Arthur Hayes argues that potential post-IPO share sales by AI companies could pressure technology stocks, tighten credit and weigh on Bitcoin. The chain of events is possible, but the assumptions about share sales and their effect on bank lending remain uncertain.
- Hayes’s figures are estimates, not announced company plans.
- Unlocked shares can increase tradable supply without diluting ownership.
- The $1.245 trillion Anthropic figure is hypothetical, not a funding requirement.
- The cited Federal Reserve data do not confirm a recent decline in bank assets.
Hayes’s thesis depends on several uncertain steps
In “D is for Dilution, ” BitMEX co-founder Arthur Hayes looks at the potential supply of shares tied to possible public listings of Anthropic, SpaceX and OpenAI. He argues that a large number of shares becoming eligible for sale could pressure technology stocks, raise borrowing costs and make banks more cautious about lending.
In Hayes’s scenario, weaker bank lending would reduce the flow of dollar-denominated credit through markets and could weigh on Bitcoin. If problems financing AI companies then grew into a broader crisis, he expects central banks might respond with monetary easing or other measures that could later support Bitcoin.
Every step is conditional. A technology-stock decline does not automatically make banks less willing to lend. Tighter credit does not guarantee that Bitcoin will fall, and a central-bank response is far from certain.
Unlocks, IPOs and the meaning of “dilution”
A share unlock lifts a restriction on existing shares, making them eligible for sale. Before a company lists publicly, its private shares are not ordinary exchange-traded shares. Hayes’s scenario therefore depends on public listings followed by the expiration of restrictions, or another route for shares to be sold. Eligibility does not mean holders will sell.
An unlock can increase the public float, meaning the number of shares available to trade, and create a potential share overhang. But it does not create new shares or, by itself, reduce existing shareholders’ ownership percentages. Dilution generally happens when a company issues new shares and increases its total share count.
An IPO can include new shares sold by the company and existing shares sold by investors. The company gets proceeds from the new shares, while selling shareholders receive proceeds from their shares. The proposed mix for the companies Hayes discusses has not been established here.
That distinction matters to Hayes’s estimate that Anthropic’s early investors could face what he calls 83% “dilution” by June 2027. If the figure refers to existing shares becoming eligible for sale, “potential float expansion” or “share overhang” is more precise. The reported estimate does not provide a share-count baseline that would let readers independently verify what the 83% measures.
What Hayes estimates for Anthropic, SpaceX and OpenAI
For Anthropic, Hayes estimates that approximately $1.245 trillion in additional buying would be needed to absorb the potential share supply without a price decline. His calculation assumes an initial valuation of $1.5 trillion. The figure is hypothetical, not a company-disclosed financing need. It equals 83% of the assumed valuation, but that arithmetic alone does not show how Hayes accounts for the shares, their price or actual investor demand.
Market capitalization is not a cash pool that must be replenished dollar for dollar to keep a stock price steady. Prices are set at the margin. The effect of newly tradable shares depends on how many holders actually sell, the level of demand and how the company is performing.
Crypto.news reported that Anthropic’s registration statement describes restrictions of 180 days for some shares and 365 days for others after an offering. Those restrictions would apply after an offering. They do not mean the shares are already available for public trading.
Hayes estimates that another 8.6 billion SpaceX shares could become eligible to trade by June 2027, roughly half the existing share count. That is his projection, not confirmation that a listing or those share releases will happen. Even if shares become eligible, holders do not have to sell them.
OpenAI is also part of Hayes’s scenario. Crypto.news reported that the company had not publicly filed an IPO registration statement when it covered his projections. Hayes’s assumptions of an initial offering of about 5% of shares, a January listing and a valuation approaching $1.5 trillion are not announced company plans. He did not specify the year for the projected January listing.
The bank-liquidity evidence does not confirm a recent decline
Hayes points to a weekly decline in U.S. commercial bank balance sheets, but acknowledges that one reading cannot establish a lasting trend. The reporting does not identify the date, figure or bank-balance-sheet series behind the observation he cited, so those details cannot be checked.
The Federal Reserve’s H.8 release cited alongside the discussion is dated December 8, 2023. Its table shows total bank assets rising from $23, 114.5 billion for the week ending November 22 to $23, 210.2 billion for November 29. Bank credit also rose over those weeks, from $17, 292.3 billion to $17, 345.2 billion.
Those figures show no recent weekly decline, and they cannot establish whether bank balance sheets have contracted since. A single weekly change in bank assets, up or down, also cannot fully measure dollar liquidity. Liquidity reflects the availability and cost of funding across financial markets.
Hayes’s broader argument is that pressure on share prices could make borrowing more expensive for AI companies, potentially leading to reduced lending and tighter liquidity. But the outcome would depend on companies’ cash flows, debt, collateral and access to other financing. A falling share price alone does not show that banks will pull back.
Bitcoin could face pressure before any policy response
Hayes’s near-term concern is that tighter credit and reduced dollar liquidity could weigh on Bitcoin and other risk assets. His longer-term view is that a serious AI-financing crisis might prompt central banks to make money or credit more available, which could improve conditions for Bitcoin.
No central-bank intervention is certain, and neither its timing nor its form is known. Even if policymakers ease, Bitcoin’s response would depend on broader market conditions and how investors interpret the measures. Easier money can support scarce assets, but it does not guarantee a price rise.
Hayes has previously linked AI-related credit risks to a forecast that Bitcoin could reach $1 million by 2030. That is a forecast, not a reliable planning assumption. Maelstrom, his investment firm, plans to wait for clearer signs of improving liquidity before increasing exposure. That is more cautious than treating the price target as a near-term signal.
Key questions and answers
-
Why does Hayes think AI share releases could affect Bitcoin?
He argues that selling pressure could weaken technology stocks, raise borrowing costs, curb lending and tighten dollar liquidity. These links form a scenario, not a proven chain of cause and effect.
-
What does the $1.245 trillion Anthropic estimate mean?
It is Hayes’s hypothetical estimate, based on an assumed $1.5 trillion valuation, of the buying he says would be needed to absorb potential share supply. It is not an Anthropic funding requirement or a literal cash amount needed to prevent a price decline.
-
Do share unlocks automatically dilute shareholders?
No. Unlocks make existing shares eligible for sale. Dilution generally results from issuing new shares. An unlock can still increase tradable supply, but eligible shares are not necessarily sold.
-
Does the cited Federal Reserve release show a recent drop in bank assets?
No. The H.8 release is from December 8, 2023, and its final weekly observations show increases in both total bank assets and bank credit. It does not verify the unspecified decline Hayes cited.
What investors can watch
Hayes’s thesis can be tested against observable signs: whether the companies pursue public listings, what their actual share-release terms say, how many eligible holders sell, and whether borrowing costs and bank lending change afterward. Recent bank data can help assess credit conditions, but no single weekly balance-sheet figure settles the question of system-wide liquidity.
Until those signals emerge, Hayes’s share-supply estimates and Bitcoin outlook remain scenarios to monitor, not confirmed company plans or reasons to treat a forecast as fact. His view of a possible Bitcoin rally amid economic chaos also depends on macroeconomic conditions, not a guaranteed price catalyst.
