The University of Pennsylvania has unveiled “Betty,” a new off-campus supercomputer that quadruples the university’s computing capacity and is designed to run AI models analyzing videos, images, texts, and databanks. Located 30 miles from campus in a Collegeville data center, Betty positions Penn to compete in what officials describe as an “arms race in computing” among top research universities seeking to attract faculty and students with cutting-edge AI capabilities.
What you should know: Betty represents a significant leap in Penn’s computational infrastructure, built in record time to meet surging demand for AI research capabilities.
- The supercomputer is an Nvidia “SuperPOD” featuring central processing units (CPUs), graphics processing units (GPUs), and data storage systems designed to improve with each query.
- Named after pioneering Penn computer scientist Frances “Betty” Holberton (1917-2001), the system consumes enough electricity to power 1,000 homes and hums louder than a commercial jet.
- Penn expects Betty to land on the Top500.org list of the world’s most powerful supercomputers.
Why this matters: Universities are locked in an intensifying competition to provide faculty with the computational resources needed to secure grants and attract top talent.
- “Top universities’ faculty are competing for grants and for students. If they don’t have the resources to compete, they won’t be in good shape,” said Jaime Combariza, a computer scientist Penn recruited from Johns Hopkins University to run the Penn Advanced Research Computing Center.
- Associate Vice Provost Michael Borda confirmed the competitive pressure: “No question, there’s an arms race in computing. We were not offering enough capacity.”
Key infrastructure challenges: Betty’s off-campus location reflects the practical realities of powering massive AI systems in urban environments.
- The facility sits in Montgomery County because Philadelphia lacks sufficient electrical capacity for such operations.
- “The hard thing in Philly is power,” said Kenneth Chaney, associate director for AI and Technology for PARCC, noting that building a city-based data center would require Penn to construct its own power plant.
- The Collegeville site is fed by two separate high-voltage substations and includes eight backup generators with several days of fuel storage.
How it works: Betty enables Penn researchers to conduct AI projects that were previously impossible or required outsourcing to external providers.
- Penn computer scientists are using the system to teach a quadruped robot to walk on a yoga ball, with ChatGPT evaluating the robot’s performance.
- Medical researchers led by Marylyn D. Ritchie are conducting genome sequencing studies of 60,000 Americans, requiring more than 100 gigabytes per person at some stages.
- The system allows undergraduate students to create full-stack web deployments using AI-generated code, capabilities that previously required years of professional experience.
The bigger picture: Betty’s deployment reflects broader trends in AI infrastructure demand that are reshaping power consumption across Pennsylvania and beyond.
- The facility offers a view of the Limerick nuclear power plant, whose owner Constellation Energy wants to increase capacity by 300 megawatts to support growing data center demand.
- Pennsylvania is planning larger AI-serving data centers in the Susquehanna Valley and at the former U.S. Steel Fairless Works in Bucks County.
- Data centers nationwide are operating at “historically low vacancy rates” due to surging demand for AI computing power.
What they’re saying: Research leaders emphasize how centralized computing resources are becoming essential for academic competitiveness.
- “Research universities, with hundreds of professors, acting like small businesses with their peculiar needs, are accelerating as fast adapters of AI technologies,” said Patrick Doherty, chief revenue officer at Flexential, which operates the data center housing Betty.
- Chaney highlighted the system’s forward-looking potential: “We always want to be looking forward. Everyone is now employing these GPUs. Soon we want to be able to deploy quantum computing. Neumorphic processes. Maybe in five years.”
Off-Campus Supercomputer to Help UPenn in AI Arms Race