Market-structure
A single thinly-traded share in a Korean pre-market session moved a tokenized SK Hynix perpetual 19% on Hyperliquid and forced $60 million of positions closed in two minutes, exposing what always-on crypto-native equity trading actually depends on
Trade.xyz's oracle worked exactly as designed, relaying a real but outsized off-exchange print into a mark price with no depth behind it, and is now discretionarily covering the resulting losses while it rebuilds pricing around its own order books.
The most important detail in this story is that nothing malfunctioned. Trade.xyz's price oracle did precisely what it was built to do: take a real, executed trade and relay it into the mark price used to value open positions. The problem was that the real trade in question was a single share, traded in South Korea's thin pre-market session, roughly 30% below the prior close. A functioning system faithfully propagated a bad signal, and $60 million in leveraged long positions got liquidated in two minutes as a direct, mechanical consequence.
This is the structural tension sitting underneath every pitch for AI-native, always-on, tokenized equity trading, including the London Stock Exchange's own 24/5 venue launched this same month. Extending trading hours and connecting AI agents directly to markets only works as well as the price discovery feeding it, and price discovery outside a name's primary listing hours is, by definition, thinner than during it. A single Korean pre-market print should not be able to set the reference price for a $60 million book of leveraged crypto derivatives, but in a 24-hour tokenized market stitched together across venues and time zones, there is no guaranteed deep, continuously-traded reference to fall back on instead.
SK Hynix being an AI chipmaker sharpens the point rather than being incidental to it. The underlying asset sits inside one of the most actively AI-thematic trading names on the planet right now, the kind of name that AI trading agents and momentum-following systems are most likely to be concentrated in. A market structure fragility that would be a minor footnote in a quiet, unloved stock becomes a $60 million liquidation event specifically because AI-adjacent names are where crowded, leveraged, always-on trading activity currently concentrates.
Trade.xyz's response, a one-time discretionary reimbursement plus a promise to weight its own order books more heavily in future pricing, is a reasonable immediate fix and also an admission that the original design assumed liquidity conditions that do not actually hold around the clock. Regulators in South Korea are reportedly already looking at leverage limits in response. The venue got lucky that it had the balance sheet to make holders whole this time; the underlying lesson, that always-on markets need pricing infrastructure built for their thinnest hours rather than their busiest ones, does not go away once this particular bill is paid.
Read the original: CoinDesk - A single thinly-traded share in a Korean pre-market session moved a tokenized SK Hynix perpetual 19% on Hyperliquid and forced $60 million of positions closed in two minutes, exposing what always-on crypto-native equity trading actually depends on. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.