Chinese scientists have unveiled an artificial intelligence drug discovery platform that compresses the initial compound screening phase of pharmaceutical research from years to seconds, achieving a throughput that is one million times faster than the previous world record for supercomputing molecular docking. The platform, named GalaxyVS, was developed jointly by the National Supercomputing Center in Tianjin and Tsinghua University’s Institute for AI Industry Research. Its existence was first reported by Science and Technology Daily, an official Chinese science publication, on Monday, and subsequently covered by the South China Morning Post.
GalaxyVS is powered by China’s new-generation Tianhe supercomputers, which provide the raw computational throughput needed to handle the platform’s extraordinary workload. The system achieves a daily throughput of 16 trillion molecular dockings, the computational process of simulating how a candidate drug molecule binds to a target protein. A single query across the platform’s library of nearly 100 billion chemical compounds can be completed in less than one minute. To put this in context, traditional computational drug screening methods that rely on standard high-performance computing clusters would require years to process a library of comparable size.
(Related: Beijing Mandates 32 Measures to Accelerate AI Drug Discovery and Surgical Robots)
DrugCLIP and the AI Foundation
At the technical core of GalaxyVS is DrugCLIP, an ultra-fast virtual screening method developed by the Tsinghua research team and formally documented in the journal Science in January 2026. DrugCLIP applies contrastive learning techniques, the same class of AI methods that power image-text models like CLIP, to the problem of molecular screening, enabling the system to rapidly evaluate the potential binding affinity between drug candidates and protein targets without running full physics-based simulations for every compound.
Li Peishun, a researcher at the National Supercomputer Center, was careful to characterize GalaxyVS as more than an incremental improvement on existing tools. “GalaxyVS is not simply an amplification of existing models,” he stated. Rather, he described it as “a complete platform that reconstructs a chemical space of nearly 100 billion elements, integrating AI models, supercomputing, high-performance retrieval and medicinal chemistry.” This framing positions GalaxyVS as a new category of research infrastructure, one that combines the breadth of a massive chemical library with the speed of AI-driven screening and the depth of medicinal chemistry expertise.
Implications for Drug R&D Timelines and Costs
The traditional pharmaceutical research and development process is one of the most expensive and time-consuming endeavors in modern science. Bringing a new drug from initial discovery to market approval typically requires more than a decade and costs several billion dollars. The initial screening phase, identifying which compounds from a vast chemical space are worth investigating further, is a critical early bottleneck. By compressing this phase from years to seconds, GalaxyVS has the potential to fundamentally alter the economics and timelines of drug discovery.
The practical applications span some of the most challenging areas of modern medicine. Li Peishun specifically cited tumors, neurodegenerative conditions such as Alzheimer’s and Parkinson’s disease, rare genetic disorders, emerging infectious diseases, and public health crises as priority application areas. The last category is particularly relevant given the global experience of the COVID-19 pandemic, during which the absence of pre-screened drug candidates significantly delayed the development of effective treatments.
(Related: China NMPA Issues AI Drug Regulation Rules to Overhaul Pharmaceutical Review by 2030)
China’s Supercomputing Advantage in AI Drug Discovery
GalaxyVS represents a convergence of two areas where China has made substantial national investments: supercomputing infrastructure and AI research. The Tianhe series of supercomputers, developed by the National University of Defense Technology, has been a flagship of China’s high-performance computing program for over a decade. By coupling this infrastructure with frontier AI research from Tsinghua University, the GalaxyVS team has created a platform that leverages both national assets to address a problem of direct economic and public health significance.
The announcement also arrives at a moment when AI drug discovery is attracting record investment globally, with major pharmaceutical companies signing multi-billion-dollar partnerships with AI-native drug discovery firms. GalaxyVS positions China as a serious competitor in this space, with a platform that combines national supercomputing resources, frontier AI research, and a chemical library of 100 billion compounds, a combination that few private sector players anywhere in the world can match.
The platform’s development also illustrates a broader pattern in China’s approach to AI research: the deliberate coupling of frontier AI techniques with national-scale computing infrastructure to achieve capabilities that exceed what any single private laboratory could produce independently. This model, in which state-funded supercomputing centers serve as force multipliers for university AI research, has precedents in China’s genomics and materials science programs, and GalaxyVS suggests it is now being applied systematically to pharmaceutical discovery. For global pharmaceutical companies evaluating partnerships or licensing arrangements in China, the emergence of GalaxyVS represents both a competitive challenge and a potential collaboration opportunity.