AI Hedge Fund SALP Blows Up: How 4x Leverage Triggered a Wall Street Liquidation Hunt
Taylor Wilson
SALP, the AI-focused hedge fund founded by former OpenAI staffer Leopold Aschenbrenner, was force-liquidated after a 25% drawdown wiped out most of its equity on 4× leverage — and Citadel swooped in at a steep discount, locking in billions in instant paper gains.
A $45 billion fund — how does one pullback blow it up?
SALP held roughly $45 billion in total assets: about $10 billion in private Anthropic equity and around $30 billion in cash. This means → the fund's real risk-absorbing capital — its net equity — was roughly $35 billion.
The problem was leverage. Through its prime broker, SALP ran 4× leverage, inflating total position value to about $120 billion. In plain terms = every $1 of its own money was supporting nearly $4 in exposure.
When its public-market stocks dropped roughly 25%, net equity cratered from $35 billion to about $5 billion. The prime broker seized the portfolio and began forced liquidation to protect itself.
Why did rivals speed up the collapse instead of helping?
According to former hedge-fund manager Martin Shkreli, word of SALP's margin call leaked days before it became public. Several funds spotted the signal and pre-emptively sold overlapping positions while shorting SALP's known holdings.
Shkreli calls this Wall Street's "Darwinian rule": once the market learns a fund is being forced to sell, competitors dump shared positions and build short bets — not standing by, but actively twisting the knife.
This reflects a brutal logic: a large forced liquidation creates guaranteed selling pressure, and front-runners profit by riding that certainty.
How did Citadel pocket billions from the wreckage?
During the forced sale, Jane Street, Millennium Management, and Citadel were invited to a closed-door auction. Citadel ultimately bought SALP's equity book at a 20% to 50% discount.
By unwinding the positions without disrupting the market, the buyer locked in roughly $3–4 billion in immediate paper gains. In plain terms = someone is forced to fire-sell at liquidation prices; you buy wholesale and sell at retail.
Shkreli noted that Citadel played a similar role during the Amaranth natural-gas blowup and other crises — further cementing its reputation as Wall Street's "buyer of last resort."
Why not just sell on the open market?
Dumping a massive position into the order book instantly signals intent. This means → quant algorithms and rival desks detect the selling pressure and front-run with shorts, driving the price down further.
Prime brokers typically use discreet inquiries or package the book for bulk sale — the goal is to sever risk quickly, not wait for a rebound.
This kind of concentrated large-block selling creates weeks of sustained downward pressure — far more damaging than a single-day crash.
AI companies are doing great — so why did the stock still crash?
Shkreli stressed: short-term price action is not set by fundamentals. It is set by the 5% of market participants trading on leverage — the marginal buyers and sellers. In plain terms = the people constantly adding and cutting positions, the ones actually moving prices, are the ones trading with borrowed money.
When the most fragile capital panics first, any small negative signal — say, one tech giant announcing a modest capex trim — can cause the entire trade structure to collapse.
He cited the Kelly Criterion — a formula for optimal bet sizing — noting that most traders run positions 2 to 10 times the mathematically justified size. Even with a 60/40 edge, chronic over-betting guarantees eventual ruin.
What does this blowup mean for the AI trade?
Shkreli compared this event to the collapses of Long-Term Capital Management (LTCM) and Amaranth, calling it one that "absolutely ranks among the biggest."
This reflects a now-visible tail risk in the AI compute and infrastructure trade: leveraged capital has been fully exposed. This means → the market has no shortage of AI bulls — what it lacks is capital that can ride out volatility without borrowing.
The key test ahead: whether AI-sector valuations can stand on their own without the support of highly leveraged marginal buyers.
Content is for reference only, not financial advice.