Lablup, the Korean company behind the Backend.AI software that schedules GPUs and other AI accelerators, ended its first day on Korea’s Kosdaq market at 14,220 won, 25.16% below its 19,000-won IPO price, according to Korea Exchange data reported by NewsTop and the exchange’s minute prices on Naver Finance. The shares had traded as high as 28,350 won, 49.2% above the offer price, five minutes after the open.
By our arithmetic, the close values Lablup at about 207.1 billion won, about $154 million at 1,344.10 won per dollar, or about 30.8 times its 6,718 million won of 2025 revenue. At the IPO price the same shares were worth about 276.8 billion won, about 41.2 times that revenue.
The key insight: As we read the filings, the first day repeated what the order book had already shown: half of the institutional bids, by count and by shares, came in below the 19,000-won floor, the price was set at that floor, and the stock closed 25.16% under it, which by our arithmetic leaves Lablup at about 30.8 times its 2025 revenue instead of about 41.2 times.
The stock opened at the 19,000-won offer price and reached its high of 28,350 won in the 9:05 minute, Seoul time, according to Korea Exchange minute bars on Naver Finance. It first ended a minute below the offer price at 12:28, touched a low of 13,310 won at 15:01, 29.9% below the IPO price by our arithmetic, and closed at 14,220 won.
Regular-session volume was about 37.3 million shares, by our sum of the minute bars. That is about 2.6 times the roughly 14.57 million listed shares and about 7.2 times the 5,172,746 shares the prospectus says were free to trade at listing, by our arithmetic. The Kosdaq index closed down 4.08% and the KOSPI down 3.59% on the same day, per Naver Finance.
Lablup’s share price by minute on its 3 November 2026 Kosdaq debut: it opened at the 19,000-won IPO price, reached 28,350 won at 9:05 Seoul time, first ended a minute below the IPO price at 12:28, touched 13,310 won at 15:01 and closed at 14,220 won. Korea Exchange minute bars via Naver Finance; percentages are our arithmetic.Animated: Lablup’s share price every five minutes on its 3 November 2026 Kosdaq debut, from the 19,000-won open to a 14,220-won close, 25.16% below the IPO price. Last trade at each five-minute mark from Korea Exchange minute bars via Naver Finance; the intraday high of 28,350 won and low of 13,310 won fall between plotted points. Chart: FourWeekMBA.
What the Order Book Showed
The weakness was visible before trading began. Lablup’s final prospectus of 21 October shows 1,129 institutional bids for 446,593,000 shares, a ratio of 284.53 to 1, and 553 of those bids, 49.0% of the count and 49.9% of the shares, were priced below 19,000 won, the bottom of its 19,000-to-22,000-won range. The company and lead manager NH Investment & Securities set the price at 19,000 won.
Retail demand was thin. Lablup’s offering results report, filed on 27 October, shows retail investors asked for 1,936,220 shares against 545,000 allotted, about 3.55 times by our arithmetic, and 3,571,220 shares across all groups for the 2,180,000 shares sold. The offering raised 41.42 billion won, about $30.8 million.
What Buyers Are Paying For
Lablup reported revenue of 3,003 million won in 2023 under Korean GAAP, then 5,932 million won in 2024 and 6,718 million won in 2025 under K-IFRS, and says first-half 2026 revenue rose 58.2% from a year earlier, according to the prospectus. It recorded a first-half net loss of 2,762 million won but positive operating cash flow of 316 million won, which it attributes mainly to non-cash costs, including depreciation of about 0.5 billion won and a derivative valuation loss of about 2.9 billion won on redeemable convertible preference shares.
Revenue is concentrated. Two customers accounted for 68.6% of 2025 revenue, and the customer count fell from 32 in 2023 to 28 in the first half of 2026, the prospectus says. It lists Samsung Electronics, Hyundai Mobis, KT, KT Cloud and Shinhan Bank among its customers. Facilities funding of 18,677,866,000 won, about 45% of the offering by our arithmetic, sits alongside R&D and overseas expansion in the use-of-proceeds plan, which the prospectus says it expects to spend in earnest from 2027.
The Partner Named as a Risk
The prospectus says Lablup was the first company in Asia-Pacific selected as an NVIDIA DGX-Ready Software Partner, a designation Nvidia gives to software it has validated on its DGX systems, and names Nvidia, Intel, Dell Technologies and VAST Data as partners.
The same document lists the chipmaker as a source of risk. It says Nvidia and other AI chip companies keep upgrading their own software for running their hardware, and that major cloud providers are strengthening their own AI infrastructure management. In our translation, it warns that “if AI infrastructure management functions are internalised into hardware or cloud services, demand for independent software such as ours may decrease”.
It cites Run:ai, a GPU orchestration company that Nvidia acquired in December 2024, as an example of a specialist absorbed into a chipmaker’s ecosystem.
Business Pill · Concentration
An evergreen Business Engineer explainer on concentration: when a few big pieces make most of the total, such as a few customers making most of a company’s revenue. It teaches the general idea only and does not discuss Lablup.
As we read the prospectus, Lablup’s position depends on staying independent of the hardware it manages: the same document that cites its NVIDIA DGX-Ready Software Partner status warns that chipmakers and cloud providers are building the same management functions into their own products.
As we read the subscription data, demand thinned at each step: 284.53 to 1 in the institutional demand forecast, about 3.55 times among retail investors by our arithmetic, and a first-day close below the offer price.
Lablup prospectus, DART, 19 October 2026 (our translation)
“the offering price does not represent a market price traded in the market after the IPO”
Three Implications
FOR KOREAN AI LISTINGS Elice closed 37.02% below its IPO price on 3 November, NewsTop reported, and Lablup closed 25.16% below its own on its first day.
FOR INDEPENDENT AI INFRASTRUCTURE SOFTWARE Lablup’s prospectus says demand for independent software such as its own may decrease if management functions are built into hardware or cloud services.
FOR LABLUP’S CUSTOMERS Two customers accounted for 68.6% of 2025 revenue, according to the prospectus, which lists Samsung Electronics, KT and Shinhan Bank among its customers.
The framework puts it this way: “That is why the AI economy is not a list of companies racing each other. It is a layered industrial stack — and at any given moment, the binding constraint sits at a different layer than it did six months ago.”
As we read it, Lablup sells into the layer between the accelerators and the applications that run on them, and its prospectus names the risk that the layers on either side, chipmakers and cloud providers, absorb that function. The framework does not discuss Lablup or this listing.
What Is Not Established
The filings do not say why institutions bid below the range, and the trading data do not show who sold on the first day. Lablup told reporters it saw market volatility and caution toward the AI sector, rather than concerns about the company, behind the demand-forecast result, NewsTop reported; we did not reach the company. Market values and multiples here are our arithmetic on a share count implied by Naver Finance’s market value, and they use 2025 revenue, not the company’s forecasts. One trading day says nothing settled about the business.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
Lablup priced at the bottom of its range after half of the institutional order book bid lower, and its first session went from 49.2% above the offer price to a 14,220-won close, 25.16% below it. By our arithmetic the market now values the company at about 30.8 times 2025 revenue, down from about 41.2 times at the IPO price, while its prospectus names its best-known partner, Nvidia, among the sources of competitive risk.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Offering terms, demand-forecast and subscription figures, financials and risk factors are from Lablup’s filings on DART (in Korean; English renderings are ours). Prices are Korea Exchange data via Naver Finance and NewsTop’s close report. Market values, multiples and dollar conversions (at 1,344.10 won per dollar, Hana Bank via Naver Finance, 3 November) are our arithmetic. We did not contact Lablup.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
