Cerebras Systems’ new memorandum of understanding with the U.S. Department of Energy positions AI infrastructure as a strategic pillar of American science and national security.
At its core, the agreement signals alignment around the White House’s Genesis Mission, a national effort to fundamentally change how scientific research is conducted by embedding AI into discovery workflows. For the DOE, which oversees some of the world’s most advanced national laboratories, the partnership reflects a growing recognition that future breakthroughs will depend on tightly coupled AI and high-performance computing rather than traditional supercomputing alone. For Cerebras, it places its wafer-scale systems squarely within the federal roadmap for next-generation computing.
The MOU sets up a framework rather than a fixed program. It enables information sharing, joint exploration of R&D pathways, and future agreements focused on secure, scalable and energy-efficient AI infrastructure. Areas under discussion span large scientific datasets, new AI and AI+HPC architectures, advances in power, packaging and cooling, next-generation memory and I/O technologies, and the development of software and programming models capable of supporting converged AI-HPC workloads. Just as important, the agreement emphasizes joint developer engagement and public-facing work in research, education and policy.
A notable element is Cerebras’ intent to support DOE and its national labs in building and deploying wafer-scale AI supercomputers as accelerators for science and security missions. This includes experimenting with advanced AI models, integrated AI-HPC workflows, and emerging “AI co-scientist” concepts designed to augment human researchers rather than simply speed up calculations.
The announcement builds on a relationship that is already deep. DOE laboratories were among the earliest adopters of the Cerebras Wafer-Scale Engine, now deployed in multiple national lab environments. Over the past decade, this collaboration has produced tangible results: AI models applied to genomics and clean energy, repeated recognition at the Gordon Bell Prize level, and prize-winning work with Argonne National Laboratory on AI-driven analysis of COVID-19 genomic dynamics.
Cerebras systems have also been used to push beyond conventional performance limits on DOE mission workloads, from molecular dynamics simulations run with Sandia, Lawrence Livermore and Los Alamos to computational fluid dynamics at NETL. These efforts highlight why converged AI-HPC is increasingly viewed as a necessity rather than an experiment.
Equally strategic is the hardware co-design work emerging from DOE programs such as Advanced Memory Technology, where Cerebras and national labs are exploring new memory systems that could dramatically expand the scale of scientific simulations and AI workloads, potentially by orders of magnitude.
Taken together, the MOU underscores a broader shift in U.S. technology policy: AI infrastructure is no longer treated as a commercial add-on but as national research infrastructure. By formalizing collaboration under the Genesis Mission, DOE and Cerebras are effectively betting that leadership in science, security and advanced computing will be determined by how well AI and HPC are fused and how quickly those capabilities can be brought to scale.
