From Materials Simulation to Experimental Astronomy, New NVIDIA AI Software Unlocks Scientific Discoveries
15:00 · June 22, 2026 · NVIDIA
At the ISC conference running in Hamburg this week, NVIDIA is introducing new software that speeds AI for science, from chemistry and materials discovery to the search for dark matter. The NVIDIA DAQIRI library and new NVIDIA ALCHEMI NIM microservices — as well as the NVIDIA cuPhoton reference code, coming soon — turn work that […]
Summary
NVIDIA has released three new software components aimed at scientific workloads: the DAQIRI networking library, the ALCHEMI suite of NIM microservices, and the cuPhoton reference code. Announced at the ISC conference in Hamburg, these tools form part of the CUDA-X collection and convert long-running CPU-bound tasks in data acquisition, molecular simulation and astronomical image processing into GPU-accelerated pipelines that operate at instrument or survey rates.
cuPhoton targets the handling of large multidimensional datasets stored in the FITS format. On GB200 NVL72 systems it delivered a 14,900-fold acceleration in loading and reading images from the Rubin Observatory’s Legacy Survey of Space and Time, together with up to 8,400-fold faster signal processing when run across 32 Grace Blackwell superchips. Princeton and Harvard researchers are adopting the code to analyse petabyte-scale observations from dark-energy surveys and other telescope campaigns.
DAQIRI streams detector output directly into GPU memory, removing the fixed-hardware bottlenecks that previously caused data loss when sensor rates exceeded storage capacity. The A-GHOST collaboration, involving CERN, the University of Chicago and University College London, uses the library to run real-time AI inference on collision events recorded by the ATLAS experiment—events that would otherwise be discarded because they exceed the experiment’s conventional storage budget.
ALCHEMI supplies domain-specific microservices for chemistry and materials science. The batched-geometry-relaxation and batched-molecular-dynamics services allow simultaneous simulation of millions of candidate structures, while an upcoming VASP microservice exploits the Multi-Process Service to achieve a 3× throughput gain on geometry-optimisation workloads. Lila Sciences has reported a 50-fold increase in high-throughput materials screening and a 30 % reduction in magnetic-property calculations when combining these services with the ALCHEMI Toolkit for training machine-learning interatomic potentials.
Why it matters
This article highlights major advancements in AI infrastructure for scientific research by NVIDIA, a key player in the global and European AI ecosystem. While not specific to the Netherlands, these tools will significantly impact how European research institutions and tech companies leverage AI for materials discovery and data analysis.
