CHICAGO, November 16, 2026 — NVIDIA has announced an expansion of its Vera CPU platform, a rack-scale computing system designed to support demanding scientific workloads including climate modelling, molecular dynamics and large-scale simulation. The announcement was made around the SC26 supercomputing conference and positions the platform as an option for research institutions and high-performance computing centres seeking greater processing capacity and energy efficiency. NVIDIA’s official announcement describes the platform and its intended applications.

The rack-scale design combines 192 Vera CPUs with NVIDIA Quantum-X800 InfiniBand networking and software intended to accelerate scientific applications. The company says the architecture is built for workloads that do not necessarily benefit from GPU acceleration but still require high throughput and efficient communication between processors. These workloads can include simulations of molecules, weather systems and complex physical processes. NVIDIA’s product release provides the technical overview.
One of the early adopters named by NVIDIA is the German Climate Computing Centre, known as DKRZ, which has selected Vera CPUs for its HLRE-5 supercomputer. Climate research relies on powerful computing systems to process observations, simulate atmospheric and ocean conditions and assess possible future scenarios. More capable computing infrastructure can allow researchers to run more detailed simulations or complete existing calculations more efficiently, although the scientific value still depends on the quality of the underlying models and data. The company’s announcement names DKRZ and its planned system.
NVIDIA says Vera is designed to deliver twice the energy efficiency of x86 CPUs for the workloads highlighted in its announcement. Such comparisons depend on the systems, software, configurations and benchmarks used, and should be assessed against independent testing as the platform becomes available. Energy efficiency is increasingly important because large computing installations consume substantial electricity and require sophisticated cooling systems. Better performance per watt could help research centres manage operating costs while expanding their computing capacity. NVIDIA’s technical release contains the company’s stated efficiency claim.
The company also announced software support, including NVIDIA HPC SDK 26.11 and a public beta of a next-generation Fortran compiler developed with high-performance computing centres and national laboratories. Fortran remains important in scientific computing because many established research applications were written in the language. Improved compiler support could help institutions accelerate existing software without having to rewrite entire codebases. NVIDIA outlines the software tools in its announcement.
Vera CPU systems are expected to be offered by Dell Technologies, HPE, Lenovo and Supermicro beginning in early 2027, according to NVIDIA. The announcement therefore represents a platform roadmap rather than immediate availability of every planned rack system. If delivered as described, the technology could strengthen the computing infrastructure available to universities, national laboratories and research organisations working on complex scientific problems. Read NVIDIA’s full Vera CPU announcement.