NVIDIA Awarded Gordon Bell Special Prize for Million-Fold AI-Accelerated Drug Discovery

At the recently held SC22, the world’s premier supercomputing conference, the Gordon Bell Prize—often regarded as the Nobel Prize of the high-performance computing (HPC) field—was announced. Researchers from NVIDIA, the University of Chicago, and other institutions were awarded the “Gordon Bell Special Prize for HPC-Based COVID-19 Research” for their joint development of an advanced model capable of processing genome-scale data.

How Has NVIDIA, a Pioneer in Accelerated Computing, Achieved Such Remarkable Success in Healthcare? Leveraging its world-leading AI computing platform and full-stack AI solutions honed through deep engagement across multiple industries, NVIDIA has delivered million-fold acceleration for AI applications in various sectors, including healthcare, over the past few years.

Over the past decade, the world has witnessed an AI revolution, with artificial intelligence triggering disruptive transformations across various industries. Empowered by AI capabilities such as machine learning, deep learning, and large language models, fields like drug discovery are experiencing a million-fold leap in efficiency. Underpinning all of this is accelerated computing.

More than half a century ago, Gordon Moore, then an employee at Fairchild Semiconductor, predicted that the number of transistors on silicon-based chips would double approximately every 18 to 24 months, thereby doubling performance while halving costs. In the decades that followed, this prediction became the guiding principle driving the development of the semiconductor industry. Today, as silicon-based semiconductors approach their physical limits in terms of dimensional scaling and the costs of continuing to develop advanced process technologies continue to rise, Moore’s Law, once revered as the gold standard, is nearing its demise.

On the other hand, AI has been developing rapidly, driven by advances in accelerated computing. According to statistics from The Economist, between 2012 and 2018 alone, the computational power used to train large models increased 300,000-fold, doubling approximately every three and a half months.

As the leader in accelerated computing, NVIDIA’s GPUs have served as an indispensable driving force behind this wave of industry transformation. AI application workloads are predominantly characterized by repetitive, compute-intensive tasks, and GPUs excel at parallel processing, enabling geometric increases in AI computational speed. NVIDIA continuously optimizes its products for AI scenarios, making GPU-centric “full-stack accelerated computing” an integral component of AI infrastructure. The convergence of accelerated computing and machine learning has sparked a revolution delivering million-fold speedups across multiple industries worldwide, including scientific computing sectors such as drug discovery, thereby surpassing Moore’s Law to become the new benchmark for the semiconductor industry.

At this year’s GTC conference, NVIDIA founder and CEO Jensen Huang pointed out in his keynote speech that over the past decade, NVIDIA’s accelerated computing has achieved a million-fold speedup in the field of AI, sparking the modern AI revolution. In the next decade, the company aims to achieve another million-fold increase in performance to address major challenges facing humanity, such as drug discovery.

Over the past decade, there has been leapfrog development in building computational power and enhancing the computational performance of applications.

First, accelerated computing and heterogeneous computing have become industry consensus, with NVIDIA building a rich software ecosystem for accelerated computing around its GPU chips. Second, data centers, due to their powerful scalability and ability to support ultra-large-scale computational tasks, have emerged as new computing units. Third, and most transformatively, the widespread adoption of AI is driving change by using neural networks to simulate and replace many scientific computing processes, thereby further simplifying computations and increasing speed.

The convergence of accelerated computing, large-scale data center expansion, and AI is driving rapid advancements in scientific and industrial computing, delivering a million-fold performance leap to tackle some of the most computationally challenging problems, such as climate change, drug discovery, and digital twins.

Currently, NVIDIA has established a full-stack data center-level accelerated computing platform that features multi-dimensional synergy across computing architectures, hardware, algorithms, software, and application frameworks, while encompassing CPUs, GPUs, and DPUs. This full-stack accelerated computing capability has positioned NVIDIA as the global “Expert in Accelerated Computing.”

NVIDIA’s CUDA libraries and NVIDIA SDKs are the core of accelerated computing. With each new SDK release, new scientific domains, applications, and industries can harness NVIDIA’s powerful computational capabilities. These SDKs address extremely complex problems at the intersection of computing, algorithms, and science. The compounding effects generated by NVIDIA’s full-stack approach have enabled million-fold acceleration. NVIDIA SDKs now serve multiple industries, including healthcare, energy, transportation, retail, finance, media, and entertainment, and continue to undergo rapid updates and expansion every year. By implementing acceleration across the full stack and at the data center level, multiple industries will benefit from AI-driven transformation and achieve million-fold leaps in performance.

Benefiting from NVIDIA’s data center-grade full-stack accelerated computing capabilities and advances in artificial intelligence, the field of drug discovery is poised for a significant boost in efficiency.

Development timelines have long been a major pain point in the field of drug R&D. On average, it takes about 10 years for a new drug to go from development to market launch. For pharmaceutical companies, shortening the R&D timeline means an earlier return on investment; for patients with severe or even terminal illnesses, the earlier availability of new drugs translates into a greater chance of survival and renewal.

