IBM says it has executed one of its key quantum computing algorithms in real time on AMD field-programmable gate arrays, moving a core function of quantum control from custom hardware to widely available reconfigurable chips. The algorithm, first announced in June, is designed to dynamically detect and correct quantum errors as they occur. By swapping in standard AMD hardware for bespoke control units, IBM is targeting cheaper and more practical hybrid systems that combine classical and quantum components.
Jay Gambetta, IBM’s vice president of Quantum, said the FPGA-based implementation runs 10 times faster than what is required for live error correction, calling the result a big deal for real-world quantum computing. The company says the demonstration puts its development timeline a full year ahead of schedule for the 2029 Starling quantum system roadmap. The performance headroom is intended to ensure that error correction can keep pace with increasingly complex workloads as qubit counts and experiment scale grow.
Quantum computers rely on qubits that can represent 0 and 1 simultaneously, but they are highly susceptible to disturbances that introduce calculation errors. IBM’s method monitors and fixes these errors in real time, helping keep systems stable during quantum operations. Running this control algorithm on mainstream AMD chips underscores the growing convergence of classical and quantum computing, enabling mixed setups where standard hardware handles orchestration and error correction on the fly. The approach promises lower cost, simpler system design, and a more straightforward path to scaling, potentially narrowing the gap between today’s nascent machines and future fault-tolerant systems suitable for business use.
The IBM and AMD partnership is expanding beyond control hardware. In August, the companies announced a close collaboration on quantum-centric supercomputing that links quantum computing with high-performance computing. In early October, they also unveiled a joint effort to provide advanced Artificial Intelligence infrastructure to Zyphra, an open-source Artificial Intelligence research and product company in San Francisco. Under a multi-year agreement, IBM will supply Zyphra with a large cluster of AMD Instinct MI300X GPUs on IBM Cloud to train next-generation multimodal foundation models.
