Skip to content

NVIDIA cuPhoton accelerates black hole research with GPU-powered data pipeline

@NVIDIA · · 40-second read

Share

    This story is from November 16, 2026. For the latest, see today's news.

    November 16, 2026 · AI Tech

    NVIDIA has released cuPhoton, a software platform that moves astronomical data processing onto to speed up black hole research. The tool was designed to handle the massive data streams from networks of radio observatories working as a virtual Earth-sized telescope, such as those that captured the first image of a black hole in the M87 galaxy.

    cuPhoton optimizes the data pipeline by streaming only the information needed, overlapping data loading with signal processing, reconstruction, and analysis. The company reports speedups of up to 3,700x for data loading and 460x for correlation, while also enabling to run analysis in parallel. The system can handle data acquisition rates up to 345 GHz and scales across multiple GPU nodes using NVIDIA's Vera Rubin platform.

    The result: analysis workflows that previously took months now complete in minutes, potentially opening new ways to study phenomena like the light rings around black holes in real time.

    Read the full story at @NVIDIA

    This is an AI-written summary. The original reporting belongs to @NVIDIA. How we write these · Spot a mistake?

    Get the AI brief every morning.

    The five AI stories that matter each morning: startups, money, jobs, tech and what's coming next. Free, and one click to unsubscribe. See our Privacy Policy.