Nvidia and the Georgia Institute of Technology have announced their first-ever AI supercomputer specially developed for student use.
Nvidia’s graphic processing units have been at the center of large language models developed by Meta, OpenAI, Alphabet, and several tech startups. Enabling a massive growth in revenue for the chipmaker over the past year, even surpassing some of its largest clients, such as Meta and Alphabet, in market cap.
According to the chip-making giants, the purpose of this move is to make supercomputing resources available to anyone. A move to enable the training of the next-generation workforce on AI. Before now, supercomputers were associated with large tech companies and highly funded tech startups.
Currently, only Georgia Tech undergraduate students will have access to the computing cluster. Nvidia’s enterprise AI software and “virtual gateways” created by Penguin Solutions will power the supercomputer.
However, students have only used the supercomputer in just one class, Foundations of Machine Learning, since the start of the semester. By the end of 2025 first half, Georgia Tech projects to grant access to all its undergraduate and graduate students.
The Georgia Tech supercomputer runs on 20 Nvidia HGX H100 systems, housing 160 Nvidia H100 GPUs, which have a high demand in the tech industry today.
For reference, a single GPU can develop a multiplication function in seconds, which would take 50,000 students and 22 years to accomplish.
According to Arijit Raychowdhurry, chair of Georgia Tech School of Electrical and Computing Engineering, the students will use the supercomputer for course projects related to large language models, supply chain management, chemical or biomedical engineering, computer vision, robotics, generative AI creativity, and design pursuits.
