AI will not make high-performance computing (HPC) obsolete, in AMD executive Scott Tease’s view. He argues that AI can contribute to scientific discovery while complex simulations continue to use traditional HPC systems for double-precision FP64 calculations. In an interview published Nov. 16, 2026, Tease framed the challenge as assigning each task to the computing approach that fits it.
Umbra’s design assigns distinct jobs to classical computing, quantum technology and software coordination.
Why Tease sees AI and HPC working together
HPC refers here to powerful computing systems used for demanding scientific simulations. FP64, or double-precision floating-point arithmetic, is a format used for high-precision calculations. Tease’s argument is that AI can be part of scientific discovery without taking over every task handled by those systems.
His view centers on choosing an approach for each part of a workload: AI where it can be applied, GPU acceleration where it suits the task, and quantum computing where that approach fits. That frames AI and HPC as tools that can have different roles in the same scientific work, rather than interchangeable labels for one kind of computing.
How AMD assigns roles in the Umbra project
AMD describes Umbra as a codesign effort involving AMD, IBM and Oak Ridge National Laboratory. IBM contributes quantum technology, Oak Ridge provides a research and computing environment, and AMD supplies classical computing technologies and software. AMD’s announcement of the Umbra effort describes these intended roles:
| Technology | Role AMD assigns it |
| AMD EPYC CPUs | Orchestration and general-purpose computing |
| AMD Instinct GPUs | Simulation, optimization and parallel scientific computation around quantum workloads |
| AMD Versal adaptive SoCs | Low-latency control, readout and decoding near the quantum processor |
| AMD ROCm | Connecting applications and hardware |
The division gives the proposed system a practical shape: CPUs coordinate work, GPUs handle parallel computation, and adaptive SoCs are assigned tasks close to the quantum processor. ROCm is intended to connect the software and hardware. These are AMD’s design roles for the project.
What AMD says is deployed and planned
AMD describes a first colocated deployment that brings IBM quantum computers together with Umbra, an HPC system powered by AMD EPYC CPUs and Instinct GPUs. AMD says a connection to the Oak Ridge Leadership Computing Facility is planned. That planned facility connection is a separate step from the colocated deployment described by AMD.