AMD to Take Equity Stake in PsiQuantum and Power Its Quantum Machines' Classical Side

AMD plans to take an equity stake in PsiQuantum, the Palo Alto startup building photonic quantum computers, and supply the classical computing that runs inside its machines — an investment whose size neither company disclosed Thursday.

The partnership, announced jointly, moves AMD from component vendor to partner and shareholder. It lands the chipmaker's full data center lineup — Instinct GPUs, EPYC CPUs and the ROCm software stack — behind a quantum bet at a moment when rivals IBM and Intel are staking their own quantum claims through very different routes.

What does the deal actually cover?

PsiQuantum had already chosen AMD's Versal FPGAs (field-programmable chips that can be reconfigured after manufacturing) for its fault-tolerant platform. The new arrangement broadens that relationship considerably.

• AMD takes an undisclosed equity stake in PsiQuantum.
• PsiQuantum's platform adds AMD Instinct GPUs, EPYC CPUs and the ROCm software stack.
• The two companies will co-design infrastructure for utility-scale quantum machines.
• Joint work will cover applications that mix quantum computing, high-performance computing and AI.

For AMD, the strategic logic is straightforward: a quantum computer big enough to do useful work needs a great deal of conventional computing wrapped around it. Real-time control, error correction and hybrid workloads all run on classical silicon. AMD wants to be the vendor — and now the shareholder — supplying that layer.

Mark Papermaster, AMD's chief technology officer, framed the bet as an extension of the company's record in supercomputers and AI infrastructure. "We see that same opportunity with quantum computing, where leadership in classical computing is essential to building systems at scale," Papermaster said in the joint announcement.

The unstated part: AMD has no qubit design of its own. The deal lets it claim a position in quantum computing without building quantum hardware — a hedge, not a moonshot.

What joint work is already running?

Some of the collaboration has produced results rather than roadmaps, which distinguishes it from many quantum announcements.

PsiQuantum researchers tuned a machine-learning decoder called PRISM to run with low latency on programmable hardware accelerators. Decoders are the classical systems that interpret error syndromes in real time, and low-latency decoding is a piece of the real-time error correction a large photonic machine will depend on. Without it, a fault-tolerant quantum computer cannot function.

The companies also reported early results in computational chemistry. PsiQuantum's methods, run on AMD GPUs, have produced early results in simulating covalent inhibitors — a class of molecules used in cancer drug research. PsiQuantum plans to build on that work and explore accelerated computing products for commercial use, starting with pharmaceutical research and drug discovery.

That is an early result, not a product. But it points to the near-term commercial logic: hybrid quantum-classical workloads in chemistry may generate revenue before a fully fault-tolerant machine exists.

PsiQuantum CEO Victor Peng pointed to a second payoff beyond quantum computing itself. The partnership, he said, will cover "new approaches for networking with silicon photonics and scaling the foundations of next-gen AI." PsiQuantum says its photonic design lets it build on high-volume semiconductor manufacturing and existing cryogenic infrastructure.

Silicon photonics — moving data with light on silicon chips — matters well beyond quantum machines, including in AI data center interconnects. Peng's comment signals the partnership's value may not hinge entirely on quantum timelines.

How crowded is PsiQuantum's cap table?

AMD joins a long list of backers, and the company's recent funding history shows how much capital photonic quantum computing absorbs.

• PsiQuantum raised $1 billion at a $7 billion valuation in September 2025, in a round led by BlackRock, Temasek and Baillie Gifford.
• Nvidia's venture arm, NVentures, was among the new investors in that round, according to participant Qatar Investment Authority.
• In May, the Commerce Department signed a letter of intent for $100 million in planned CHIPS Act funding for PsiQuantum's work on electro-optic materials, high-temperature single-photon detectors and low-loss photonic packaging.
• Under that CHIPS arrangement, the government takes a minority, non-controlling stake.

The detail worth noting: Nvidia — AMD's chief rival in AI accelerators — is already a PsiQuantum investor through NVentures. AMD's direct equity stake and supply agreement gives it a closer operational position, but the startup is not exclusive to anyone. PsiQuantum also announced the partnership publicly on X on November 12, 2026, describing plans to co-develop "technologies underpinning the future of computing."

Where does this fit in the quantum race?

The announcement landed a day after IBM updated its own quantum roadmap, and the contrast between the two approaches is instructive.

IBM, which builds superconducting machines, said Wednesday that its Nighthawk r3 processor will be built to run 10,000 operations on 520 programmable qubits, with plans to link two of them into a 1,040-qubit system in 2027. That is a hardware roadmap: bigger processors, more qubits, a dated milestone.

AMD is betting on the opposite layer. The company, which began shipping its 256-core Venice EPYC chips this week, is placing its chips on the classical hardware and software around the qubits rather than on a qubit design of its own. Intel took a third route last month, agreeing to buy quantum software firm Classiq.

Three incumbents, three strategies: IBM builds qubits, Intel buys software, AMD funds a startup and sells it the surrounding compute. Which approach pays off depends on whether utility-scale quantum machines arrive on photonic, superconducting or other substrates — a question no company has answered definitively.

What comes next?

The two companies are weighing a presence at the Illinois Quantum and Microelectronics Park in Chicago, where PsiQuantum is the anchor tenant. They plan to share more about the partnership at the SC26 supercomputing conference, which runs November 15-20 in the same city.

Expect the investment amount to surface eventually — undisclosed today does not mean undisclosed forever, and regulatory or funding filings may force the number into the open. The harder question is delivery: PsiQuantum's photonic approach promises leverage from conventional semiconductor manufacturing, but a utility-scale fault-tolerant machine remains unproven at any price. AMD's stake is a wager that the classical half of that machine will be valuable even if the quantum half takes longer than promised.