AI infrastructure management and operations software company Lablup is set to target the global market with its upcoming KOSDAQ listing. The company plans to expand its business by leveraging an integrated platform that encompasses GPU resource management, AI model training, inference, serving, and token management, all while maintaining a growth strategy based on recurring revenue.
On October 19, Lablup held a press conference at the Conrad Seoul in Yeouido, where it outlined its growth strategy and future business plans following its IPO. Founded in 2015 as a platform for academia, Lablup has evolved its technology from research-focused computational efficiency to managing large-scale computational resources. Its proprietary AI infrastructure operations platform, Backend.AI, is central to supporting the establishment and operation of AI infrastructure for businesses and institutions.
Backend.AI integrates the entire process of AI infrastructure operations, from managing AI accelerator resources like GPUs and NPUs to model development, training, inference, and serving. Recently, the company has expanded its business scope to include AI agents and token management.
Lablup emphasizes its operational experience accumulated in real industrial settings as a competitive advantage. The platform is currently in use at over 120 sites, including Samsung Electronics, Hyundai Mobis, KT, Shinhan Bank, and Samsung Medical Center, with experience managing over 2,000 GPUs at a single site. According to Lablup, it has increased GPU utilization from 30% to 85% while reducing GPU budgets by 40%.
The company has also built a track record in operating large-scale AI models. It has participated as an infrastructure partner in the government’s independent AI foundation model project, supporting model training using over 2,500 NVIDIA B200 and H100 GPUs since the first phase, demonstrating its capabilities in large-scale infrastructure operations.
Furthermore, Lablup is expanding its collaborations with global companies. Since winning NVIDIA's AI Conference Inception Award in 2018, it has maintained a partnership and was named the first 'DGX-Ready Software Partner' in the Asia-Pacific region in 2021. It is currently conducting joint performance validation of AI accelerators with Intel and recently participated as an Intel GPU D-Series ecosystem partner at the 2026 OCP Global Summit.
In terms of business model, Lablup features a growth structure based on recurring revenue. The company operates an annual subscription 'Land-and-Expand' model, where the expansion of clients' GPU resources and the replacement of next-generation GPUs lead to additional license revenue. As of last year, recurring revenue accounted for 94.7% of total revenue, with a recent two-year average net revenue retention rate of 131.9%.
Following its listing, Lablup plans to aggressively enter the global AI infrastructure market, particularly targeting the 'Sovereign AI' market, where demand for data sovereignty and independent AI infrastructure is growing. The company aims to focus on global neo-cloud providers as key customers and expand its business into North America, Europe, Japan, Southeast Asia, and the Middle East through partnerships with local resellers in the U.S., U.K., and Singapore.
Lablup is also pursuing business diversification through new products. It has developed AI that extends its operations from data center-focused AI infrastructure to PCs and workstations, and has unveiled the 'Continuum Suite,' which supports multi-AI model routing and token usage and cost management. In the future, it plans to evolve this into the 'Lablup Token Factory Suite,' a platform that encompasses infrastructure management, model inference, and token operations.
Shin Jeong-kyu, CEO of Lablup, stated, "Over the past decade, we have validated our technological competitiveness based on our unique AI infrastructure operations technology and collaboration experience with global companies. With our KOSDAQ listing, we aim to expand our global business and broaden our scope from GPU-centric infrastructure management to AI model and token operations, growing into a global AI infrastructure software company."
Lablup plans to offer a total of 2.18 million shares, with a proposed price range of 19,000 to 22,000 won, aiming to raise between 41.4 billion and 48 billion won. Demand forecasting for institutional investors will be conducted until today, with general subscriptions scheduled for October 22-23. The company is set to list on November 3, with NH Investment & Securities serving as the lead underwriter.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.