Anthropic IPO tests Arthur Hayes's $1.245 trillion AI selling thesis and its Bitcoin liquidity link
Anthropic's IPO puts an AI selling theory to the test
Anthropic has announced a proposed IPO, an initial public offering where a private company sells shares to the public for the first time. The offering creates a real possibility of insider selling, but CryptoSlate reports that the share-release schedule does not make a drain on Bitcoin liquidity inevitable.
The report examines whether the offering supports a thesis from former BitMEX executive Arthur Hayes, who argues that falling AI share prices could trigger a wider credit squeeze that eventually reduces the dollar liquidity supporting crypto.
Key numbers from the offering
- Anthropic's Nov. 10 announcement proposes 825 million Class A shares: 801,351,354 offered by the company and 23,648,646 by existing stockholders.
- Anthropic receives no money from the existing holders' sales; those sellers keep the proceeds.
- The preliminary price range is $70 to $78 per share.
- Underwriters have an option to buy another 123.75 million shares.
- Under the preliminary IPO and a concurrent private placement, legal Class A and Class N shares together rise from about 11.9 billion to over 13 billion, an increase of roughly 9.66%.
Hayes's $1.245 trillion dilution estimate
In a Nov. 5 essay, Hayes argues that insider sales would weaken AI stocks, make borrowing more expensive or unavailable, slow bank balance sheet growth, and reduce the dollar liquidity that supports crypto prices. He estimated 83% dilution by June 2027, meaning existing owners' stakes would shrink sharply as new shares are issued.
He calculated that roughly $1.245 trillion in new capital would be needed to absorb potential selling at a hypothetical $1.5 trillion IPO valuation. CryptoSlate notes this is a theoretical figure. The actual outcome depends on what holders sell, the prices buyers pay, and demand for the stock.
How the staged share releases work
New company shares increase the total share count and raise money for Anthropic. Secondary shares from existing holders already exist; selling them transfers ownership and pays the seller. A later unlock only removes a restriction on selling, without creating new shares or forcing anyone to sell.
According to the preliminary resale disclosures, about 316 million additional shares become eligible for sale from the first trading day. Further quantities include 299 million at each of 27 and 48 days after the US prospectus date, and 349 million at each of 62 and 76 days.
Restrictions differ by holder group. Non-founder personnel and certain other holders face an end at the earlier of a first-quarter results trigger or a 180-day endpoint. Co-founders and certain larger holders follow an earlier third-quarter results trigger or a 365-day endpoint, with staged early-release exceptions.
About 15.5 billion Class A-equivalent securities are covered by lock-ups, a lock-up being a period when holders cannot sell. This pool includes Class N shares, awards, and potential note conversions, and is larger than the freely tradable float.
What bank credit data shows so far
The Federal Reserve's Nov. 13 H.8 release shows bank credit rising $58 billion in the later week. The release does not identify any AI-specific lending, so the credit side of Hayes's thesis has not yet shown the squeeze he describes.
What is confirmed and what is not
Confirmed: the share counts, price range, underwriter option, release schedule, and the Fed's bank credit figure all come from Anthropic's announcement and the Fed release.
Not confirmed: Hayes's $1.245 trillion figure is his own estimate based on a hypothetical valuation. Actual sales, tighter credit, and any transmission from AI share selling to Bitcoin liquidity all remain conditional. CryptoSlate's own illustration notes the liquidity link in the thesis is unproven.
Why this matters for crypto
If Hayes is right, AI insider selling could tighten credit and reduce the dollars flowing into crypto. Anthropic's offering is the first testable step in that chain. So far, the evidence is mixed: the share supply is set to grow, but the Fed data shows bank credit rising, not contracting, and no AI-specific lending has been identified.