NVIDIA Runs Simulation of Fault-Tolerant Quantum Computing at Jülich Supercomputing Centre

thequantuminsider.com Oct 19, 2026

NVIDIA Runs Simulation of Fault-Tolerant Quantum Computing at Jülich Supercomputing Centre

Curated by Shubham · TechCapsules · Oct 19, 2026

AI-summarised brief · reviewed before publication

NVIDIA used Europe’s JUPITER exascale supercomputer to simulate a 104‑qubit fault‑tolerant quantum processor, focusing on magic‑state cultivation under realistic hardware noise. The run employed 8,192 GPUs on NVIDIA GH200 superchips and its quantum‑computing software, achieving the largest comprehensive fault‑tolerant simulation to date. Results indicate the studied approach could cut the number of qubits required for magic‑state production by roughly 90 % versus earlier techniques. Researchers say the high‑fidelity data will let designers evaluate processor architectures before hardware is built, shortening development cycles. NVIDIA’s GB200 system at its Accelerated Quantum Research Center was used to scale tests before moving to JUPITER. The work aims to accelerate quantum applications in drug discovery, finance and materials science and enable faster commercialization of quantum solutions.

💡 Why It Matters

· By slashing the qubit overhead for magic‑state generation, the simulation clears a major engineering hurdle, making scalable fault‑tolerant machines far more attainable.

Read full article on thequantuminsider.com

NVIDIA and Jülich Run 104-Qubit Magic State Cultivation Simulation on JUPITER Supercomputer

NVIDIA and the Jülich Supercomputing Centre executed a 104‑qubit fault‑tolerant quantum simulation on JUPITER, Europe’s first exascale machine, during GTC Berlin 2026. The simulation modeled magic‑state cultivation under realistic hardware noise, validating a protocol that cuts physical qubit overhead by up to 90% versus traditional cascade distillation. The work, detailed in NVIDIA’s technical report, complements Google Quantum AI’s cultivation theory and demonstrates the CUDA‑Q platform’s scalability across European supercomputing hubs. It marks the first exascale‑level validation of a key fault‑tolerant building block.

💡 By proving that a 104‑qubit protocol can run efficiently on an exascale system, the study removes a critical scalability barrier for practical quantum computers. It signals that large‑scale, noise‑aware quantum simulations are now within reach, accelerating the transition from theoretical models to deployable quantum hardware.

therobotreport.com · Oct 19

Unleash AI Innovation: The Power of NVIDIA RTX PRO 6000 Blackwell Workstation…

PNY Technologies announced the availability of NVIDIA RTX PRO 6000 Blackwell Workstation Editions, including a Max‑Q variant that can house up to four GPUs delivering 14.2 petaflops of AI compute and 384 GB of GPU memory. The systems target AI training, inference, data‑science preprocessing, and physical‑AI workloads such as robotics simulation, offering local processing power comparable to data‑center resources. Integrated with NVIDIA’s software stack—CUDA‑X, RAPIDS, NGC, AI Workbench, and Blueprints—the workstations enable rapid prototyping, fine‑tuning of generative models, and seamless scaling to cloud or on‑premise clusters. Enterprise users can embed the machines into existing IT environments, supporting high‑resolution creative tools like Adobe Photoshop Neural Filters and DaVinci Resolve face tracking while reducing cloud latency and costs.

💡 By delivering data‑center‑grade AI performance on a desktop, the RTX PRO 6000 Blackwell workstations let organizations iterate locally, cutting reliance on expensive cloud compute and accelerating time‑to‑innovation.

techcrunch.com · Oct 19

Nvidia launches new open models to run agents in the enterprise and…

Nvidia unveiled a new family of open‑weight AI models aimed at both enterprise agents and edge devices. The flagship Nemotron 3.5 Super VL, a 120‑billion‑parameter model, is positioned for always‑on enterprise agents and claims accuracy comparable to the larger Nemotron 3 Ultra while outperforming Anthropic and OpenAI models on specialized tasks such as document intelligence and financial accounting. For local deployment, Nvidia introduced two lightweight Bolt models, including a 74‑billion‑parameter Bolt 74B‑A9B for multistep reasoning and a smaller Bolt E4B for conversational agents, designed to run on Nvidia RTX Spark‑powered AI PCs and DGX Spark workstations now shipping through partners like Microsoft and Dell. Nvidia emphasizes that the models can be post‑trained on private data and run on Nvidia infrastructure without sacrificing capability or deployability.

💡 By offering open, high‑performance models that can be fine‑tuned and run on‑premises, Nvidia gives enterprises and consumers a viable alternative to closed‑source AI services, potentially reshaping data‑privacy and cost dynamics in AI adoption.

semiconductor-digest.com · Oct 15

NVIDIA Launches Semiconductor Simulation Tools for AI-Defined Fabs

NVIDIA announced the expansion of its CUDA-X library suite, adding cuPotential for accelerated atomic‑scale simulations, cuEST for electronic behavior modeling, and cuDSS for complex engineering calculations. The move targets AI‑defined fabs, where AI agents optimize manufacturing processes and improve yield. Early adopters such as Synopsys, Cadence, COMSOL, and Voltai report speedups ranging from 2‑3× to 30×, enabling faster digital twin testing before production changes. NVIDIA’s libraries aim to automate more engineering tasks across chip design and fabrication.

💡 By slashing simulation times, NVIDIA’s tools let fabs iterate designs faster, directly addressing the bottleneck of scaling advanced chip production. This acceleration could shorten time‑to‑market for next‑generation processors.

quantumcomputingreport.com · Oct 15

Riverlane and NVIDIA Demonstrate Accelerated QEC Decoder Calibration Using GPU Processing

Riverlane and NVIDIA released benchmark results showing that NVIDIA H100 GPUs can accelerate the recalibration of quantum error‑correction decoders. The Local Clustering Decoder, adapted for GPU parallelism, evaluates candidate weight settings against live syndrome data, reducing surface‑code training epochs from minutes to seconds. Faster calibration compensates for qubit parameter drift such as frequency shifts and gate‑fidelity loss, improving logical error rates by several percent up to a factor of two versus uncalibrated priors. The workflow draws on optimization methods pioneered by Google Quantum AI and positions GPUs as offline learners while Riverlane’s Deltaflow handles real‑time decoding. The demonstration underscores the expanding role of heterogeneous classical hardware in the quantum control stack. The results were validated on a 127‑qubit superconducting processor.

💡 Accelerated decoder tuning shrinks the downtime between experiments, enabling tighter feedback loops that push quantum devices closer to fault‑tolerant operation.

semiconductor-digest.com · Oct 14

Siemens, NVIDIA Accelerate Computational Lithography Up to 12×

Siemens announced that its Calibre software now incorporates NVIDIA’s cuLitho GPU‑accelerated library, delivering up to 12× faster computational lithography simulations. The partnership targets the “runtime wall” that hampers high‑NA EUV and curvilinear patterning, such as Inverse Lithography Technology, by providing 10–17× speed‑ups for key workloads. Faster simulations enable more optimization iterations, earlier hotspot detection, and tighter process margins, making full‑chip curvilinear patterning viable for next‑generation AI and memory chips.

💡 By slashing lithography runtimes, the collaboration unlocks deeper design‑time optimizations that directly improve yield and reduce time‑to‑market for the most advanced semiconductor nodes.