IBM Claims a Quantum Edge Conventional Computers Cannot Match

News

1. Temasek, Seraphim Back Pixxel in $100 Million Round

2. UN Rights Chief Warns Advanced AI Could Threaten Humanity

3. Tata-owned JLR to Cut About 4,000 Bobs Amid Tariff, China Pressure

4. Anthropic Walks Away From $6 Billion Decart Deal

5. Hyundai India Targets 20% Women in Executive Workforce by 2030

6. Security Breach Drains $320 Million in Bitcoin From Liquid Network

IBM said on Thursday, 30 July, that one of its quantum computers achieved “quantum advantage” for the first time, completing a verifiable benchmark in about 15 minutes that conventional computers cannot practically reproduce.

Quantum advantage is reached when a quantum computer performs a task more efficiently, cheaply or accurately than any known classical alternative and produces a result that researchers can trust.

IBM based its claim on three papers released with the University of Chicago, Israeli quantum software company Qedma and Finnish startup Algorithmiq. The experiments took different routes to the same goal: finding calculations that strain classical computers while building checks into the quantum process.

The University of Chicago experiment offers the clearest test. Researchers used 97 physical qubits to run a calculation involving 70 logical qubits, which encode information in a way that allows errors to be detected and suppressed.

The circuit performed 2,415 logical two-qubit operations and 468 more complex operations known as T gates. Encoding the calculation and discarding runs that failed internal checks reduced the effective gate-error rate by about tenfold, according to the paper.

The result carried a fidelity lower bound of 28.4% with 95% confidence. Fidelity measures how closely the state produced by a quantum computer matches the ideal one.

“We are now firmly in the quantum advantage era,” IBM Research Director Jay Gambetta said. “We have demonstrated a quantum computation beyond the practical reach of classical computers that establishes, with statistical confidence, a lower bound on how faithfully it was executed.”

The experiment used a structured form of random circuit sampling, a benchmark in which a quantum computer generates patterns that become increasingly difficult for a conventional machine to simulate.

Previous sampling experiments ran into a basic problem. Once the calculation became too large for a classical computer to reproduce, the classical machine could no longer provide a reliable answer against which to check the quantum result.

IBM and the University of Chicago worked around that problem by starting with a circuit that could still be simulated conventionally. They then inserted T gates that made the calculation much harder while retaining checks capable of detecting errors as it ran.

That allowed the researchers to estimate the fidelity of the difficult calculation directly, instead of inferring it from smaller or simpler experiments.

The result is still a computing benchmark rather than a useful scientific or commercial calculation. Its value lies in showing that quantum hardware can cross a computational boundary without leaving researchers unable to judge whether it worked.

There are other qualifications. The three papers were posted as preprints this week and have not been peer reviewed. The University of Chicago paper also describes its certificate as device-dependent, meaning the result rests partly on assumptions about the machine that performed the calculation.

IBM’s two other demonstrations were closer to scientific applications, though their claims are less conclusive because no definitive classical answer is available.

Qedma, working with IBM, Japan’s RIKEN research institute and quantum software company BlueQubit, studied how a two-dimensional magnetic model responded to repeated pulses of energy.

Using Qedma’s error-mitigation software on an IBM Heron processor, the researchers followed the model’s behavior across circuits containing as many as 74 qubits. Leading tensor-network simulations failed to converge, while another classical method remained highly sensitive to the shortcuts used in the calculation, despite running on advanced graphics processors and Japan’s Fugaku supercomputer, according to the study.

Selected parts of the experiment were repeated on Quantinuum’s trapped-ion quantum computers, which produced similar results. The cross-check reduced the likelihood that the findings were caused by a quirk in IBM’s hardware or Qedma’s error-mitigation software.

The work could eventually help researchers study optoelectronics, superconductors and other advanced materials. Those uses remain prospective, however, and the experiment does not by itself establish a commercial advantage.

Algorithmiq took a different approach. Its researchers used a 56-qubit experiment on IBM hardware to simulate the movement of information through a material containing regions with different physical properties.

At least three groups using leading classical methods produced conflicting answers, while the quantum results remained consistent across machines and different levels of noise, according to the paper.

The problem has been listed on IBM’s Quantum Advantage Tracker for eight months. No classical method has yet reproduced reliable results across the full range studied, the companies said.

Algorithmiq is also releasing a classical simulation package called Monoprop as open-source software, giving researchers another tool with which to test future advantage claims.

The invitation to challenge the results matters because quantum advantage rarely provides a permanent finish line. Better algorithms can rapidly restore the classical lead.

A BlueQubit experiment previously listed on IBM’s tracker illustrates the risk. Its researchers initially estimated that a classical computer would need 3.2 million years to reproduce a task completed by a quantum machine in about two hours. Within months, new classical methods reduced the calculation to between an hour and a few seconds, beating the quantum system.

Google faced a similar challenge in 2019 after its 53-qubit Sycamore processor completed a sampling task in 200 seconds. Google estimated that a leading supercomputer would need 10,000 years, a finding published in Nature.

IBM countered that a better classical method could finish the calculation in about two-and-a-half days. Subsequent improvements have continued to narrow the cost of simulating some quantum experiments.

IBM has submitted its latest results and data to the Quantum Advantage Tracker, where outside researchers can propose stronger classical solutions. The tracker currently labels such experiments “active advantage candidates,” meaning quantum methods appear competitive but further benchmarking is needed.

The announcement comes as IBM increases its financial commitment to quantum computing. The company plans to invest more than $10 billion over five years in research, manufacturing, partnerships and acquisitions.

IBM last week agreed to acquire HRL Laboratories from Boeing and General Motors, adding expertise in silicon-spin qubits to its longstanding work on superconducting circuits. Financial terms were not disclosed.

The quantum push has taken on greater importance after IBM shares plunged more than 25% on 14 July, their largest one-day fall on record, following a profit warning that cited customers shifting spending from software to AI hardware and memory chips.

IBM’s next major target is Starling, a fault-tolerant quantum computer scheduled for 2029. The company says it will use 200 logical qubits to perform as many as 100 million quantum operations.