The Elice IPO was supposed to be an easy win. Korean retail investors put more than 10 trillion won in deposits on the table for a piece of the AI cloud company. Then on 20 October 2026 the stock closed its first day on the Kosdaq 32% below its offer price.
Bloomberg, which first framed the debut as one of South Korea’s worst on record, put the blame on a hype cycle that collided with a global AI selloff. Here is what happened, why it happened, and what to watch next.
Key takeaways
- Elice priced its shares at 90,500 won, the top of its range, and raised about 201.1 billion won (roughly $151 million), the biggest Kosdaq IPO of 2026.
- The retail tranche was 408.26 times oversubscribed. Institutions bid 388.84 times over.
- The stock opened at 75,700 won, fell as low as 56,500 won, and closed at 61,500 won on 20 October 2026, a 32.04% loss.
- The selloff came alongside worries about debt-funded AI data centres, led by concerns over US cloud firm CoreWeave.
What is Elice, and why did investors want in?
Elice started in 2015 as a coding education platform, founded by KAIST doctoral students. It still runs an AI training platform, Elice LXP. But the story that sold the IPO was its second act: GPU cloud services and modular data centres.
In plain terms, Elice is a “neocloud”. That means it rents out access to high-end AI chips by the hour, the same model as CoreWeave in the US. The company says its modular data centres can be built in under three months, against 24 to 36 months for a conventional site. That claim has not been independently verified.
The numbers looked good on paper. According to Seoul Economic Daily, Elice posted 39.5 billion won in revenue and 5.2 billion won in operating profit in 2025. In the first half of 2026 alone it booked 34.4 billion won in revenue and 10.3 billion won in operating profit. A profitable AI infrastructure company is rarer than you might think.
How bad was the Elice IPO debut?
Bad from the opening bell. Here is how the day went against the 90,500 won offer price.
| Moment (20 October 2026) | Price (won) | vs offer price |
|---|---|---|
| Offer price | 90,500 | 0% |
| Opening price | 75,700 | about -16% |
| 9:30 a.m. KST | 63,000 | -30.39% |
| Intraday low | 56,500 | about -38% |
| Close | 61,500 | -32.04% |
For context, Korea’s exchange has let new listings trade between 60% and 400% of the offer price on day one since June 2023, as KED Global reported. That floor sits at 54,300 won for Elice. At its low, the stock came within about 2,200 won of the lowest price the rules allowed.
The company was valued at just over 1 trillion won at the offer price. By the close, roughly a third of that had gone.
Why did a 408x oversubscribed IPO fall so hard?
Oversubscription measures how many people want a ticket. It does not tell you how many want to hold on to it once the doors open. Picture a concert where everyone in line plans to sell their ticket outside the venue. Once they all show up at the same time, the price collapses.
Korea’s IPO market has trained retail investors to expect a first-day pop. Under the wider price band, new listings returned 123% on average on day one in the months after the 2023 rule change. Many subscribers wanted that pop, not the company. When the opening price came in below the offer, the rush to the exit fed on itself.
Timing made it worse. According to News1, worries about CoreWeave’s heavy borrowing and profitability had spread selling across global AI stocks. In Seoul, Samsung Electronics was down 21.2% and SK hynix 31.1% for October by the day Elice listed. TechNode Global, citing Reuters, noted that other planned data centre listings had been pulled or paused as investors worried about borrowing costs.
Our take, and it is opinion: Elice was priced for a calm market and listed into a storm. When AI’s biggest US names are being questioned over debt, a small Korean neocloud priced at the top of its range had little room for error.
Who actually lost money?
Mostly retail subscribers who held through the open. At 408 times oversubscribed, most people received only a small allocation. But anyone who sold at the close lost 29,000 won on every share bought at 90,500 won.
The bigger cost may be trust. The 10 trillion won in deposits shows how much cash Korean households have been willing to park on IPO bets. A debut like this makes the next hot listing a harder sell.
Valuations running ahead of fundamentals is a common theme in AI right now. We looked at a similar dynamic in Town’s $1B valuation four months after launch, and at the fight over AI chip supply in Amazon’s TorchNeuron push against Nvidia.
What happens next
Elice plans to spend the proceeds on regional edge AI data centres, including sites in the Seoul metro area and the Honam region, and on expansion into the United States, Japan and Singapore. Whether it can deliver on those plans matters more than one bad Tuesday.
- Quarterly results: Elice’s first earnings as a listed company will test whether its first-half profit growth holds.
- Lock-up expiries: Pre-IPO shareholders are blocked from selling for set periods. When those end, extra supply can weigh on the price.
- CoreWeave and AI credit: As long as debt-funded AI infrastructure is under a cloud, neoclouds everywhere will trade on that sentiment.
- Korea’s IPO pipeline: Watch whether upcoming tech listings price lower to avoid a repeat.
FAQ
How much did Elice shares fall on their first day?
Elice closed at 61,500 won on 20 October 2026, down 32.04% from its 90,500 won offer price, after touching an intraday low of 56,500 won.
How much did the Elice IPO raise?
About 201.1 billion won (roughly $151 million) from 2.22 million new shares, making it the largest Kosdaq IPO of 2026.
Why did Elice drop despite huge demand?
Many subscribers wanted a quick first-day gain and sold at once. Global worries about debt-funded AI data centres, especially around CoreWeave, made buyers scarce.
If you subscribe to IPOs, read the prospectus and the market mood, not just the subscription ratio. A crowded line tells you about demand for tickets, not for the company. Elice’s first quarterly report is the next real test.
This article is for information only and is not investment advice.
1 comment