IBM and UChicago Demonstrate Verified Quantum Advantage Using 70 Logical Qubits

For decades, the promise of quantum computing has been accompanied by a quiet, persistent paradox: if a machine solves a problem so complex that no conventional supercomputer can check the math, how do we know the answer is actually right? This "verification gap" has haunted the industry, threatening to relegate quantum processors to the realm of expensive, untrustworthy novelties.

That paradox may have just found a structural solution. Researchers from IBM and the University of Chicago have demonstrated a computation using 70 "logical" qubits that they say firmly establishes a new era of "trusted quantum advantage."

The joint experiment, published as a preprint and announced in late July 2026, executed a highly complex quantum circuit in approximately 15 minutes. According to the research team, the same calculation would require prohibitive, practically impossible runtimes on the world's leading classical supercomputers.[1][2]

But the phrase "quantum advantage" is heavily loaded marketing terminology, often deployed prematurely. Previous claims of "quantum supremacy"—most notably by Google in 2019—were eventually undercut when classical algorithms were optimized to solve the exact same benchmark problems in a fraction of the time.[3]

To understand why this announcement is structurally different, it helps to look at how the industry previously measured performance. Historically, researchers used a benchmark called Random Circuit Sampling (RCS). RCS asks a quantum processor to spit out highly complex, random statistical patterns.[2]

The issue with RCS is that it scales terribly for verification. At 30 or 40 qubits, a classical computer can simulate the quantum circuit and verify the output. But push past 50 qubits, and the classical simulation chokes. Researchers are left blindly trusting the quantum hardware's output without any mathematical proof that the machine did not just succumb to thermal noise and spit out garbage.[1][3]

The IBM and UChicago team abandoned RCS for a new protocol they call "doped Clifford sampling." Instead of random noise, they built a highly structured quantum circuit using "Clifford gates"—operations that are remarkably easy for classical computers to simulate and verify.[1]

Once this verifiable, predictable "backbone" was established, the researchers strategically injected—or "doped"—the circuit with 468 non-Clifford "T gates." These specific gates are the computational heavy lifters that push the problem beyond the reach of classical simulation.[1][3]

To keep the fragile quantum states from collapsing under the weight of their own errors, the team embedded the entire operation within a "spacetime code." This required using 27 additional "ancilla" helper qubits to constantly monitor the system for errors across both physical space and computational time.[1]

The result was a computation operating across 97 physical qubits that functioned as 70 error-detected "logical" qubits. By constantly checking the predictable Clifford backbone, the system could mathematically certify a lower bound for its own accuracy—even though the final output was too complex for a classical machine to verify directly.[1][2]

This encoded structure allowed the logical computation to achieve effective error rates that were 10 times lower than the raw physical error rates of the IBM Heron processor itself.[2]

It is crucial to distinguish between what was shipped and what is often hyped. The 70 logical qubits used in this experiment are "error-detected" data qubits, not the fully "fault-tolerant" logical qubits that will eventually be required for universal, error-free quantum computing.[3]

The UChicago collaboration was not an isolated event. IBM coordinated the release with two other ecosystem partners, Qedma and Algorithmiq, who published parallel demonstrations of quantum hardware outperforming classical simulation in modeling complex materials and magnetic properties.[3]

Together, these demonstrations signal a pivot in quantum research. The focus is shifting away from raw physical qubit counts and theoretical supremacy, and toward "utility-scale" operations where the hardware can actually be trusted to do useful scientific work.[2][3]

If the spacetime coding and doped Clifford sampling techniques hold up to peer review and classical algorithmic counter-attacks, they provide a blueprint for the next decade of quantum development.[1]

The era of blindly trusting quantum black boxes may be ending. As the hardware scales, the math required to keep it honest is finally scaling with it, opening the door for verifiable breakthroughs in materials science and chemistry.