Key facts at a glance

• Elice shares closed at 61,500 won on their first day on the Kosdaq, down 32% from the IPO price of 90,500 won.
• The South Korean AI cloud and modular data-centre company raised 201.1bn won, or about $151m, by selling 2.22 million new shares at the top of its indicated range.
• The stock opened at 75,700 won, touched a low of 58,000 won and closed at 61,500 won.
• The debut was the worst on record for a Korean IPO raising $100m or more, according to market data.
• The retail tranche was 400 times oversubscribed, with individual investors placing about 10.3tn won in subscription deposits.
• Retail investors received 25% of the offering, worth about 50.3bn won, and lost about $12m based on the closing price, according to calculations cited in market reports.
• Elice's IPO was the biggest on the Kosdaq this year; only K Bank's $337m listing in March raised more in Korea.
• The benchmark Kospi closed 4% lower, with Samsung Electronics down 3% and SK Hynix down 4%.
• Elice was founded in 2015 by three KAIST doctoral students as a coding education platform and later moved into AI cloud infrastructure.
• First-half revenue rose 200% to 34.4bn won, with operating profit of 10.3bn won, according to its prospectus.
• Vertex Ventures and Altos Ventures back the company.
• Chief executive Jae Won Kim said the IPO proceeds will fund more AI data centres in South Korea and abroad.

A bruising first trading day

Shares in Elice, a South Korean AI cloud and modular data-centre company, fell 32% on their first day of trading in Seoul on Tuesday, closing at 61,500 won after pricing at 90,500 won. It was the worst debut on record for a Korean initial public offering of $100m or more, according to market data, a striking result for a company that had been marketed as a profitable participant in the global AI infrastructure buildout. The first-day slump was a sharp reversal for a business that had positioned itself at the centre of demand for AI computing capacity, and it showed how quickly sentiment can turn in a market where retail investors had aggressively chased new listings.

The listing raised 201.1bn won, or about $151m, by selling 2.22 million new shares at 90,500 won each, the top of its range. The stock opened at 75,700 won on the Kosdaq, fell as low as 58,000 won and closed at 61,500 won. Elice's IPO was the biggest on the Kosdaq this year; in Korea, only K Bank's $337m listing in March raised more. The size and top-of-range pricing suggested that underwriters expected strong demand, but the first-day drop showed that demand did not survive the transition to secondary trading. For a newly listed company, that gap between subscription appetite and aftermarket support can be painful because early buyers set the tone for trading and can amplify selling pressure if they retreat.

A 400-times oversubscribed flop

Demand had looked strong before the listing. The retail tranche was 400 times oversubscribed, and individual investors put about 10.3tn won into subscription deposits. Most of them received few shares. Retail investors got 25% of the offering, worth about 50.3bn won. Based on the closing price, they lost about $12m on the day. The disparity between the oversubscription and the aftermarket performance is central to the Elice story: a heavily oversubscribed book can reflect speculative appetite rather than durable institutional support, especially when investors are chasing a hot sector.

“Retail investors may have blindly believed that IPO stocks would always go up,” said Jo Hyunik, an investment division manager at Iroom Investment Advisors in Seoul. The comment captures frustration among market participants who warn that retail demand for new listings can become disconnected from valuation and liquidity. In Elice's case, the retail allocation was small relative to the deposits committed, magnifying the loss relative to the capital tied up in subscriptions. Investors who committed large sums to the offering received only a fraction of the shares they sought, and the shares they did receive fell sharply from the IPO price.

The benchmark Kospi closed 4% lower on the day. Samsung Electronics fell 3% and SK Hynix 4%. The broad decline amplified pressure on new listings and made it harder for Elice to hold its IPO price. A weak tape can be especially punishing for a newly listed technology company because early buyers are more sensitive to sector sentiment and less willing to build positions during a selloff. The market backdrop also complicated the company's effort to present itself as a profitable exception in an otherwise volatile corner of the AI trade.

