The SETI Institute, renowned for its decades-long search for intelligent life beyond Earth, has announced it will adopt NVIDIA’s new IGX Thor platform to enhance real-time signal processing at the Allen Telescope Array (ATA) in Northern California.

The collaboration represents a major step toward bringing cutting-edge AI capabilities to radio astronomy at an unprecedented scale.

According to Dr. Andrew Siemion, the Bernard M. Oliver Chair for SETI at the Institute, precision is key to their search.

“We search for any kind of electromagnetic emission that is inconsistent with the known or expected astrophysical and instrumental backgrounds,” Siemion told The Debrief in an email. “This sort of agnostic search is incredibly computationally intensive and algorithmically quite challenging with traditional approaches.”

“This new IGX Thor system will allow us to readily couple state-of-the-art AI processing with our raw astronomical data streams,” he added, “making our searches more sensitive and more complete.”

The ATA consists of 42 antennas continuously scanning the sky for radio signals that could reveal cosmic phenomena—or potentially, signs of intelligent life. Historically, processing such enormous volumes of data was time-consuming and labor-intensive. With NVIDIA’s IGX Thor platform, SETI researchers can now process and interpret signals directly at the telescope, dramatically reducing the time required to detect unusual patterns.

“NVIDIA IGX Thor enables us to run AI inference and GPU-accelerated signal processing workloads closer to the edge,” said Luigi Cruz, Staff Engineer at the SETI Institute, in a press release. “Its compact form factor and power efficiency make it an ideal development platform for our next-generation pipeline, which is based on NVIDIA Holoscan.”

“Our group has demonstrated that AI-based approaches to analyzing radio data can be incredibly effective,” Siemion added. “These algorithms aim to identify electromagnetic emissions with a high degree of spectral or temporal coherence that match our expectations for distant cosmic technosignatures, while excluding signals consistent with instrumental interference.”

He noted that new algorithms in development are designed to detect an even broader class of anomalies “free of any preconceived notion of what a technosignature might look like.”