Recommended Articles
Deals & Exits

Elice Group Crashes 28% on KOSDAQ Debut Despite a 389x Oversubscribed IPO

By Angeled · October 20, 2026 · Deals & Exits
Elice Group Crashes 28% on KOSDAQ Debut Despite a 389x Oversubscribed IPO
Elice Group Crashes 28% on KOSDAQ Debut Despite a 389x Oversubscribed IPO

The listing that went wrong

Elice Group, a South Korean AI infrastructure company, suffered one of the worst first-day performances for a major Korean IPO in years when it debuted on the KOSDAQ on October 20, 2026. The stock opened at 75,700 won — 16.4% below its 90,500 won offer price — and kept sliding, closing roughly 28% below the offer (intraday lows touched 61,000 won, a 32.6% discount), according to Reuters and Investing.com.

That makes Elice's debut the worst opening-day performance for a South Korean IPO raising at least $100 million since at least 2022 — and it was the biggest KOSDAQ listing of 2026 by proceeds.

The paradox: record demand, record decline

What makes the collapse striking is how hot the IPO looked on paper. Elice priced at 90,500 won — the very top of its 70,400–90,500 won range — raising 201.1 billion won (about $150 million) through 2.22 million new shares. Institutional investors bid for 388.84 times the shares on offer, and retail investors oversubscribed the offering 408.26 times, placing more than 10 trillion won in deposits.

The company's fundamentals were not obviously the problem either. Elice reported first-half 2026 revenue of 34.4 billion won — more than triple the prior-year figure — with operating profit of 10.3 billion won and EBITDA of 23.2 billion won, a 68% margin. For the full year 2025, revenue was 39.5 billion won with 5.2 billion won in operating profit. This was no pre-revenue moonshot.

What Elice does — and what the money is for

Elice began in 2015 as an AI education platform and has expanded into AI cloud services, GPU-based data centers, and AI software for businesses. Its flagship pitch is its Portable Modular Data Centers (PMDC): modular facilities that can reportedly be built in about three months — versus two to three years for traditional data centers — at up to 40% lower total ownership cost.

The company said it would use the IPO proceeds primarily to build data-center capacity in Korea and expand to the United States, Japan, and Singapore.

Why the tape turned

The stock market's mood changed between the book-building and the listing. In the intervening days, CoreWeave's bonds traded at yields above 15%, hedge funds unwound AI positions for six straight sessions, and Australia's Nvidia-backed Firmus shelved its planned $5 billion IPO, citing scrutiny of AI valuations and the sustainability of infrastructure investment. The KOSPI itself fell 2.5% on listing day, with SK Hynix down 3.5%.

The pattern suggests the market's skepticism is no longer about whether AI demand exists — it is about the returns on the massive capital expenditure pouring into AI data-center infrastructure, and about who ultimately pays for it. Oversubscribed books reflect past expectations; the opening-day tape reflects repriced ones.

A cooling global AI trade

Elice's flop lands amid a broader reassessment of AI valuations. Markets in the US, Europe, and Asia have shown investors pulling back from richly priced AI stocks, and a flagship new listing stumbling this badly tends to dampen enthusiasm for other AI companies queued to go public. The episode shows how quickly the AI trade can flip from scarcity bidding to discount pricing — within a single week.

Why it matters for founders

The Elice debut is a case study in how the IPO window rewards timing, not just fundamentals. A profitable, fast-growing company priced at the top of its range still fell 28% because macro sentiment about its sector rotated between book-building and listing day. For founders planning public exits in AI infrastructure: lock in demand late, watch the tape of comparable names (CoreWeave's bond yields were the canary here), and remember that oversubscription books are a lagging indicator of sentiment. Private valuations that extrapolate from peak AI multiples are on notice.

Why it matters for investors

The 389x-book-to-28%-decline whiplash is a warning about IPO allocation mechanics in AI-adjacent names: frenzied retail deposit bidding can coexist with institutional second thoughts on day one. The deeper signal is about AI infrastructure capex itself — investors are now demanding proof that data-center spend converts into durable earnings, not just capacity. Price new AI infrastructure issues on return-on-invested-capital discipline, not on demand-for-compute narratives, and size positions for the volatility that repricing cycles bring.

Angeled.com connects founders with investors.

Sources: Reuters (via Investing.com, Cryptopolitan, Hokanews), Crypto Briefing, Trade2Spike, AltcoinDaily

Read Next