ORNL, IBM and AMD Launch Quantum Error Correction Platform

Oak Ridge National Laboratory, IBM and AMD have opened a co-design platform for quantum error correction that puts IBM quantum systems, including the 120-qubit IBM Quantum Nighthawk, next to ORNL’s Umbra high-performance computer, powered by AMD Instinct MI355X GPUs and 5th Gen AMD EPYC CPUs, the companies said in IBM and AMD newsroom releases on 5 November 2026.

The launch is a shared research environment, not a claim of new logical-qubit records. IBM’s framing is that early error-correction advances now need to be scaled with more classical computing, and that DOE’s Quantum Genesis Mission — through its Q Competition — is aiming at 100 logical qubits running 100,000 hard operations by late 2028. IBM said its own published experiments have reached up to 70 logical qubits and 438 hard operations, a gap it described as “roughly 230-fold” in hard-operation count.

Classical compute next to the qubits

IBM said the first deployment lets researchers evaluate tight integration of classical and quantum machines for decoding, optimisation and software orchestration. Nighthawk’s square-lattice connectivity, circuit throughput and real-time pulse control are presented as the quantum side of that testbed; Umbra’s AMD GPUs and CPUs are the classical side. The partners also plan to connect the platform to the Oak Ridge Leadership Computing Facility (OLCF), so quantum work can sit beside broader AI and scientific computing stacks rather than in isolation.

Travis Humble, director of the Quantum Science Center at ORNL, said solving quantum error correction is essential if quantum computing is to become a practical tool for scientific discovery, and that with Umbra the centre will “test new ideas, evaluate approaches at scale, and build the foundation for more reliable quantum computers.” Jay Gambetta, IBM Fellow and director of IBM Research, said the question is “no longer if error correction will work on quantum hardware” but how to accelerate and optimise it so scientists can tackle meaningful problems. AMD’s Mark Papermaster and Ralph Wittig cast the company’s role as supplying the classical foundation — CPUs, GPUs, adaptive SoCs and open software — that future hybrid systems will need.

How this sits with earlier quantum milestones TSN has covered

The co-design pitch is continuous with work TSN has already tracked on IBM’s software stack. IBM’s Directed Execution beta pushed error-mitigation choices back toward the user’s side of the Qiskit Runtime pipeline; today’s announcement is about putting the classical horsepower for decoding and orchestration in the same facility as the qubits. Classical simulation of error-correction ideas has also moved: NVIDIA’s JUPITER magic-state work showed large-scale simulation of quantum-error techniques on GPU supercomputers, while Google Quantum AI’s magic-state cultivation on Willow reported hardware progress on a key fault-tolerance resource. ORNL’s platform is the co-location version of that same race — hardware, HPC and software asked to fail and recover together.

IBM also said the effort builds on earlier ORNL collaborations, including quantum-plus-Frontier workloads on magnetic materials and fusion-fuel molecule configurations, and on AMD work accelerating sample-based quantum diagonalization and FPGA-style decoding demos. Those are prior results, not today’s new data.

Co-location is the product until the numbers move

The honest reading of the 5 November releases is architectural, not experimental. A 120-qubit Nighthawk sitting beside Umbra does not by itself close IBM’s stated hard-operation gap to the DOE Q Competition target. What it does create is a place where concatenated codes, space-time codes, algorithm-aware decoders and hardware-aware compilation can be timed against real pulse-level feedback and real GPU/CPU spend, instead of being argued from separate papers.

For UK and European readers, the story is still a US national-lab / vendor co-design move under DOE’s Quantum Genesis framing. There is no UK deployment claim in the releases. The transferable lesson is narrower: if your organisation is planning quantum-ready scientific workflows, the bottleneck the partners are designing around is classical — decoding latency, orchestration and the cost of mid-circuit measurement — not only qubit count. That matches how procurement teams already talk about hybrid AI/HPC stacks: the interesting contract is the one that specifies the interface between accelerators, not the one that only quotes peak FLOPS or peak qubits.

Watch next whether OLCF connectivity is demonstrated with named workloads, whether IBM restates the 70-logical-qubit / 438-operation baseline with new numbers from this platform, and whether AMD publishes concrete Versal or ROCm recipes for near-qubit decoding. Until those land, the confirmed fact is the platform launch itself: quantum error correction research now has a co-located IBM–ORNL–AMD testbed aimed at the 2028 Genesis timeline, without a new result sheet attached to the announcement.

Sources

• Oak Ridge National Lab, IBM, and AMD Launch New Co-Design Platform for Quantum Error Correction (IBM Newsroom, 5 November 2026)
• AMD Builds the Classical Foundation for Quantum’s Future (AMD Newsroom, 5 November 2026)

Related stories

• IBM Quantum Launches “Directed Execution” in Beta
• NVIDIA Simulates 104 Qubits on JUPITER to Test a Quantum Error Fix
• Google Quantum AI Shows “Magic State Cultivation” Working on Its Willow Chip