On September 20, 2026, IBM Research published a paper demonstrating Spacetime Probabilistic Error Cancellation (Spacetime PEC), a hybrid error-mitigation protocol that reduced sampling overhead by up to 63 times compared to conventional Probabilistic Error Cancellation (PEC). The experiment ran on a 27-qubit heavy-hex superconducting processor, IBM’s standard architecture for its cloud-accessible quantum systems. The result, posted on the arXiv preprint server, addresses one of the most persistent bottlenecks in near-term quantum computing: the exponential number of circuit executions required to cancel noise.
What They're Actually Building
Probabilistic Error Cancellation is a leading error mitigation technique that works by running a noisy quantum circuit multiple times with randomly inserted recovery operations, then classically post-processing the results to produce an unbiased estimate of the ideal outcome. The catch: the number of samples needed grows exponentially with circuit depth and gate count, making PEC impractical for all but the smallest circuits. IBM’s Spacetime PEC tackles this by integrating Quantum Error Detection (ED) into the protocol. Error detection uses ancillary qubits to flag certain errors during computation, allowing the system to discard runs where errors are detected. By combining detection with cancellation, Spacetime PEC reduces the effective noise level before applying PEC, slashing the sampling overhead.
On benchmark circuits such as GHZ state preparation and quantum volume tests, the team measured a 63× reduction in the number of shots required to reach a target accuracy compared to standard PEC. The technique was tested on IBM’s 27-qubit Falcon-family processor with heavy-hex connectivity, the same chip available to IBM Quantum Network users. IBM’s 27-qubit processor has typical two-qubit gate error rates around 0.5–1%, making error mitigation essential for any circuit beyond a few hundred gates. IBM’s heavy-hex lattice reduces crosstalk and simplifies error correction, but it also limits qubit connectivity, making error mitigation more challenging. Spacetime PEC works within these constraints by using the existing qubits for both computation and detection, avoiding the need for additional physical qubits. The error detection step uses stabilizer measurements on ancilla qubits to identify runs with bit-flip or phase-flip errors, discarding them before applying PEC.
IBM’s roadmap targets a 100,000-qubit system by 2033, but near-term utility depends on squeezing more performance from noisy devices. Spacetime PEC is a step in that direction, though it remains a research demonstration, not yet integrated into the Qiskit Runtime service. IBM’s Qiskit Runtime already includes a PEC implementation; Spacetime PEC could be integrated as an upgrade, potentially reducing the cost of running error-mitigated jobs by a factor of 63 for suitable circuits. The hybrid approach is notable because it leverages the same physical qubits for both computation and error detection, avoiding the overhead of dedicated ancilla arrays. The detection step uses post-selection, meaning some runs are discarded, but the net effect is a dramatic reduction in total samples.
Winners and Losers
IBM’s advance puts pressure on competitors in the error mitigation race. Google Quantum AI has demonstrated similar hybrid techniques, including probabilistic error cancellation on its Sycamore and Willow processors, but has not reported overhead reductions of this magnitude. Google’s published overhead reductions have been in the range of 10–20× for similar circuits. Quantinuum and IonQ, which use trapped-ion qubits with inherently lower error rates (two-qubit gate fidelities above 99.8%), rely less on aggressive error mitigation, but they still face overhead challenges as they scale. A 63× reduction in sampling cost could make IBM’s superconducting platform more economically attractive for cloud users, where compute time is billed per second. Cloud quantum computing is typically priced per task or per second of QPU time; a 63× reduction in shots directly translates to a 63× cost reduction for the end user, assuming the same per-shot pricing. This could shift the competitive dynamics in the quantum cloud market, where IBM currently holds a leading share.
For quantum software startups like Q-CTRL and Riverlane, which build error suppression and mitigation middleware, IBM’s in-house protocol could be seen as a competitive threat, but it also validates the market for error mitigation tools. Cloud providers AWS and Microsoft Azure, which offer IBM backends, stand to benefit from improved performance. The bigger loser might be any quantum hardware company that cannot demonstrate comparable overhead reductions, as customers increasingly factor error mitigation efficiency into platform selection. If IBM can deliver a 63× reduction in sampling cost, its per-job pricing could undercut competitors, assuming comparable hardware performance.
The Bigger Picture
In 2026, quantum computing remains firmly in the NISQ era, with error mitigation serving as a bridge to fault tolerance. The U.S. government continues to fund quantum research through the CHIPS Act and the National Quantum Initiative, while the EU Quantum Flagship funds similar efforts. IBM’s result arrives amid a flurry of error mitigation milestones: in 2025, Google’s Willow processor demonstrated improved coherence and error rates, and Quantinuum achieved 56 logical qubits using error correction. Spacetime PEC is not error correction—it does not create logical qubits—but it makes noisy physical qubits more usable. The technique also aligns with the broader industry trend toward hybrid quantum-classical workflows, where classical post-processing is already a given. The technique could also be applied to other qubit modalities, though IBM has only demonstrated it on superconducting qubits.
The quantum computing market is projected to reach $8.6 billion by 2027, according to some estimates, with error mitigation software and services representing a growing segment. IBM’s advance could accelerate adoption by making quantum experiments cheaper. However, the technique still requires classical post-processing, which adds latency. Real-time error mitigation remains a challenge. The 63× overhead reduction is significant, but it’s measured on specific circuits and may not translate uniformly to all algorithms. The next step is to see whether IBM can integrate this technique into its cloud service and demonstrate practical speedups for real-world applications like quantum chemistry or optimization.
The Signal
The signal here is that hybrid error mitigation—combining detection and cancellation—can dramatically lower the cost of extracting useful results from noisy quantum hardware. A 63× reduction in sampling overhead is not incremental; it’s a meaningful improvement that could make certain quantum algorithms feasible on near-term devices where they were previously too expensive. However, the result is a single research demonstration, and the overhead reduction factor will vary by circuit. The real test will be whether IBM can productize Spacetime PEC in Qiskit Runtime and show consistent gains across a range of workloads. If IBM can demonstrate this on a 100-qubit device with similar gains, it would be a major validation. For now, it’s a promising paper, not a product.
In short: IBM’s Spacetime PEC reduces error-mitigation sampling overhead by 63×, making near-term quantum computations more feasible. The quantum computing industry will watch closely to see if this research translates into a practical tool for enterprise users.