[Zoom-In on SMEs] Xiilab’s Blueprint for a Transition to VLA-Centered AI… Growing Confidence in ‘AstraGo’
Xiilab, a vision artificial intelligence (AI) company, has begun full-scale expansion into the “physical AI” business, positioning visual-language-action (VLA) technology as a new growth engine. The company plans to extend its capabilities into new domains based on the digital twin model technology it has developed. This shift is supported by the stable performance of its AI infrastructure business—its primary cash cow—and the growing momentum of its digital twin business.
According to the Financial Supervisory Service’s electronic disclosure system on Thursday, Xiilab’s consolidated sales rose 47.6 percent year-on-year to ₩3.8 billion in the third quarter. During the same period, the company reduced its operating loss to ₩3 billion. The company attributes the improvement to stable sales of AstraGo and growth in its digital twin business. Xiilab is also expanding its scope by increasing the supply of its own vision-language model (VLM) AI.
AstraGo, launched in May of last year, is a GPU cluster management solution developed by Xiilab. Backed by technology that improves GPU efficiency fourfold and reduces AI environment setup time by 95 percent, AstraGo has secured a partnership with global chipmaker Nvidia. Xiilab also introduced Nvidia’s latest GPU server, the DGX B300, in Korea for the first time, fueling growth through increased supply to large corporate data centers.
AstraGo has achieved 220 license sales to date, significantly increasing its revenue base. Compared to 50 licenses sold last year, the company recorded a compound annual growth rate (CAGR) of 322 percent and a 95 percent license renewal rate. Its customer base has expanded from large enterprises to affiliate data centers. This growth is driven by the establishment of a recurring revenue model based on subscription licensing.
With confidence in its stable revenue performance, Xiilab is accelerating efforts to secure new growth engines. At an online briefing held last month, Xiilab announced that physical AI would become a mid- to long-term strategic focus. The roadmap consists of three key pillars: AI factories, AI semiconductors, and AI robots. By integrating technologies across these areas, the company aims to complete a physical AI system applicable to real industrial environments.
Xiilab is developing a VLA model that analyzes language, vision, and behavior by applying robotic simulation to its VLM. The company plans to apply this technology to various manufacturing platforms, including AI factories. “Based on our digital twin business, we will strengthen our ability to simulate physical devices such as robots,” a company official said. “We plan to integrate this data into our VLM so it can control VLA models.”
In addition, the Vision AI division is promoting the development of specialized AI models tailored for specific industries. Xiilab plans to expand its line of device-based lightweight AI products in response to rising demand for image analysis solutions in sectors such as media, construction, and transportation. Vision AI will evolve as an input layer for physical AI implementation and be developed into a form that can be directly applied in real-world environments.
Xiilab previously secured ₩19 billion in funding for its new business initiatives through a capital increase carried out in August. The rights offering, conducted via shareholder allocation, was successful with a subscription rate of 257.95 percent during the public offering process involving existing shareholders. In securing new capital, the company also strengthened its financial structure. Consolidated equity totaled ₩9.6 billion at the end of last year but rose 167.3 percent to ₩25.7 billion by the end of the third quarter.
댓글
댓글 쓰기