks often apply for risk licenses.
Private equity can buy narratives, public equity must buy litigable sentences.
When moats are written as fear, fear can also open doors for competitors.