AMD corporate vice president of HPC and sovereign AI Scott Tease stated that high-precision workloads will exist for a long time and that AI will never replace high-performance computing (HPC). Speaking to Tom's Hardware Premium, Tease argued that while AI is becoming an essential part of scientific discovery, the industry must move toward a heterogeneous compute stack that integrates AI, HPC, and quantum computing into a single, interconnected fleet.
Tease noted that the research community is shifting away from viewing AI as a threat to HPC, instead focusing on how to deploy it alongside traditional simulations. He argued that attempting to use disconnected "islands" of CPUs or GPUs is insufficient for modern research requirements. To address this, AMD is collaborating with IBM and Oak Ridge National Laboratory to deploy its classical computing stack in the Umbra HPC system alongside IBM’s 129-qubit Quantum Nighthawk quantum computer.
The push for integrated systems comes as a point of contention regarding hardware precision. While AI relies on low-precision data formats, HPC requires high-precision FP64 compute for accuracy. Tease suggested that Nvidia has prioritized low-precision formats for AI workloads at the expense of native FP64 performance. For comparison, Nvidia's Rubin GPU offers 33 TFLOPs of FP64 compute, which is lower than the 34 TFLOPs provided by the four-year-old H100. In contrast, AMD's Instinct MI355X peaks at 78.6 TFLOPs of FP64, while the upcoming MI430X is expected to peak at 288 TFLOPs.
Tease emphasized that the future of scientific computing lies in the coexistence of these technologies to solve complex problems in fields like chemistry and human health. "There [are] going to be a lot of problems where quantum will not be the be-all, end-all," Tease said. "So having those things coexist, having them in a single workload, is where we want to get to."