From coding classes to AI cloud

Elice was founded in 2015 by three KAIST doctoral students as a coding education platform. It later moved into AI cloud infrastructure, selling servers, liquid cooling, networking and software. It now builds modular data centres for dense clusters of AI chips. The shift from education to infrastructure reflects the broader transformation of the AI market: demand has moved from software and models toward the physical systems that train and run them. Modular data centres are designed to be deployed quickly and configured for high-density computing, a requirement that becomes more important as AI chips consume more power and generate more heat.

Unlike many AI infrastructure startups, Elice is profitable. Its prospectus showed first-half revenue up 200% to 34.4bn won and an operating profit of 10.3bn won. Vertex Ventures and Altos Ventures back the company. Profitability is notable in a sector where capital-intensive data centre buildouts often produce losses and require repeated fundraising. It also gives Elice a different investment case from loss-making neoclouds that have relied on debt and equity to fund GPU capacity. Still, profitability in one half does not guarantee that a company can maintain margins while expanding into new data centres and competing with larger infrastructure providers.

The company's pitch to investors rested on the idea that AI infrastructure demand will remain strong and that Elice can capture a slice of it with modular, full-stack offerings. Its products span servers, cooling, networking and software, which means it is not simply renting out computing power. That broader stack could help it differentiate itself from pure GPU rental businesses, but it also exposes the company to execution risk across multiple hardware and software categories. The first-day trading reaction suggests that investors are not yet willing to award a premium for that strategy while the neocloud sector is under pressure.

What the IPO proceeds will fund

“We’re going all-in on becoming a full-stack AI infrastructure leader globally,” said chief executive Jae Won Kim. “This IPO is about giving ourselves the financial firepower to execute.” Kim said Elice will spend the money on more AI data centres in South Korea and abroad. The plan puts the company in direct competition with larger infrastructure providers, but it also ties its fortunes to a sector where investor sentiment has become volatile. The proceeds give Elice capital to build capacity, but the market will judge whether that capacity can be filled profitably as competition intensifies.

The expansion plan also comes at a delicate moment for AI infrastructure financing. Data centres require significant upfront investment, and returns depend on utilization rates, power costs and the ability to price computing capacity at attractive levels. Companies that expand too slowly may miss demand, while those that expand too quickly may face idle capacity if the AI investment cycle cools. Elice's profitable first half gives it a stronger starting point than many peers, but the first-day share price decline shows that investors are focused on the risks as much as the growth story.

Caught in the neocloud selloff

Elice is a neocloud, one of the companies that rent out AI computing power. Investors are losing faith in the group. CoreWeave's shares and bonds fell sharply last week over its debt. Australia's Firmus dropped its planned ASX float, and Nscale faces investor pushback on its IPO valuation. Those developments weighed on the broader neocloud trade and provided a difficult backdrop for Elice's listing. When investors begin to question the balance sheets and business models of larger players, new entrants can be punished even if their own financials look different.

The neocloud model can be attractive when demand for AI compute is rising and financing is cheap. It becomes more fragile when investors focus on debt loads, utilization rates and the risk of oversupply. Elice's modular data-centre approach is aimed at dense clusters of AI chips, a niche that requires sophisticated cooling and power management. But even a profitable company can be caught in a sector-wide de-rating if investors decide that AI infrastructure spending has outpaced near-term returns. The selloff in related companies therefore became a direct headwind for Elice's debut, overshadowing the company's revenue growth and operating profit.

“Sentiment toward AI infrastructure has been highly volatile and often disconnected from underlying fundamentals,” said Albert Yong, chief investment officer at Petra Capital Management. That volatility was visible in Elice's first-day trading: the stock opened below its IPO price, fell further and closed near the low end of its range. For IPO investors, the lesson is that a strong subscription book does not guarantee a strong debut, especially when the sector is under pressure. The Elice listing also illustrates how quickly a company can move from being viewed as a growth opportunity to being judged against the weakest parts of its peer group.

Kim Dojoon, chief executive of Zian Investment Management, said Elice's finances differ from CoreWeave's and that investors need not see it as “particularly risky”. He said it was not yet clear whether Elice could stay as profitable. The debate over Elice now centres on whether its first-half profitability can persist as it spends IPO proceeds on new data centres and expands abroad. The first-day slump occurred despite the company's profitable first half and a retail book that was 400 times oversubscribed.

Source:TNW | Investors-funding News