Penn Unleashes "Betty" Supercomputer, Escalating AI Research Arms Race

The University of Pennsylvania has launched a powerful new off-campus supercomputer named "Betty," a move that quadruples the university's computing capacity and signals its serious commitment to the escalating "arms race" in academic artificial intelligence research.[1][2] Developed in partnership with NVIDIA, the supercomputer is designed to handle complex AI applications, enabling researchers to analyze massive datasets and generate increasingly refined results.[3][4] Housed in a data center in Collegeville, Pennsylvania, approximately 30 miles from Penn's main campus, Betty represents a significant strategic investment to attract top-tier faculty and students in the competitive landscape of higher education.[4][2]

At the heart of Betty's impressive power is an NVIDIA DGX SuperPOD, a high-performance computing (HPC) and AI cluster built to NVIDIA's reference specifications.[5][3] This system boasts 31 nodes, each containing eight NVIDIA Blackwell B200 GPUs, for a total of 36,704 GPU cores.[5] To facilitate the immense data flow required for large-scale AI models, the system is interconnected with a dedicated NDR400 InfiniBand fabric, a high-speed networking solution.[5][3] This architecture allows a single experiment to scale across the entire SuperPOD, providing a peak performance of 8.5 PetaFLOPS.[5] Complementing the GPU-centric AI system is a primary classical HPC system from Dell EMC, powered by AMD processors and connected via an NDR200 InfiniBand network.[5] The decision to house this powerful hardware off-campus was a practical one, driven by the need for specialized energy and cooling systems that would be difficult to implement in an urban campus environment.[4] The Collegeville facility, operated by Flexential, is specifically designed for high-density computing and is part of the NVIDIA DGX-Ready Data Center program.[6][7]

The introduction of Betty is a direct response to the growing demand for computational resources in modern research.[3] As one university official noted, the scale of AI research has grown to a point where it is no longer feasible for individual schools to maintain their own systems.[3] Betty unifies HPC and AI resources into a centrally managed system accessible to all of Penn's 12 schools and various research centers, democratizing access to cutting-edge technology.[5][6] This shared resource model is intended to maximize the usage of the powerful computers, which might otherwise sit idle in individual labs.[4] The supercomputer is named in honor of Frances "Betty" Holberton, one of the six original programmers of the ENIAC, the world's first general-purpose electronic digital computer, which was developed at Penn in the 1940s.[4][1] This naming choice pays homage to the university's historical role in the dawn of computing.[4]

The implications of this new supercomputer for research at Penn are vast and immediate.[2] Previously, researchers needing this level of computational power would have had to seek time on external systems, which are often fully booked.[1] Now, projects that were once impractical can be conducted entirely in-house.[1] For example, medical researchers are leveraging Betty for a massive genome study, sequencing the genes of 60,000 Americans, a task that generates over 100 gigabytes of data per person in some stages.[1][2] In the realm of robotics, computer scientists are using the system to train a quadruped robot to walk on a yoga ball, with ChatGPT providing performance evaluations.[2] The enhanced capacity also benefits students, allowing undergraduates to create full-stack web deployments with AI-generated code, a task that previously demanded years of professional experience.[2] These examples highlight the transformative potential of Betty across a wide range of disciplines, from biomedical informatics to engineering.[5][1]

The launch of Betty is a key component of the University of Pennsylvania's broader strategic framework, "In Principle and In Practice," and aligns with the mission of the recently established Penn AI Council.[8][9][6] This initiative aims to foster interdisciplinary collaboration and drive responsible AI innovation that translates into real-world solutions.[8] The university's investment in high-performance computing is indicative of a larger trend among top research institutions, which are increasingly recognizing that cutting-edge computational infrastructure is a critical asset for attracting and retaining top talent and securing research grants.[2][10] As universities like Stony Brook and Virginia Tech also roll out new supercomputing clusters, the "arms race in computing" continues to intensify, pushing the boundaries of scientific discovery and solidifying the role of AI in the future of academic research.[2][11][12]