AMD has just scored one of its biggest wins yet in the AI and supercomputing space. The U.S. 
Department of Energy (DoE) has officially tapped the chipmaker to power two next-generation supercomputers that could redefine America’s scientific computing capabilities. The deal marks not only a massive milestone for AMD but also a strategic shift in how the government is diversifying its AI hardware partners beyond the NVIDIA ecosystem.
According to a report from Reuters, AMD’s upcoming Instinct MI355X and the newer MI430 AI chips will form the backbone of the two government-backed systems, codenamed Lux and Discovery. Both projects are part of a nearly $1 billion investment by the DoE, which aims to enhance the nation’s high-performance computing (HPC) infrastructure for cutting-edge research, climate modeling, and national security simulations.
The first of these systems, Lux, is set to go online within six months – a record deployment timeframe. Built in collaboration with Hewlett Packard Enterprise (HPE), Oracle, and Oak Ridge National Laboratory, Lux will run entirely on AMD’s MI355X chips. According to AMD CEO Lisa Su, this rapid rollout demonstrates how mature AMD’s HPC and AI ecosystem has become. The system will serve as a real-world benchmark for how AMD’s accelerators perform in large-scale AI workloads.
The second system, Discovery, slated for completion in 2028, is even more ambitious. It will feature a custom variant of AMD’s Instinct MI430 – a next-gen chip specifically tuned for AI-driven simulations and exascale performance. Discovery has reportedly been in planning for over a year, with AMD now confirmed as the primary computing provider. The project underscores the government’s confidence in AMD’s hybrid architecture approach, which integrates CPUs and GPUs in a unified memory structure, as seen in AMD’s MI300A super APU.
While NVIDIA remains a dominant player in AI compute, its focus on data center and cloud customers – and the limited FP64 capabilities of its upcoming Blackwell and Rubin architectures – may have given AMD a unique advantage here. The DoE’s past success with AMD-based systems like Frontier, currently one of the world’s fastest supercomputers, likely influenced the decision. Simply put, the government knows AMD’s ecosystem and trusts its roadmap.
Analysts suggest this partnership could signal a broader trend. With NVIDIA’s supply chain stretched thin for at least two years, major institutions are exploring alternatives that offer both flexibility and scalability. AMD’s combination of open software, competitive performance, and strong power efficiency makes it an appealing choice for public-sector projects requiring predictable long-term supply.
Though NVIDIA will almost certainly reappear in future government collaborations, AMD’s current momentum is undeniable. For now, Team Red holds the spotlight – and the Department of Energy seems more than happy to let them lead the next era of supercomputing innovation.