Market / Stocks

Down 30 percent on debut, up 30 percent the next morning: Elice cannot decide what it is worth

South Korea's AI infrastructure firm raised $151 million in an IPO investors queued hundreds of times over to join, then lost nearly a third of its value on day one and won most of it back on day two, still below its offer price.

Seoul, South Korea · October 21, 2026

There is a particular kind of embarrassment reserved for a stock that everyone fought to buy and nobody wants to hold. Elice Group, a South Korean AI infrastructure and cloud company, lived it on Tuesday. Its listing raised 201.1 billion won, about $151 million, priced at 90,500 won a share at the top of its range after institutional orders came in at roughly 389 times the shares on offer and retail demand ran to about 400 times. Then trading started, and the stock fell more than 30 percent on its debut, closing about a third below the price all those investors had queued for.

Then Wednesday morning reversed the film. Elice surged about 30 percent, hitting its daily limit at 79,900 won in the reporting of the rebound, as Asian technology stocks recovered on the strength of SpaceX's results and returning appetite for AI infrastructure names. Even after a move that size, the shares sat below the 90,500 won offer price. Down 30, up 30, still underwater: in two sessions the market managed to say the IPO was overpriced, oversold and worth chasing, in that order, without the company itself changing at all.

The wider scene explains some of the violence. Elice was priced earlier in October, before the latest selloff in AI stocks, and it has landed in a month when the data centre listing trade has been falling apart in public: Blackstone backed AirTrunk shelved a Singapore flotation, Nvidia backed Firmus gave up on its Australian listing, and SoftBank backed SB Energy delayed a huge Nasdaq debut. Investors are asking harder questions about who funds AI infrastructure, on what terms, and when the spending turns into profit. A small Seoul cloud company with real growth, its first half revenue roughly tripled on the figures carried with the debut reporting, walked into that mood and got priced by it twice in two days.

Our call is that Elice is a warning about the queue, not about the company. Subscription multiples in the hundreds measure how rationed a deal is, not what the shares are worth once everyone can sell, and the gap between those two numbers is where debut losses live. Until an AI infrastructure listing prices, opens and holds its price for a full week, every new deal in this sector will be priced with a sceptic's discount built in, and bankers will have to sell that discount to founders as the cost of listing in this market. Elice's two days are the new reference point: the demand was real, and it meant almost nothing.