AI galaxy hunters are adding to the global GPU crunch

Source: TechCrunch

NASA announced that it will launch the Nancy Grace Roman Space Telescope into orbit in September 2026, eight months ahead of schedule. The new telescope is expected to deliver 20 000 TB of data to astronomers over its lifetime.

This will add to the 57 GB of breathtaking imagery downlinked daily from the James Webb Space Telescope, which began operations in 2021, and to the upcoming survey by the Vera C. Rubin Observatory in the Chilean Andes, slated to begin later this year and projected to gather 20 TB of data each night.

For comparison, the Hubble Space Telescope, once the gold standard, delivers just 1–2 GB of sensor readings per day. As the volume of data has exploded, astronomers are increasingly turning to GPUs to process and analyze it.

GPU‑Accelerated Astronomy

Brant Robertson, a UC Santa Cruz astrophysicist, has witnessed this shift firsthand. Over the past 15 years he has collaborated with NVIDIA to apply GPUs to space‑related problems, from advanced simulations of supernova explosions to tools for analyzing the torrent of data from new observatories.

“There’s been this evolution—from looking at a few objects, to doing CPU‑based analyses on large scales of the data set, to then doing GPU‑accelerated versions of those same analyses,” Robertson told TechCrunch.

The Morpheus AI Model

Robertson and former graduate student Ryan Hausen developed a deep‑learning model called Morpheus that can scan large data sets and identify galaxies. Their early AI analysis of Webb data uncovered a surprising number of a specific type of disc galaxy, adding a new wrinkle to theories about the universe’s development.

Morpheus is now being updated:

• Architecture shift: Moving from convolutional neural networks to transformer models (the same architecture behind large language models).
• Impact: The new design will enable analysis of several times more sky area, dramatically speeding up processing.

Generative AI for Ground‑Based Observations

Robertson is also exploring generative AI models trained on space‑telescope data to improve the quality of observations collected by ground‑based telescopes, which suffer from atmospheric distortion. Since launching an 8‑meter mirror into orbit remains challenging, software‑based enhancements to Rubin Observatory data represent a promising alternative.

The Global GPU Crunch

Despite these advances, Robertson feels the pressure of growing demand for GPU access:

• He built a GPU cluster at UC Santa Cruz with NSF funding, but the hardware is becoming outdated as more researchers adopt compute‑intensive techniques.
• The Trump administration’s budget proposal sought to cut NSF funding by 50 %, threatening further support for such infrastructure.

“People want to do these AI, ML analyses, and GPUs are really the way to do that,” Robertson said. “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.’”

---