Elice had a subscription ratio of 400-to-1, yet it fell 32% on its first day of trading. It's a profitable company, so why did this happen? Will it affect upcoming AI IPOs?
2026-10-21
The subscription ratio reflected a race for allocation, not demand
Elice's first-day close of 61,500 won is no coincidence. Before the listing, one analyst issued a negative view with a fair value of 62,353 won, about 31% below the 90,500 won offer price (Stockhub, 2026-10). The market fell to almost exactly that fair value in a single day. The moment the offer price was set at the top of the range, the shares had effectively already been sold too expensively.
Even so, institutional book-building demand was 388.8-to-1 (Stockhub, 2026-10). In Korea's IPO market, a high subscription ratio is closer to meaning "we inflated our orders to get a bigger allocation" than "we want to buy at this price." When much of the demand is short-term money that sells its allocation on day one, a higher ratio actually means more selling on the first day. Customer concentration, with about 66% of first-half revenue coming from a single government agency, was also a discount factor.
It's a trend across the IPO market, not a single stock
Elice is not an exception. The 11 stocks listed in the second half of this year had an average return of -25.61% versus their offer prices as of the end of August, and 9 of the 11 were trading below their offer prices (Herald Business, 2026-08-27). In the first half, 14 of 17 listings were also below their offer prices as of early July (Korea Times, 2026-07-06). Analysts said money for IPOs has dried up as the index has become concentrated in large-cap chip stocks.
What sets Elice apart is its industry. The same month, Australian data center company Firmus withdrew its listing, and CoreWeave, which builds data centers with debt, is being asked to pay junk-level yields on its bonds. Connect these three and the structure becomes clear.
- The data center business requires massive upfront investment before generating profits
- When the equity (IPO) funding window closes, funding must come from debt
- As borrowing costs rise, profitability looks worse and the IPO window narrows further
This vicious cycle is the real meaning of a case where "even a profitable company couldn't hold up." Elice posted operating profit of 10.3 billion won in the first half (Stockhub, 2026-10).
Historically, the IPO window has signaled tops first
During the 2000 dot-com bubble and the 2022 rate-hiking cycle, the IPO market froze first and listed company shares followed lower. IPO demand sits at the very front of risk appetite, so a string of IPO failures is read as a sign that the market is putting a lower value on growth stories. Conversely, the point when the IPO window reopened often served as a signal that a bottom was in.
What should investors watch
| Check | Why it matters | Elice case |
|---|---|---|
| Offer price position within the range | At the top, upside has been taken in advance | Priced at the top of the range |
| Customer and revenue concentration | Losing one customer means an earnings shock | One government agency about 66% |
| Gap with analyst fair value | Lower bound for the first-day price | Offer price about 45% above fair value |
| Industry funding environment | Peer withdrawals and bond yields | Firmus withdrawal, weak neocloud bonds |
- It's better to be conservative on AI infrastructure and data center IPO subscriptions for now. Don't read subscription ratios as a demand signal; check the share of institutional lock-up commitments first.
- By contrast, large listings of AI model and software companies that already generate cash flow are a different matter. This selloff reflects doubts about "infrastructure operators building with debt," not about "AI demand."
- The time to increase exposure to this sector is when the IPO window shows signs of reopening, that is, when new data center-related listings hold their offer prices. Until then, avoid chasing related IPOs on their first day of trading.