Lablup expands GPU management and prepares agent infrastructure tools
The AMW Read
The interview adds customer and product details for an infrastructure provider, but its reported results and planned expansion do not establish a segment-wide shift.
Lablup expands GPU management and prepares agent infrastructure tools
Lablup is expanding its AI infrastructure business around software that manages GPUs and servers from multiple manufacturers. CEO Shin Jung-kyu told Edaily that the company operates its platform at more than 120 sites and has managed more than 2,000 GPUs at one site. Lablup also plans to begin beta testing NeuVM, a CPU virtualization tool for managing AI agent resources and security, by the end of 2026.
The company occupies the compute execution layer, where enterprises must allocate hardware across AI development and inference workloads. Supporting both NVIDIA and AMD equipment gives Lablup a position in mixed hardware environments, though the interview does not establish how its performance compares with rivals. Lablup says one customer raised GPU utilization from 30% to 85% and cut related spending by 40%; those figures describe one company-reported case, not a general result. The expansion into agent operations also builds on its Continuum products for managing AI inference systems.
For enterprise buyers, the practical test is whether Lablup can maintain hardware-specific optimization while fitting into existing systems. The company says it has spent the past 18 months making its products more modular. For investors, overseas distribution is a near-term milestone: Lablup expects partners Insight and AMAX to start selling early in 2027, while it trains partners to handle sales, delivery, and technical support. Those expected sales remain distinct from demonstrated overseas revenue growth.
