From Materials Simulation to Experimental Astronomy, New NVIDIA AI Software Unlocks Scientific Discoveries
NVIDIA Blog · Chris Porter · 2026-06-22
NVIDIA announced at ISC 2026 three new AI software tools — cuPhoton for astronomical data, DAQIRI for real-time detector streaming, and ALCHEMI NIM microservices for materials simulation — delivering speedups ranging from 3x to 14,900x across scientific computing workloads.
Appears in
Extraction
Topics: ai-for-sciencegpu-accelerationmaterials-discoveryastrophysicshpc
Claims
- NVIDIA's cuPhoton accelerated loading and reading of astronomical FITS images from the Rubin Observatory's LSST by 14,900x on GB200 NVL72 systems, with up to 8,400x faster signal processing using 32 Grace Blackwell superchips.
- CERN's A-GHOST project uses NVIDIA DAQIRI to run real-time AI on ATLAS collision data that would otherwise be rejected due to storage constraints, covering over 99% of previously discarded data.
- ALCHEMI's VASP microservice achieves a 3x speedup for geometry optimization by running multiple calculations on a single GPU via NVIDIA Multi-Process Service.
- Lila Sciences achieved 50x acceleration in high-throughput materials screening and a 6x speedup in training/inference with 3x memory reduction using ALCHEMI's specialized TensorNet kernels.
Key quotes
The work showcases using a powerful computing stack assembled to accelerate discovery at a scale no individual scientist could achieve alone.