AMD expands PsiQuantum partnership to build quantum computing infrastructure

The expanded partnership centers on co-designing the hardware and software that will sit around PsiQuantum’s quantum machines. AMD brings its CPUs, GPUs, FPGAs and its ROCm software platform to the table. PsiQuantum brings its photonic quantum technology, which uses particles of light as the basis for computation. The goal is to make the two worlds work together as a single system.

Each AMD product has a distinct role in that kind of setup. CPUs are general-purpose processors. GPUs handle massive parallel workloads, which is why they power most AI training today. FPGAs are reprogrammable chips that can be rewired for specialized, time-sensitive tasks after they leave the factory. ROCm is AMD’s software stack for running high-performance and AI workloads on its hardware.

The companies say the collaboration will focus specifically on hybrid quantum-classical applications and artificial intelligence. A hybrid approach means splitting a problem so each type of computer handles the part it does best. The first commercial targets are pharmaceutical research and drug discovery. That choice is not random. Simulating how molecules behave is one of the problems most often cited as a natural fit for quantum machines.

The two companies are also in discussions about establishing a footprint at Chicago’s Illinois Quantum and Microelectronics Park. Those talks have not been finalized, so the location remains a possibility rather than a plan locked in concrete. PsiQuantum CEO Victor Peng framed the work around three priorities.

The phrase doing the heavy lifting in this announcement is “fault-tolerant.” Quantum bits, or qubits, are notoriously fragile. Small disturbances from the environment can introduce errors into a calculation. A fault-tolerant machine is one that can detect and correct those errors as it runs. Getting there requires a lot of classical computing muscle. Error correction involves constantly monitoring the system and making fast decisions about what went wrong. That is precisely the kind of real-time processing where chips like FPGAs and GPUs earn their keep.

“Utility-scale” is the other key term. It signals machines large and reliable enough to solve useful problems, not just run lab demonstrations. PsiQuantum’s bet on photonics also shapes the infrastructure challenge. Because its approach is built on light, networking and connecting components with silicon photonics is central to scaling up, which explains why Peng singled it out.

For AMD, the deal is a way to secure a seat in quantum computing without building a quantum processor itself. The undisclosed investment size leaves an open question about how deep AMD’s commitment really runs. Analyst commentary on the immediate impact has been limited so far. Several things are worth watching from here. One is whether the Chicago site moves from discussion to a signed commitment. Another is whether AMD eventually discloses the size of its equity stake.