Science at the speed of light. 🧪 🌌 Unveiled at #ISC26, new CUDA-X software (cuPhoton, DAQIRI, and ALCHEMI) that turns days of research into real-time insights. From a 15,000x boost in telescope data analysis to accelerated materials discovery, we’re helping scientists unlock breakthroughs faster than ever. Read the full blog: nvda.ws/4w7DFTO

Turn local weather data into real-time operational decisions With AI data assimilation tools in NVIDIA Earth-2, you can pipe proprietary and local sensor data directly into high-res forecasting models like CorrDiff, StormCast, and HealDA. No full model retraining required. Read the full Earth2Studio code walkthrough: nvda.ws/4cE3NOQ

Introducing CUDA Rust! CUDA Rust lets you write GPU kernels natively in Rust, not just launch them from it. Two paths: cuda-oxide for SIMT kernels compiled to PTX, and cutile-rs for Tile-based programming on stable Rust. Both can catch aliasing errors at compile time. Technical blog: nvda.ws/4hm1bHS

CUDA Toolkit 13.4 is now available. It adds Windows on Arm support, early developer support for NVIDIA Rubin, and MPS V3 controls for partitioning GPU compute and memory across workloads, plus new CUDA Python APIs and CCCL performance improvements. Here's all the technical details: nvda.ws/4A8VzbC

NVIDIA × Fair Math Challenges are LIVE on FHERMA! We're excited to collaborate with @NVIDIAHPCDev to bring GPU acceleration layer support to the FHE ecosystem using NVIDIA's cuPQC. To kickstart this initiative, we're launching two new developer challenges with GPU compute: - Polynomial Multiplication: Build high-performance polynomial multiplication on cuPQC's bigint backend.  (Great entry point for GPU/CUDA devs!) - Key Switching in CKKS: Optimize one of FHE's biggest bottlenecks using GPU-native arithmetic. Whether you're an FHE researcher or a CUDA/GPU dev, you can help build open-source infrastructure that projects worldwide will use. Join the challenges: fherma.io #FHE #Cryptography #CUDA #GPU #cuPQC #NVIDIA #OpenSource

SP1 is integrating the NVIDIA cuPQC cryptography library. cuPQC powers core operations in SP1's GPU prover, making proving performant across @NVIDIA accelerated computing and expanding the range of GPUs that can prove efficiently on the Succinct Prover Network.

NVIDIA cuPQC v0.6 is here! Introducing cuPQC-BigInt and cuPQC-NTT, GPU-accelerated math libraries for building cryptographic systems. Dive into big integer arithmetic (modular multiplication, inversion, exponentiation, and reduction) and number theoretic transforms for zero-knowledge proofs, fully homomorphic encryption, RSA, elliptic curve, and isogeny-based cryptography. Read the blog: nvidia.github.io/cuPQC/relea… Get started: developer.nvidia.com/cupqc

10 million downloads for NVIDIA Warp 🎉 Warp started with a simple idea: you shouldn’t have to leave Python to get real GPU performance for physics and simulation. Since then, developers have used it to accelerate work across physics simulation, computational engineering, geometry processing and robotics. Thank you to everyone who downloaded it, broke it, filed an issue or sent a PR. On to the next 10M!

CuPy v14.2 is out! 🎉 Highlights: 🔹 SciPy-compatible sparse array classes 🔹 Windows on Arm (RTX Spark) support 🔹 Initial integration with nvmath-python Check the release notes for details 👇

NVIDIA cuPhoton is now available on GitHub 🔭 Working with data from telescopes, X-ray sources and laser experiments often means spending months or years just getting the data ready to use. cuPhoton moves that pipeline onto the GPU to speed up that process, from ingest and cleaning to analysis, bringing it down to hours or minutes. Get started: github.com/nvidia/cuPhoton

RT @NVIDIARobotics: A milestone for the Holoscan community. 🎉 Over a million developers and researchers are using the Holoscan SDK to adva…

NVIDIA nvmath-python v1.0 is now GA. Get the raw performance of CUDA-X math libraries (cuFFT, cuBLAS, cuSPARSE) directly inside your NumPy, CuPy, and PyTorch-based code. 🔹 Scale CPU ➡️ Single-GPU ➡️ Multi-node 🔹 Stateful APIs for reusable planning & autotuning 🔹 Fast JIT kernel fusion with numba-cuda Details: developer.nvidia.com/blog/ru…

The CUDA MCP Server is available now. This NVIDIA hosted MCP server gives your agent a search tool over indexed, current CUDA documentation and code examples, curated by NVIDIA engineers. Ask a CUDA question, the agent searches the corpus and answers in-line, without leaving your agent of choice. Compatible with Claude Code, Codex, Cursor, Antigravity, or any MCP-compatible agent. One line to connect 👇developer.nvidia.com/nsight-…

📰 @nvidia's Earth2Studio now ships with native Earthmover Marketplace data sources. @NOAA's ERA5, @ECMWF IFS, or any dataset on the Marketplace is available to FourCastNet, CorrDiff, StormCast: the AI weather models surpassing the accuracy of physics-based predecessors 🙌

NVIDIA cuEST 0.2.0 is live. Our cuEST (CUDA-X Electronic Structure Theory) library now delivers faster, GPU-accelerated quantum chemistry calculations, along with many new features. Explore the latest capabilities and start accelerating atomic-scale simulations with cuEST 0.2.0, available today. 🔗 nvda.ws/44APhD4

Rare events are difficult to study precisely because conventional sampling rarely encounters them. NVIDIA researchers combine guided diffusion with probability correction to better estimate extreme-event likelihoods. Beyond climate, the approach could advance research in aerospace, robotics, finance, materials, molecular simulation, and infrastructure resilience. Explore the research: nvda.ws/4aNJL3u

Excited and honoured to have @nvidia as part of RAISE Hackathon 2026. Let the games begin. NVIDIA didn't set out to build the backbone of AI, CUDA launched in 2006 as a way to make gaming GPUs programmable for general computing. Nobody was thinking about transformers yet. Nearly two decades later, that same parallel-processing bet is why almost every model trained this year ran on their chips. Sometimes the platform outlives the plan. #RAISEHackathon