SK Biopharm arms itself with 128 Nvidia B200 chips, extending SK-Nvidia AI alliance into drug development

SK Biopharm has secured 128 units of Nvidia's next-generation Blackwell architecture-based B200 AI accelerators, building out a large-scale proprietary AI computing infrastructure.

The move marks the first tangible result of the SK Group-Nvidia partnership — forged primarily around semiconductors and telecommunications — expanding into drug development.

The acquisition directly channels SK Group Chairman Chey Tae-won's group-wide AI transformation drive into pharmaceutical and biotech R&D, while also accelerating SK Biopharm President Lee Dong-hoon's strategy to grow the company into a "big biotech."

The 128 B200 units deliver 576 petaflops per second at 8-bit low-precision (FP8) computation — roughly 0.58 exaflops — making it by far the largest self-built AI infrastructure assembled by a domestic pharmaceutical or biotech company for pure research purposes.

SK Biopharm plans to deploy that computing power entirely toward TR-AX, its AI-driven translational research initiative and the linchpin of its drug development strategy. Translational research bridges preclinical findings from cell and animal studies to human clinical trials, narrowing the gap between the two stages to reduce the failure rate of drug candidates and speed up the identification of new indications.

The investment aligns with a vision Lee laid out at a press conference last month for the in-licensing of cenobamate. At the time, he said AI was "not an option but a prerequisite" for becoming a big biotech, adding that the company's "ultimate and most critical goal is to focus above all on building TR-AX — a system in which the entire translational research process is run entirely by AI agents."

Building on the B200 infrastructure, SK Biopharm plans to develop its own biotech- and healthcare-specific foundation model and accelerate the full R&D cycle — from target identification and candidate screening to clinical success-rate simulation.

The infrastructure's applications extend well beyond the research lab. SK Biopharm plans to use the computing resources to develop a proprietary large language model for company-wide operations, aiming to maximize productivity and decision-making efficiency across the organization.

Particularly sensitive data — including patient information and drug candidate details — will be processed within a closed, self-controlled infrastructure, allowing the company to proactively meet the strict data governance and security compliance requirements of global regulators.

SK Biopharm also intends to open the infrastructure as an "open innovation AI platform" for joint research with promising external biotech firms, academic institutions and global research organizations. The company envisions using its computing capacity to attract global partnerships and, over the longer term, to develop new business models such as AI-based drug discovery research services.

Having diversified its epilepsy commercialization pipeline through the 1.1 trillion won ($812 million) in-licensing of cenobamate, SK Biopharm has now added cutting-edge AI infrastructure to complete what it calls the two wings of its big biotech ambition.

Some analysts say the early acquisition of large-scale AI infrastructure reflects the synergy between the holding company's aggressive AI transformation drive and SK Biopharm's solid financial fundamentals.

With Chey declaring "all-round AI ecosystem expansion" as a core pillar of the group's rebalancing strategy, SK Biopharm — which has demonstrated strong cash generation through the direct US sales success of its epilepsy drug cenobamate (sold in the United States as Xcopri) — was identified as the group affiliate best positioned to lead the adoption of its AI capabilities.

Earlier this year, Lee was the only CEO from a Korean pharmaceutical or biotech company to receive an official invitation to a private reception hosted by Nvidia CEO Jensen Huang at the JP Morgan Healthcare Conference in San Francisco in January — the same event at which Eli Lilly and Nvidia announced the world's first joint AI innovation laboratory, giving Lee a firsthand look at the convergence of AI and drug development. When Huang visited South Korea in June, Chey's eldest daughter, Chey Yoon-jung, head of SK Biopharm's strategy division, joined the dinner meeting at a Gangnam restaurant, drawing considerable attention.

That network has now translated into tangible results through a cloud GPU-as-a-service subscription to "Haein," SK Telecom's data center cluster and the first in South Korea built on B200 chips. The pharmaceutical affiliate's ability to move quickly and secure the latest AI computing resources — in short supply worldwide — reflects tight group-level coordination and execution on AI infrastructure.

"We made a preemptive investment decision based on the judgment that large-scale proprietary AI infrastructure is the critical foundation that will determine future drug development competitiveness," Lee said. "We will further strengthen TR-AX — the core capability in early-stage drug R&D — on the back of our global-standard B200 infrastructure, and dramatically improve the speed and efficiency of research and development by maximizing AI use across the entire drug development process."