Space telescopes are creating a massive demand for GPU compute

NASA will launch the Nancy Grace Roman space telescope in September 2026. The upcoming space telescope is expected to deliver 20,000 terabytes of data over its life, creating a massive data deluge that will force astronomers to move beyond manual analysis. This volume represents a significant scale increase compared to existing and upcoming observatories. For comparison, the legacy Hubble Space telescope delivers 1 to 2 gigabytes of sensor readings each day, while the active James Webb Space Telescope, which began work in 2021, delivers 57 gigabytes of imagery daily. Meanwhile, the upcoming Vera C. Rubin Observatory in Chile is expected to gather 20 terabytes of data each night.

To address this massive data volume, UC Santa Cruz astrophysicist Brant Robertson is working to adapt astronomical tools for modern computing. Robertson, who has spent 15 years working with Nvidia on GPU applications, co-developed a deep learning model called Morpheus with Ryan Hausen to identify galaxies. Morpheus is switching its architecture from convolutional neural networks to transformers to expand the area it can analyze. Robertson is also working on software tools to improve observations from ground-based telescopes, which are distorted by Earth’s atmosphere. According to Robertson, the field has evolved from analyzing a few objects to performing CPU-based analyses on large datasets, and now to running GPU-accelerated versions of those same analyses. He noted that researchers want to perform AI and machine learning analyses, and graphics processing units (GPUs) are the necessary tool for the job.

However, astronomers face a global shortage of GPUs alongside potential funding instability. Robertson has relied on the National Science Foundation (NSF), a US government agency, to fund a GPU cluster at UC Santa Cruz, but the infrastructure is aging as demand for compute-intensive research rises. This infrastructure faces further risk as the Trump administration proposed cutting the NSF budget by 50%. Robertson emphasized the need for researchers to find creative solutions to secure resources. “You have to be entrepreneurial…especially when you’re working kind of at the edge of where the technology is. Universities are very risk averse because they just have constrained resources, so you have to go out and show them that, ‘look, this is where we’re going as a field’,” said Robertson, a UC Santa Cruz astrophysicist.

Why it matters

Astronomers are facing an unprecedented data deluge from new space and ground observatories, making GPU-accelerated AI models essential for scientific discovery. However, a global GPU shortage and proposed federal funding cuts threaten the computing infrastructure required to process this incoming cosmic data.