AI galaxy hunters analyzing cosmic data intensify the global GPU crunch

AI Galaxy Hunters Face Computing Challenges

The study of galaxies has entered a new era of data abundance that's creating unexpected pressure on GPU availability. NASA announced the Nancy Grace Roman space telescope will launch in September 2026, eight months ahead of schedule, and is expected to deliver 20,000 terabytes of data over its lifetime 1

1

. This astronomical influx adds to the James Webb Space Telescope's daily downlink of 57 gigabytes of imagery since 2021, and the Vera C. Rubin Observatory's anticipated 20 terabytes of data each night starting later this year 1

1

. By comparison, the Hubble Space Telescope delivers just 1 to 2 gigabytes daily, highlighting how dramatically the volume of cosmic data has expanded 1

1

.

Deep Learning Models Transform Astrophysics Research

Brant Robertson, a UC Santa Cruz astrophysicist, has witnessed this transformation firsthand while working with data from these missions. Over 15 years collaborating with Nvidia, Robertson has applied GPU technology first to advanced simulations testing theories about supernova explosions, and now to developing tools for analyzing the torrent of space telescope data . Robertson and then-graduate student Ryan Hausen developed Morpheus, a deep learning model that can examine large data sets and identify galaxies automatically . Their early AI analysis of Webb data identified a surprising number of specific disc galaxies, adding new insights to theories about the formation of early galaxies and the development of the universe .

Computational Models Evolve With AI Advances

Morpheus itself is adapting to modern AI techniques. Robertson is switching its architecture from convolutional neural networks to transformer architecture, the same technology behind large language models . This upgrade will enable the model to analyze several times more area than currently possible, significantly accelerating its work. Robertson is also developing generative AI models trained on space telescope data to enhance observations from ground-based telescopes, which suffer from atmospheric distortion . Despite advances in rocketry making it difficult to launch 8-meter mirrors into orbit, using software to improve Rubin's observations offers a practical alternative.

Global GPU Crunch Hits Scientific Research

The challenge astronomers now face extends beyond software development. Robertson is experiencing the pressure of global demand for GPU access firsthand . While he used National Science Foundation funding to build a GPU cluster at UC Santa Cruz, it's becoming outdated even as more researchers seek to apply compute-intensive techniques to their work. The situation is compounded by the Trump administration's proposal to cut the NSF's budget by 50% in its current budget request . "People want to do these AI, ML analyses, and GPUs are really the way to do that," Robertson said, noting that researchers must be entrepreneurial when working at the edge of technology, especially as universities remain risk-averse due to constrained resources .

Unprecedented Cosmic Data Requires AI Solutions

When the first images from the James Webb Space Telescope began returning in 2022, Robertson and his colleagues encountered an overwhelming sight. "There were galaxies everywhere," Robertson recalled, describing how the abundance of distant galaxies genuinely shocked the team 2

2

. Robertson leads a team at UC Santa Cruz studying how the earliest galaxies formed after the Big Bang, work that has broken the record for the most distant known galaxy multiple times, each time pushing observation closer to the universe's first light 2

2

. JWST, the most powerful observatory ever launched, observes in infrared and captures light that has traveled for more than 13 billion years, with each deep-field image crowded with hundreds of thousands of galaxies 2

2

. "These datasets are far too large and complex for humans to analyze by hand," Robertson explained, noting that even teams of experts would take years to accomplish what now needs to happen in days 2

2

. Without computation at this scale, the cosmic data would simply pile up, making AI and GPU technology essential for modern astrophysics and understanding the universe.