The headline number: 99%, but for a specific task
IBM’s new 99% figure refers to average quantum-state transfer fidelity across four long-range interconnects linking two IBM Quantum Nighthawk R1 processors, not to universal “99% accuracy” for every quantum computation a future machine might run . That distinction matters because quantum reliability is a stack: moving information, applying gates, preserving entanglement, reading out states, correcting errors and coordinating software all have to work together before a system becomes broadly useful.
In the experiment IBM described on October 27, two 120-qubit superconducting processors were connected inside a dilution refrigerator using four so-called L-couplers, each built around an 80-centimeter superconducting aluminum cable . IBM says the design is intended for modular systems, where many processors act less like isolated chips and more like linked tiles inside a larger quantum computer . TipRanks summarized the result as a “major quantum computing breakthrough,” noting that IBM transferred information between the two processors with 99% accuracy on average and reported a best connection of 99.3% .
The achievement is important because scaling quantum computers by simply making a single chip larger runs into physical, fabrication and control limits. IBM’s bet is that a future fault-tolerant system will need multiple chips connected by high-quality quantum channels, much as classical high-performance computing depends on interconnects as well as processors . Network World described the work as a milestone for scaling, reporting that the links enabled multiple quantum channels between large superconducting processors and that the setup also transmitted entanglement between chips .
Why interconnects are a hard problem
Quantum information is fragile because a qubit’s useful state can be disturbed by heat, noise, unwanted coupling, imperfect gates or measurement itself. Linking two chips is therefore not the same as connecting two classical circuit boards. The interconnect must move a quantum state coherently, without turning it into ordinary classical information, and it must do so while operating at extremely low temperatures.
IBM’s L-couplers are designed to connect processor modules inside modular cryogenic systems, using long-range couplers rather than only short on-chip couplers . The company says this architecture can support larger error-correction codes and larger algorithms across more entangled qubits . In practice, that means the 99% transfer result is not just a speed or wiring statistic; it is a test of whether quantum information can cross a physical distance large enough to make modular machines plausible.
ScienceAlert emphasized the same point from a systems perspective: the couplers are meant to be installed inside cryogenic units, with IBM’s longer-term goal being a fault-tolerant quantum computer by 2029 . The report also noted that the result remains an intermediate step, because the researchers observed issues such as magnetic interference and because further efficiency improvements are still required before the links are ready for full-scale use .
The fine print: transfer fidelity is not gate fidelity
The most important caveat is that IBM’s 99% number applies to state transfer across the link. When the system performed two-qubit CZ gates between qubits located on separate modules, the average fidelity was lower: IBM reported 94.75% across the four links . That number is still technically notable, but it is much closer to one error in twenty operations than one error in one hundred.
Post-Hype Tech highlighted this gap directly, arguing that the 99% figure is real but narrower than a casual headline might imply . Its audit noted that the preprint shows quantum states moving across the links at about 99% fidelity, while cross-link two-qubit operations averaged around 95%, and that the demonstration did not yet run error correction or a full computation across the link .
That does not erase the milestone. It frames it. Quantum computing progress often arrives as a sequence of constrained reliability improvements rather than a single commercial switch being flipped. A good interconnect does not automatically create a dependable quantum computer, but a dependable modular quantum computer cannot exist without good interconnects.
Where this fits in IBM’s roadmap
IBM says the connection design is compatible with Nighthawk and can scale toward IBM Quantum Starling and beyond . TipRanks reported that IBM links the work to its plan for Quantum Starling, a 2029 system target associated with 200 logical qubits and 100 million quantum operations . These are roadmap ambitions, not evidence that commercial quantum advantage is already routine.
The company’s public framing is also careful in one useful way: IBM says it has “room to improve all of the metrics” reported and identifies future work on yield, reproducibility, error rates and scaling to more couplers . In other words, the result is a building block. It points toward a possible architecture for fault tolerance, but it does not by itself prove that large enterprise workloads can now be run reliably.
For business readers, this is the line to hold. The announcement strengthens IBM’s case that modular superconducting machines are viable, but it does not mean quantum computers are ready to replace classical systems in chemistry, finance, logistics or materials science at scale. The immediate value is technical de-risking: IBM has shown that long-range chip-to-chip quantum transfer can work with high fidelity in a configuration designed for larger machines.
The broader context: reliability and manufacturability are converging themes
The same 72-hour news cycle also brought a separate but related quantum hardware story from German startup NVision, which announced a peer-reviewed Nature paper on a manufacturable molecular qubit and a partnership with HyperLight to integrate thin-film molecular qubits onto photonic integrated circuits . Reuters, via MarketScreener, reported that NVision’s approach borrows from manufacturing techniques used in smartphone displays and emerged from the company’s work on medical-device technology .
That NVision development is not IBM’s result, and it should not be folded into the 99% claim. But it helps explain why IBM’s milestone matters: the quantum race is shifting from isolated demonstrations toward systems engineering, manufacturability and repeatable reliability. Whether the path is superconducting modular links, molecular qubits on photonic chips or another architecture, the core question is becoming less “can one device behave quantum mechanically?” and more “can many devices be manufactured, connected and controlled well enough to finish useful workloads?”
What to watch next
The next tests are straightforward to describe and difficult to pass. IBM will need to improve cross-module gate fidelity, demonstrate error correction across linked processors, increase the number of couplers, and show that the same performance can be reproduced across more complex systems . Independent peer review and third-party replication would also make the 99% claim more durable as an industry benchmark.
For now, the result deserves neither dismissal nor hype. It is not a general-purpose 99% accurate quantum computer. It is a credible advance in one of the engineering layers that a scalable quantum computer must master. If quantum computing eventually becomes useful for science and industry, milestones like this are the kind of plumbing progress that will make the breakthrough look inevitable in hindsight.
Comments
Be the first to comment.