▶ From Unified GPU and NPU Management to AI Training and Inference on a Single Platform…Subscription Licensing Drives Sustainable Growth
▶ Proven Capability to Operate GPU Clusters of More Than 2,000 GPUs at a Single Site Across Key Industries Including Finance, Cloud, and Manufacturing
▶ Offering 2.18 Million Shares…Institutional Bookbuilding Scheduled for September 28–October 2

<August 25, 2026> Lablup, an AI infrastructure management and operations software company led by CEO Jeongkyu Shin, announced today that it has filed a securities registration statement with the Financial Services Commission, officially launching its initial public offering process for a KOSDAQ listing.

Lablup plans to offer 2,180,000 shares, with an indicative price range of KRW 20,000–23,000 per share. The anticipated offering size is therefore KRW 43.6 billion–50.1 billion. Institutional bookbuilding will take place from September 28 through October 2, followed by the retail subscription period on October 12–13. NH Investment & Securities is the lead underwriter.

Founded in 2015, Lablup specializes in AI infrastructure management and operations software that improves the efficiency of high-performance computing resources, including GPUs. Its flagship product, Backend.AI, provides a unified platform for managing GPUs and AI accelerators required for AI model development, training, deployment, and operations, while automatically allocating resources according to individual users and workloads.

A key technological differentiator is Lablup’s proprietary GPU partitioning and virtualization technology. It partitions high-cost GPUs at the workload level, enabling multiple users to utilize them concurrently and thereby increasing resource efficiency. In actual enterprise deployments, GPU utilization has increased from 30% to 85%—a gain of approximately 55 percentage points, or 2.8 times.

The technology has also been validated in large-scale AI infrastructure environments. As the infrastructure partner for Upstage’s consortium in the government-led “Sovereign AI Foundation Model Project,” Lablup supported the operation of GPU clusters comprising more than 500 GPUs in the first phase and more than 880 GPUs in the second phase. Having also advanced to the third phase, the company has demonstrated its ability to reliably operate GPU environments with more than 1,000 GPUs.

Backend.AI is also hardware- and environment-agnostic. It manages heterogeneous accelerators—including GPUs from NVIDIA, AMD, and Intel as well as domestic and international NPUs—on a single platform, while supporting on-premises, public-cloud, and hybrid environments across major operating systems. In addition, rather than relying on one-time implementation revenue, the company uses a subscription licensing model designed to grow alongside customers’ GPU expansion and broader AI adoption.

Backend.AI is currently deployed at more than 120 sites in Korea and overseas. Its customer base includes major enterprises and institutions such as Samsung Electronics, Hyundai Mobis, KT, kt cloud, Shinhan Bank, LIG D&A, Samsung Medical Center, the Korea Internet & Security Agency, and the Artificial Intelligence Industry Convergence Agency. The company is also expanding internationally, with customers in the United States, the United Kingdom, Brazil, and Thailand.

Lablup’s revenue grew from KRW 3.0 billion in 2023 to KRW 6.7 billion in 2025, more than doubling in two years. The company also recorded operating profits for two consecutive years, in 2024 and 2025. Lablup plans to use IPO proceeds to advance R&D, expand globally, and strengthen its infrastructure. In particular, it aims to broaden Backend.AI support in line with the increasingly diverse AI accelerator ecosystem—including GPUs and NPUs—and accelerate expansion into North America, Japan, Southeast Asia, Europe, and the Middle East.

“AI infrastructure is shifting from an era focused on how many GPUs a company can secure to one focused on how effectively it can use the GPUs it already has,” said Jeongkyu Shin, CEO of Lablup. “Following the listing, we will make focused investments in R&D and global sales to become a company that leads the global AI infrastructure software market.”