Sandia National Laboratories has deployed four Cerebras CS-3 systems for a testbed.

The cluster, which will eventually comprise eight Cerebras CS-3 nodes, was installed in October of this year and will be used to expand research into AI workloads for national security missions.

Dubbed Kingfisher, the cluster has been funded by the National Nuclear Security Administration (NNSA) as part of its Advanced Simulation and Computing Artificial Intelligence for Nuclear Deterrence strategy (ASC AI4ND).

Justin Newcomer, senior manager of the ASC program at Sandia said: "The Cerebras CS-3 system positions us to be able to develop large-scale trusted AI models on secure internal Tri-lab (Sandia, Lawrence Livermore and Los Alamos Laboratories) data without many of the memory and power challenges that GPU systems face."

The Cerebras CS-3 nodes feature Cerebras' third-generation wafer scale engine architecture - WSE-3 - which the company describes as an alternative to traditional accelerators used for AI.

“We’re excited to expand our collaboration with Sandia with the deployment of this new Cerebras CS-3 cluster, which consists of four of our 3rd-generation wafer-scale systems,” said Andy Hock, senior vice president of product and strategy at Cerebras. “Building on Sandia’s and Cerebras’ history of record-setting AI and HPC performance and award-winning research, we look forward to seeing how this new, powerful cluster will enable Sandia researchers to uncover new breakthroughs across science, energy, national security and more.”

The WSE-3 is described as being approximately the size of a dinner plate. The company uses optical lithography and other methods to create a large number of transistors, cores, and individual processor ‘dies’ on the wafer, which unlike other chips is not then directed into individual dies.

WSE-3 was announced in March 2024 and has four trillion transistors and 900,000 "AI Cores," along with 44GB of on-chip SRAM. When combined with the Cerebras CS-3 system, the company claims the chip is capable of 125 peak AI petaflops. That can be scaled up to 2,048 CS-3s in a single cluster, for a purported 256 exaflops of AI compute.

“The deployment of the Cerebras CS-3 system at Sandia is a significant milestone in our journey to lead in AI and machine learning innovation,” said Jen Gaudioso, director of the ASC Program at Sandia. “This advanced testbed aligns perfectly with the DOE’s FASST (Frontiers in Artificial Intelligence for Science, Security, and Technology) initiative, enabling us to explore and develop cutting-edge AI technologies that are crucial for future national security missions.”

The Cerebras systems will not only be used for AI, but will also do more traditional modeling and simulation workloads.

Sandia National Laboratories has two main labs - one in Albuquerque, New Mexico, and the other in Livermore, California.

Sandia is operated and managed by National Technology and Engineering Solutions of Sandia, LLC., a wholly owned subsidiary of Honeywell International, and operates as a contractor for the US NNSA. The laboratories focus on nuclear weapons, national security programs, energy, and global security programs.

Sandia is also home to Intel's Hala Point system - a 1.15 billion neuron system - which is claimed to be the world's largest neuromorphic system.

In May of this year, Sandia was testing Submer's immersion cooling at its HPC center.

Cerebras submitted a draft registration statement with the US Securities and Exchange Commission for a proposed initial public offering (IPO) in August.