AI Galaxy Hunters Add to Global GPU Shortage

NASA's Nancy Grace Roman space telescope, launching September 2026, will produce 20,000 terabytes of data, intensifying the GPU crunch. Astronomers increasingly rely on GPUs for AI analysis of massive datasets from observatories like James Webb (57 GB daily) and Vera C. Rubin (20 TB nightly). UC Santa Cruz astrophysicist Brant Robertson, who developed the Morpheus deep learning model with Ryan Hausen, is now shifting its architecture from convolutional neural networks to transformers to handle larger areas. He is also training generative AI on space telescope data to correct ground-based observations distorted by Earth's atmosphere. Despite building a GPU cluster at UC Santa Cruz via NSF funding, Robertson faces pressure from global demand and proposed 50% NSF budget cuts. He urges universities to be entrepreneurial in adopting GPU-accelerated AI for astronomy.