AMD expands PsiQuantum collaboration and plans equity investment
The partners will combine CPUs, GPUs and adaptive computing, with more details due at Chicago's SC26 conference on Nov. 15 to 20.
Julian Okafor
Commercially useful quantum computers will need conventional processors to correct errors and run scientific workloads. Advanced Micro Devices, Inc. NASDAQAMD, the designer of processors for data centres, personal computers and embedded systems, announced an expanded collaboration with PsiQuantum on Nov. 12.
PsiQuantum, a privately held Palo Alto quantum-computing developer founded in 2016, is building systems based on silicon photonics. The companies plan to co-design infrastructure for fault-tolerant quantum computers and applications combining quantum computing, high-performance computing and AI.
AMD also plans an equity investment in PsiQuantum, but the Nov. 12 release gives no amount, ownership percentage or closing date. It discloses no chip order value, delivery schedule or revenue commitment, with more partnership details promised at SC26 in Chicago on Nov. 15 to 20.
Shares in NASDAQAMD closed at US$552.90 at 4:00 p.m. EST on Nov. 11. That price preceded the partnership announcement, so it provides a reference price rather than a market response to the agreement.
AMD Adds CPUs And GPUs To Existing Adaptive Computing Work
The existing relationship centred on AMD Versal field-programmable gate arrays, which PsiQuantum selected for development of its fault-tolerant platform. The expanded arrangement adds AMD Instinct graphics processors, EPYC central processors and the ROCm software stack to the collaboration.
The Nov. 12 announcement places AMD on the conventional computing side of PsiQuantum's architecture, spanning infrastructure and application development. It also includes exploration of silicon-photonics networking and foundations for next-generation AI, although it attaches no separate product timetable to that work.
"Our work with PsiQuantum brings the breadth of AMD compute and software to this next frontier, combining our strengths to accelerate the path to utility-scale quantum computing," AMD chief technology officer Mark Papermaster said in the release.
A preprint submitted on Nov. 10 provides a specific technical result behind the relationship. Its authors describe PRISM, a machine-learning decoder that uses information about where errors occur in photonic quantum-computing protocols.
The abstract reports an optimized implementation on an AMD VP1902 FPGA with latency of 6.59 microseconds. It also reports an amortized decoding time of 1.56 microseconds per logical block, showing the scope and speed of the error-correction task being assessed.
Across its benchmarks, the authors say PRISM matches the logical error rates of integer-programming-based decoders while running significantly faster. Those results concern a decoder; they establish no commercial revenue, delivered utility-scale quantum computer or customer acceptance milestone.
Drug Discovery Is The Initial Application
The Nov. 12 release says PsiQuantum's computational-chemistry methods are already running on AMD hardware and performing beyond state-of-the-art calculations. The partners identify drug discovery as their initial application focus, including simulation of covalent inhibitors relevant to oncology research.
The release calls the simulation results breakthroughs but supplies no numerical comparison, named pharmaceutical customer or contract value. It says PsiQuantum plans to explore integrated accelerated-computing solutions for commercial applications, preserving a distinction between present research and future sales.
Quantum-computing theorist Scott Aaronson offered an earlier caution on commercialization in a Dec. 21, 2025 assessment. He expressed confidence in quantum-computing theory while remaining uncertain how many years implementation would take and which hardware approach would succeed first.
Aaronson identified simulation of physics and chemistry among quantum computing's established prospective applications. He cautioned that benefits for optimization and machine learning could be modest and take longer to win in practice, a qualification to the partnership's broader claims about hybrid AI computing.
AMD Already Has A Large Data Centre Business
AMD's Nov. 3 results put third-quarter revenue at US$13.4 billion, up 45 per cent from a year earlier. Data centre revenue reached US$8.5 billion, up 97 per cent, driven by EPYC processor demand and the launch of Instinct MI450 graphics processors.
The Nov. 12 partnership release gives no financial contribution that can be compared with those sales. It describes an expansion across AMD's computing portfolio while leaving the size of both its planned investment and potential chip business undisclosed.
Rigetti Computing, Inc. NASDAQRGTI, the Berkeley developer of superconducting quantum computers, reported third-quarter revenue of US$1.4 million in its Nov. 10 results. Its operating loss was US$33.2 million, illustrating the spending accompanying another hardware developer's commercialization effort.
Those figures extend the thread in Mugglehead's earlier report on Rigetti's quarterly loss and hardware timetable. Rigetti reported US$497.1 million in cash, cash equivalents and available-for-sale investments at Sept. 30, with no debt.
Read more: Rigetti Computing reports US$1.4M revenue and US$33.2M operating loss
AMD and PsiQuantum also plan to explore a footprint at Chicago's Illinois Quantum and Microelectronics Park, where PsiQuantum is the anchor tenant. The Nov. 12 release describes exploration, without a separate facility commitment or investment budget.
The next scheduled disclosure is SC26 in Chicago, running Nov. 15 to 20, when the partners say they will share more about the collaboration. Investment terms, infrastructure commitments and a delivery timetable remain the commercial details to test against that presentation.
