SK BIOPHARMACEUTICALS announced on the 2nd that it has introduced 128 B200 units based on NVIDIA’s Blackwell architecture. The equipment was secured through “Haein,” South Korea’s first B200-based data center cluster built by SKTelecom. Rather than building the infrastructure itself, SK BIOPHARMACEUTICALS utilizes a GPUaaS (GPU as a Service) model to access the necessary computational resources.
The computational performance of the 128 B200 units introduced this time reaches 576 petaflops (PFLOPS) per second based on FP8, equivalent to approximately 0.58 exaflops. SK BIOPHARMACEUTICALS expects that by securing large-scale computational resources, it will be able to develop AI models optimized for research objectives and data characteristics more quickly and reliably.
SK BIOPHARMACEUTICALS plans to focus on utilizing this AI infrastructure for “TR-AX (Translational Research-AI Transformation),” which signifies the AI transformation of translational research. Translational research is the process of linking the results of preclinical studies—such as animal experiments—to clinical outcomes in humans.
TR-AX is SK BIOPHARMACEUTICALS’ core research capability for maximizing the value of assets in the preclinical and early clinical stages. SK BIOPHARMACEUTICALS plans to apply its proprietary specialized foundation models to actual research projects to further enhance its research capabilities across the board—including translational research, as well as diseases, targets, and compounds.
In addition, the company plans to build a dedicated large language model (LLM) that can be used across various business areas—not just R&D—to boost work productivity and decision-making efficiency, and to establish its own independent AI utilization system. Rather than merely introducing infrastructure, the company will continuously operate sufficient model training and inference resources to accelerate the research cycle—from experimentation and analysis to retraining. Through this, SK Biopharm aims to boost productivity across the entire new drug development value chain, from candidate compound discovery to research decision-making support.
Data security and governance will also be strengthened. The company plans to operate sensitive research data from the bio and healthcare sectors on an infrastructure under its own control to enhance security levels and systematically respond to future regulatory and compliance requirements.
Furthermore, the company intends to expand the scope of this infrastructure beyond its own research to include joint research with external companies, research institutions, and universities, thereby utilizing it as an AI research platform that supports open innovation and collaboration among industry, academia, and research institutes.
In the medium to long term, the company plans to leverage this investment to create new business opportunities. While evaluating the commercialization potential of its AI infrastructure and related technologies, the company intends to actively seek opportunities for joint research and strategic partnerships. In the long term, the plan is to identify various business models—including AI-based research services—to maximize the value of its computing assets and accumulated AI capabilities.
Lee Dong-hoon, President of SK BIOPHARMACEUTICALS, stated, “We decided to make this preemptive investment based on our assessment that a large-scale in-house AI infrastructure is a core foundation that will determine our competitiveness in future new drug development.” He added, “Based on our world-class B200 infrastructure, we will further strengthen TR-AX—a core competency in early-stage new drug R&D—and dramatically increase the speed and efficiency of R&D by maximizing the use of AI throughout the entire new drug development process.”
