The vast expanse of space has long fascinated humans, and the latest generation of space telescopes is about to deliver a data deluge of unprecedented proportions. With the Nancy Grace Roman space telescope launching in September 2026, eight months ahead of schedule, astronomers are eagerly anticipating the torrent of information it will bring. The new telescope is expected to deliver 20,000 terabytes of data to scientists, dwarfing the 57 gigabytes of daily data from the James Webb Space Telescope and the 20 terabytes of nightly data from the Vera C. Rubin Observatory. This explosion of data is set to transform the field of astronomy, and GPUs are emerging as the key to unlocking its secrets.
GPU Power: The Key to Unlocking Space Data
For a long time, astronomers relied on the Hubble Space Telescope to provide a wealth of information about the cosmos. However, its 1 to 2 gigabytes of daily sensor readings were pored over by hand, a labor-intensive process that could only be sustained for so long. The advent of larger space telescopes like the James Webb Space Telescope and the Vera C. Rubin Observatory, which collects 20 terabytes of data each night, has made it clear that manual analysis is no longer feasible. That’s where GPUs come in – powerful processors that can handle complex calculations at incredible speeds.
Brant Robertson: A Pioneer in GPU-Powered Space Research
UC Santa Cruz astrophysicist Brant Robertson has spent the past 15 years working with Nvidia to apply GPUs to the problems of understanding space. His work has taken him from advanced simulations testing theories about supernova explosions to developing the tools to analyze large data sets from the newest 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,” he explained in an interview with TechCrunch. Robertson’s work has been instrumental in demonstrating the power of GPUs in space research, and his team’s deep learning model, Morpheus, has identified a surprising number of disc galaxies in Webb data.
Morpheus: A Deep Learning Model for Galaxy Analysis
Developed by Brant Robertson and his team, Morpheus is a deep learning model designed to analyze large data sets and identify galaxies. Its early AI analysis of Webb data revealed a significant number of disc galaxies, adding a new wrinkle to theories about the development of our universe. What’s more, Morpheus is not just a static model – it’s being upgraded to take advantage of the latest advancements in AI research. Robertson is switching its architecture from convolutional neural networks to the transformers behind the rise of large language models, which will result in the model being able to analyze several times the area than it can currently, speeding up its work.
Generative AI Models: The Future of Ground Telescope Observations
Robertson is also working on generative AI models trained on space telescope data to improve the quality of observations collected by ground telescopes. These models use machine learning algorithms to generate high-quality images from distorted data, effectively “correcting” the images to show more detail. This is particularly important for ground-based telescopes, which have to contend with the distorting effects of Earth’s atmosphere. By using AI to improve the quality of observations, researchers can get a clearer picture of the universe, and make new discoveries that would be impossible with traditional methods.
The GPU Crunch: A Global Phenomenon
The demand for GPU access is not limited to the United States – researchers around the world are clamoring for access to these powerful processors. The Trump administration’s proposal to cut the NSF’s budget by 50% has only exacerbated the problem, leaving many researchers scrambling to find alternative funding sources. Brant Robertson’s experience with the National Science Foundation has been typical – he’s used the organization to build a GPU cluster at UC Santa Cruz, but it’s becoming outdated even as more researchers want to apply compute-intensive techniques to their work. “You have to be entrepreneurial…especially when you’re working kind of at the edge of where the technology is,” he said. “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.'”
Adding Global GPU Capacity: A Solution to the Crunch
So, what can be done to address the GPU crunch? One solution is to increase global GPU capacity – building more powerful processors and making them available to researchers. This can be achieved through a combination of public and private funding, as well as partnerships between governments, industry, and academia. By working together, it’s possible to create a global network of GPU-powered research facilities, providing access to these powerful processors for researchers around the world.
Practical Steps to Add Global GPU Capacity
So, how can we increase global GPU capacity? Here are some practical steps that can be taken:
- Establish partnerships between governments, industry, and academia to provide funding for GPU research.
- Develop and implement new GPU architectures that are more efficient and powerful.
- Build and maintain a global network of GPU-powered research facilities.
- Provide training and education for researchers on how to use GPUs effectively.
- Develop new software and tools to make it easier for researchers to access and use GPUs.
Conclusion
The advent of large space telescopes like the James Webb Space Telescope and the Vera C. Rubin Observatory has created a data deluge that is set to transform the field of astronomy. GPUs are emerging as the key to unlocking this data, and researchers around the world are clamoring for access to these powerful processors. By increasing global GPU capacity, we can address the crunch and make it possible for researchers to analyze large data sets and make new discoveries that would be impossible with traditional methods. The future of space research looks bright – and it’s powered by GPUs.
