International Edition
Latest News
Technology

NVIDIA Unveils DGX Spark and Rubin HPC at SC26 Conference

NVIDIA is launching the DGX Spark HPC Developer Kit this week at the SC26 supercomputing conference in Chicago. Designed for academics, researchers, and developers, the kit provides a four-node cluster capable of running multi-node scientific workloads and AI…

NVIDIA Unveils DGX Spark and Rubin HPC at SC26 Conference

NVIDIA is launching the DGX Spark HPC Developer Kit this week at the SC26 supercomputing conference in Chicago.

Designed for academics, researchers, and developers, the kit provides a four-node cluster capable of running multi-node scientific workloads and AI agents directly at a desk. The system aims to eliminate the wait times typical of shared GPU clusters by providing consistent, local access to high-performance computing resources.

The kit integrates the GB10 Grace Blackwell Superchip into a compact, desk-ready form factor. Each of the four nodes connects to its two neighbors in a ring configuration. An NVIDIA HPC SDK installation playbook is expected to follow shortly to support the hardware.

Lambda is incorporating NVIDIA Photonics co-packaged optics (CPO) switches to manage the demands of massive AI clusters. These switches connect NVIDIA GB300 NVL72 racks, which house more than 10,000 GPUs. According to the NVIDIA Blog, this implementation aims to deliver lower power consumption, higher resiliency, and longer mean time between interruptions for large-scale AI factories.

“Networking has to keep pace with the demands of increasingly large AI clusters,” said Dave Ward, president of cloud services at Lambda, in a statement reported by the NVIDIA Blog. “Bringing co-packaged optics into a deployment of this scale is an important step toward building the infrastructure needed to support the next generation of AI.”

NVIDIA is also expanding its Slurm support to simplify the integration of quantum processors at supercomputing centers. By enabling QPU builders to create a single interface for their hardware using a simple file, the update builds on collaborations with IBM and Pasqal. This effort responds to a market where quantum deployments have grown nearly sevenfold over the past five years, with approximately 30 centers worldwide now hosting more than 50 quantum processors.

The company unveiled the Rubin HPC GPU at SC26 as part of the NVIDIA Vera Rubin platform. Engineered for science, engineering, and analytics, the GPU features more than 50,000 CUDA cores and utilizes seventh-generation NVIDIA NVLink interconnect technology. The architecture includes sparse data acceleration to maintain flow, specifically targeting faster simulation times in fields like molecule design and material science.

As the CUDA platform marks its 20th anniversary at SC26, NVIDIA has released updated CUDA-X libraries, including cuPhoton, cuPotential, and cuSpa. These tools are designed to help AI agents process vast datasets, supporting research from astrophysics to single-cell biology.

The release arrives as the company honors the developers who established GPU computing as a standard foundation for modern science, a platform that has grown significantly since its inception two decades ago.

About the author: Anika Shah - Technology

MSc in Computer Science, senior reporter. Anika focuses on AI ethics, cybersecurity, and emerging hardware—frequently moderating panels at CES and Web Summit. “Anika Shah decodes tech breakthroughs and startup disruption shaping tomorrow’s digital landscape.”