we_are_coded.by CODE · The world, decoded
БГ

NVIDIA · Infra

NVIDIA at SC26: Nemotron for Science, an expanded Vera CPU platform and Rubin HPC, plus a quantum platform with IBM and Fermilab

NVIDIA announced on 17 November Nemotron for Science, an expanded Vera CPU platform for high-performance computing and the Rubin HPC GPU, plus a plan for a quantum platform with IBM and Fermilab.

Source
NVIDIA
Published
The event
Checked on
Responsible editor
Cvetelin Ivanov
How we work
MethodCorrections

An agent from KTH repairs broken GROMACS setups on its own. In one overnight run, on a single DGX Spark, it fixed 81% of the test cases. That is the number NVIDIA brought to SC26 in Chicago, and it is about the small work that used to eat a researcher's time.

Fact

The facts: on 17 November 2026 NVIDIA announced at SC26 in Chicago Nemotron for Science, the Rubin HPC GPU and an expanded Vera CPU platform for high-performance computing, plus a plan for a quantum platform with IBM and Fermi National Accelerator Laboratory. According to NVIDIA, Nemotron for Science combines the open Nemotron 3.5 Super VL model with a new open chemistry dataset and post-training recipes for life sciences and chemistry, and runs locally on a DGX Spark, a workstation or your own supercomputer. According to NVIDIA, an agent built with the toolkit took a single prompt at Argonne National Laboratory, planned and tested hypotheses, chose which simulations to run and identified two new binding sites. According to NVIDIA, an agent ported a routine from the WRF model to GPUs and sped it up more than 5x using about 150 million tokens. According to NVIDIA, Rubin HPC supports precisions from FP64 to FP4, adds sparse-data acceleration and connects over seventh-generation NVLink; with more than 50,000 CUDA cores it has about three times as many as Hopper and is expected in 2028. DKRZ, the German Climate Computing Center, has selected Vera CPUs for its HLRE-5 supercomputer, expected in early 2028. According to NVIDIA, a simulation on Jülich's JUPITER supercomputer showed that a technique called magic state cultivation can cut the operations needed for one logical step tenfold. NVIDIA, IBM and Fermilab announced plans for a quantum computing co-design platform to support error-correction and applications research. According to NVIDIA, a new CUDA-Q decoding algorithm, TCBP-JOBS, cuts error-correction research computation from 12 weeks to 45 minutes.

Here is what is really moving. The agent is not a chat that answers you. It plans, runs code, looks at the result and tries again, on your own machine. For that loop NVIDIA sells CPUs.

The model plans the experiment. The chip feeds it.

The quantum part is with IBM and Fermilab. The plan is a co-design platform to support error-correction and applications research, and NVIDIA and IBM are pairing DGX B200 systems with IBM quantum computers, including those powered by IBM Quantum Nighthawk. According to NVIDIA, the new TCBP-JOBS algorithm cuts error-correction research computation from 12 weeks to 45 minutes, and the goal is to bring decoding latency from about 300 microseconds down to 10.

The numbers are NVIDIA's and there are no independent ones. The direction, though, is visible without them: what is being sold is the loop in which the agent plans, runs and checks, and the chip is only the current for it.

The visual is generated code-art, no third-party images.