To accelerate drug discovery, NVIDIA has developed a dedicated AI-accelerated computing software platform called NVIDIA Clara Discovery. This solution integrates GPU-accelerated and optimized frameworks, tools, applications, and pre-trained models, combining capabilities in artificial intelligence, data analytics, simulation, and visualization. It supports interdisciplinary workflows across the drug development process, including cheminformatics research, protein structure prediction, virtual screening of drug candidates, and molecular dynamics simulations. By leveraging accelerated computing, researchers can simulate millions of molecules at once and screen hundreds of potential drug candidates simultaneously, thereby reducing costs and improving efficiency.

Furthermore, NVIDIA’s genomics sequencing analysis acceleration software, NVIDIA Clara Parabricks, can significantly enhance the speed and accuracy of genomic analysis. Meanwhile, its large language model (LLM) framework, NVIDIA BioNeMo, enables the training and deployment of supercomputing-scale large biomolecular language models, helping scientists gain deeper insights into diseases and identify treatments for patients.

Empowered by NVIDIA’s full-stack accelerated computing platform for the healthcare industry, pharmaceutical R&D companies worldwide are bridging the former computational divide and achieving exponential gains in research and development efficiency:

• With the assistance of Clara Discovery, AI drug discovery company Entos leveraged its proprietary OrbNet deep learning architecture to accelerate simulations of chemical reactions between proteins and drug candidates by 1,000-fold, completing in three hours a workload that would have otherwise taken more than three months.

• Parabricks, a startup now part of NVIDIA, leverages NVIDIA DGX AI supercomputers to decompose genetic information into tiny individual fragments for processing when detecting key markers and outliers in genomic sequences. This approach has successfully reduced tasks that previously took days to under half an hour, achieving an efficiency improvement of more than 50- to 80-fold.

• Schrödinger, a global leader in chemical simulation software development, has enhanced the speed and accuracy of its computational drug discovery platform by adopting NVIDIA DGX systems, enabling rapid and precise evaluation of billions of molecules to accelerate the development of new therapies.

• Biotechnology company Recursion has deployed BioHive-1, a supercomputer based on the NVIDIA DGX SuperPOD reference architecture, enabling it to complete deep learning projects within a single day—a task that previously took over a week using its existing cluster.

• Startup Peptone uses the Cambridge-1 supercomputer, an NVIDIA DGX SuperPOD cluster built on NVIDIA DGX systems, BlueField-2 DPUs, and NVIDIA InfiniBand networking, to perform high-throughput inference on millions of proteins in parallel within hours. Based on these computational results, it develops proprietary innovative drugs targeting specific intrinsically disordered proteins (IDPs).

• PrecisionLife, a startup, leverages NVIDIA GPUs to analyze data from 100,000 patients within just a few hours—a feat previously unattainable. This capability enables the identification of subgroups within large patient populations that share matching disease drivers, disease progression patterns, and treatment responses, thereby helping researchers select appropriate drug development targets, determine optimal therapies for individuals, and identify suitable candidates for clinical trials.

• Leveraging NVIDIA’s accelerated computing platform, the AI-driven biopharmaceutical technology company Insilico Medicine completed the early drug discovery process—from target identification and molecular generation and design to in vivo and in vitro efficacy confirmation, safety assessment, and nomination of preclinical candidate compounds—in less than 18 months. This represents a two-thirds reduction in time compared to the approximately four and a half years required by traditional methods, with significantly lower costs.

• “AI + Cryo-EM” Driven Novel Drug R&D Company Shuimu Future Leverages NVIDIA GPU Computing Platform to Boost Efficiency in Sample Screening, Quality Monitoring, and Data Acquisition by Over 10-Fold During Cryo-EM Image Preprocessing, Significantly Reducing Drug Development Costs.

• Leveraging the NVIDIA GPU computing platform, SuiKun Intelligence, a next-generation startup at the intersection of machine learning and biotechnology, has achieved more than a tenfold increase in computational efficiency and model training speed for its AI4D online service platform. This advancement significantly enhances the generation and screening of target-specific molecules, as well as the prediction efficiency of drug-likeness and developability, thereby substantially reducing late-stage drug R&D investments and improving the success rates of clinical trials and market approval.

Leveraging its data center-grade full-stack capabilities, NVIDIA also offers a comprehensive suite of accelerated computing solutions for the healthcare sector. In addition to Clara Discovery, NVIDIA provides specialized solutions tailored to diverse medical application scenarios, including Clara Holoscan for medical devices, Clara Parabricks for genomics, and Clara Guardian for medical imaging and patient care.

From traditional pharmaceutical giants to startups, an increasing number of global healthcare companies are choosing the NVIDIA accelerated computing platform to enhance AI productivity and reduce R&D costs. The vision of Million-X—a million-fold leap in computational performance—has already taken root in healthcare and many other fields critical to humanity’s future well-being. As long as humanity continues to explore technology and discover the unknown, the dream of accelerated computing will never cease.