IBM, Oak Ridge National Laboratory and AMD have introduced a co-design platform that brings a quantum processor and a high-performance computing system into a shared research environment for quantum error correction. Announced on November 5, the first deployment connects IBM’s 120-qubit Quantum Nighthawk processor with Oak Ridge’s Umbra system, which uses AMD Instinct MI355X GPUs and fifth-generation EPYC CPUs. The project targets a central obstacle in quantum computing: turning fragile quantum operations into computations that can remain reliable as systems become larger.
Quantum computers process information using quantum bits, or qubits, whose states can represent information in ways that differ from conventional binary bits. Those states are highly sensitive to noise from the surrounding environment and imperfections in control hardware. An error can corrupt a calculation before it finishes. Error-correction methods attempt to detect and correct those faults by distributing quantum information across multiple physical qubits and repeatedly measuring selected properties without directly destroying the encoded information.
That process creates a substantial classical computing workload. Researchers must decode measurement results, identify likely errors and decide which corrective operations should follow. As quantum processors grow and circuits become more complex, the supporting computation can become a bottleneck. IBM says the new platform aims to let researchers study how quantum hardware, classical accelerators and software should be designed together, rather than treating the quantum processor as a standalone machine.
Quantum error correction needs more classical computing
The initial system places IBM’s Quantum Nighthawk processor alongside Oak Ridge’s Umbra high-performance computing system. IBM identifies Nighthawk as a 120-qubit processor. Umbra is powered by AMD Instinct MI355X graphics processing units and fifth-generation AMD EPYC central processing units. The AMD hardware supplies the classical computing capacity for research into error correction, decoding optimisation and software orchestration. The companies describe the arrangement as a co-design platform, meaning that researchers can evaluate the interaction between the quantum device and the conventional computing resources supporting it.
This is not simply a matter of sending completed quantum jobs to a distant supercomputer. IBM says its Quantum System API provides researchers with low-level access to the quantum hardware, including pulse-level control, mid-circuit measurements and the ability to respond to measurement results during a calculation. Researchers can inspect the signals and operations used by the device and test how control strategies affect error-correction methods. The aim is to study the full workflow, from hardware behaviour through decoding and software execution.
A shared platform links Nighthawk and Umbra
The partners intend to compare different approaches by measuring what each method requires in qubits, execution time, samples and classical processing. That comparison matters because an error-correction scheme that works in a small demonstration may become too expensive or slow when scaled. GPU accelerators could help with some decoding and optimisation workloads, but the most suitable division of work between quantum and classical processors remains an active research question. The platform gives the partners a place to test those trade-offs on integrated equipment.
IBM’s announcement links the work to the US Department of Energy’s Quantum Genesis Mission. The mission’s stated target is to reach 100 logical qubits capable of performing 100,000 hard operations by late 2028. Logical qubits are error-corrected units built from multiple physical qubits; they aim to preserve quantum information more reliably than an individual physical qubit. The distinction is important because a processor’s physical-qubit count alone does not show how many dependable operations it can perform.
The 2028 target remains a demanding research goal
IBM says its existing progress includes experiments involving as many as 70 logical qubits and 438 hard operations. The company calculates that reaching the mission’s target would require roughly a 230-fold increase in the hard-operation count. That comparison shows the scale of the remaining work. It should not be read as a providey forecast for the new platform or as evidence that the target has already been achieved. The platform is a research tool intended to help investigators test methods that may contribute to that progress.