Elice IPO: The Price Was Set Before CoreWeave Cracked. The Shares Opened After.

On Oct. 6, the price of the Elice IPO was locked in at 90,500 won a share. On Oct. 14, CoreWeave, the American company that rents out AI chips, lost 15.5% of its value in one day as its bonds sank. On Oct. 20, Elice shares started trading in Seoul. They closed at 61,500 won, 32% below the price buyers had paid.

Three dates, one problem. The price was decided before the bad week. The trading started after it.

A record nobody wanted

Elice raised 201.1 billion won, about $151 million. It was the biggest listing on Korea’s Kosdaq market this year. According to Bloomberg, it was also the largest opening-day drop ever for a Korean IPO of $100 million or more. At its lowest point that day, the stock was down 37.6%.

The demand had looked huge. Ordinary investors put up about 10.3 trillion won, roughly $7.7 billion, in deposits to apply for shares. Their part of the offering was oversubscribed 408 times, ZDNet Korea reported. Before that, 2,367 institutions had bid, and almost all of them were willing to pay the top of the price range.

So how does an Elice IPO that popular turn into a record loss?

Why the price could not move

In the U.S., a company and its bankers usually set the IPO price the night before trading starts. The NYSE’s own guide describes it that way. The price reflects the market of that evening.

Korea runs the steps in a different order. First the final price. Then ordinary investors sign up. Then the shares list. For Elice, that left 14 days between the price (Oct. 6) and the first trade (Oct. 20).

Those 14 days were not quiet. CNBC reported CoreWeave’s 15.5% drop on Oct. 14, as investors worried that AI companies were borrowing too much. That same week, the cost of insuring CoreWeave’s debt against default reached distress levels. The company was carrying more than $51 billion of debt, according to CNBC.

Two other details made a sharp first day easier. First, only about 13.7% of the institutional orders came with a lock-up, a promise not to sell for a set period, Korea Economic Daily reported when the price was fixed. Most institutions were free to sell on day one. We don’t know who actually sold. But most of them were allowed to.

Second, Korea lets a new stock move anywhere from 60% to 400% of its offering price on its first day, under Korea Exchange rules. For Elice, the floor was 54,300 won. The day’s low, 56,500 won, came close to it.

What Elice actually sells

Elice started in 2015 as a coding-education company. Today it is mostly a “neocloud”: a company that buys expensive AI chips, called GPUs, and rents their computing time to others. CoreWeave is the best-known example. Cloud services made up about 7% of Elice’s revenue in 2023. In the first half of 2026, they made up 76%.

That is why CoreWeave’s bad week mattered in Seoul. Both companies sell the same thing. Both pay a lot up front for chips that get old fast, and both need the rent to keep coming in. I looked at how much of the AI build-out runs on borrowed money in AI Debt Bubble: Five Pieces Behind This Week’s Bond Rescue.

Elice’s customer list is short, though. In the first half of 2026, about 22.5 billion won, or 65.6% of its revenue, came from one customer: the government-run National IT Industry Promotion Agency, Korea Economic Daily reported. Total first-half revenue was 34.4 billion won, with an operating profit of 10.3 billion won.

Buying chips has also squeezed its cash cushion. Its current ratio, which compares what a company can turn into cash within a year with what it must pay within a year, fell from 266.9% at the end of 2023 to 52.3% by June 2026, the same paper reported. Below 100% means short-term bills are bigger than short-term assets.

What to watch

Elice now has 201.1 billion won of new money to put into data centers. The first trading day asked one question: can it find paying customers outside the government before its chips grow old? Until private customers show up in the numbers, the market will judge Elice by CoreWeave’s bond yields, not by the 408-to-1 subscription.

My take: A price tag printed in a different market.

Not advice. Just how I see it.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top