The breakthrough: a 99% link between quantum chips
IBM has reported a major step toward modular quantum computing: two IBM Quantum Nighthawk processors were connected using four meter-scale l-couplers, achieving an average quantum state transfer fidelity of 99% . In plain language, IBM is not saying that every quantum computation now runs with 99% accuracy. It is saying that fragile quantum information could be moved from one superconducting processor to another over the new inter-chip links with 99% average fidelity .
That distinction matters. The story is still important because quantum computers are difficult to scale as single, monolithic chips. If useful machines eventually need hundreds, thousands or more physical qubits, one promising route is to connect multiple processors into a larger modular system. IBM’s latest test addresses one of the hard engineering questions behind that route: whether quantum states can travel between chips without being destroyed by noise, loss or thermal instability.
The company says this is the first time superconducting quantum processors have been connected using multiple meter-scale quantum channels with an average quantum state transfer fidelity of 99% . ScienceAlert’s coverage similarly describes the experiment as linking two 120-qubit chips with four superconducting aluminum L-couplers, reaching 99% average transfer accuracy and a peak of 99.29% . Network World also reports that the two 120-qubit chips were connected through four connectors, each using a superconducting cable about 31.5 inches, or 80 centimeters, long .
What IBM actually demonstrated
The experimental setup used two full IBM Quantum Nighthawk R1 processors and four l-couplers inside a dilution refrigerator . Each l-coupler included an 80-centimeter superconducting cable plus the supporting on-chip and control components needed to transmit a quantum state between modules . IBM says the processor chips were full Nighthawk chips, even though the experiment wired out only a subset of qubits for development purposes .
The headline metric is state transfer fidelity. IBM reports transfer fidelities up to 99.3%, with a 99% average across the four links . ScienceAlert gives the same framing: the researchers linked the two 120-qubit chips and achieved an average transfer accuracy of 99%, peaking at 99.29% . ForkLog’s account also emphasizes that the best coupler reached 99.29% and that the result demonstrates a route for integrating individual superconducting chips into larger computational systems .
The team also tested more than simple transmission. IBM says it used distant controlled-Z, or CZ, gates built by transferring states across modules and achieved an average two-qubit gate fidelity of 94.75% over the four links . ForkLog highlights the same number and adds that the maximum inter-processor CZ operation reached 96.42% . That difference between 99% transfer fidelity and roughly 95% cross-module gate fidelity is central to understanding the result.
In other words, the link is better at moving a quantum state than at performing the full inter-processor two-qubit operations that complex algorithms would require. Post-Hype Tech makes that limitation explicit, calling the result promising while noting that operations across the link are closer to 95% and that IBM has not yet shown error correction or a full computation running across the two chips .
Why modular quantum computing matters
The breakthrough fits IBM’s broader strategy of scaling quantum computers through modularity. Instead of putting every qubit on one increasingly complicated chip, modular designs connect multiple chips, ideally allowing a system to grow while keeping individual processor units manageable. IBM says high-fidelity links between processing units are required for larger algorithms and more powerful error-correction codes across higher numbers of entangled qubits .
This is not just a matter of adding more qubits to a spec sheet. Quantum information is fragile. Qubits must be controlled, cooled, measured and protected from environmental noise. As chips grow, wiring, control electronics, calibration and thermal management become harder. A modular approach can reduce some of those pressures, but only if the connections between modules are reliable enough to carry quantum information and support entangling operations.
IBM’s l-couplers are designed for that role. The company says the new inter-chip connection is compatible with IBM Quantum Nighthawk and can scale toward IBM Quantum Starling and beyond . Network World reports that the approach is compatible with Nighthawk and with IBM’s planned fault-tolerant Starling computer, which IBM has targeted for 2029 . ScienceAlert likewise frames the result as part of IBM’s effort to link quantum chips inside cryogenic systems on the path to a fault-tolerant system by 2029 .
The relevance is practical. If modular links become dependable and repeatable, quantum engineers can think less like chip designers trying to fit everything onto a single die and more like system architects assembling processors, couplers, cryogenic infrastructure and error-correction layers into a machine. That is why a successful 80-centimeter quantum connection is noteworthy even though it sounds short by everyday networking standards.
What the 99% number does not mean
The clearest risk in this story is overstatement. A casual reading of “IBM achieves 99% accuracy” could suggest that IBM has solved quantum computing errors generally. It has not. The 99% figure applies to quantum state transfer over the new l-coupler links . It does not mean that arbitrary algorithms run correctly 99% of the time across a 240-qubit machine.
The cross-module gate result is lower: IBM reports an average two-qubit gate fidelity of 94.75% over the four links . ForkLog notes the same average and a maximum of 96.42% for those inter-processor operations . For many useful quantum algorithms, errors can accumulate quickly, so improving gate fidelity remains essential.
There is also the matter of peer review. ScienceAlert notes that the research had not yet been peer-reviewed and was available as a preprint . Post-Hype Tech reaches a similar conclusion, saying the 99% figure is real for moving a qubit’s state but narrower than it sounds, while the inter-chip operations needed for computation are nearer 95% .
IBM itself points to further work. The company says it plans to improve l-coupler yield and reproducibility, reduce error rates, and scale to more couplers that can be connected and disconnected between larger and more complex processors . ScienceAlert also reports that researchers observed magnetic interference affecting signal quality and that further efficiency improvements are needed before the couplers are ready for practical use .
Entanglement across chips: an important supporting result
The transfer number is not the only important measurement. IBM says the distant CZ gate allowed it to generate entangled states shared between the two modules, including the use of two l-couplers at the same time to create independent entangled states . Network World reports that the cables were able to transmit quantum entanglement between the chips and stayed operational through three warm-up and cool-down cycles .
ForkLog adds that the researchers demonstrated distributed entangled states, including four- and six-qubit GHZ states spanning the processors . That is important because scalable quantum computing is not just about moving data. It requires creating and manipulating correlations between qubits in ways that classical systems cannot efficiently reproduce.
However, entanglement by itself is not a finished computer. Post-Hype Tech notes that the work shows entangled states shared across two chips but does not yet show error correction or a computation running across the link . This is best understood as a hardware milestone: a crucial component has been demonstrated, not the whole architecture.
The market and enterprise angle
The story also lands at a time when quantum computing is moving from pure research toward commercial roadmaps, cloud access and enterprise pilots. Network World reports that IBM CEO Arvind Krishna told investors that three independent demonstrations achieved quantum advantage on IBM hardware during the summer and that IBM had signed 16 enterprise engagements so far this year, including with AT&T and General Motors . Those claims are separate from the l-coupler experiment, but they show why investors and enterprise technology leaders watch hardware milestones closely.
For businesses, the near-term impact is not that IBM’s quantum systems are suddenly ready to replace classical computers. The better reading is that the infrastructure required for future fault-tolerant quantum machines is becoming more concrete. If modular links improve, they could support larger error-corrected systems, and those systems are the ones expected to matter for demanding applications in materials, chemistry, optimization and cryptography.
Still, the timeline remains uncertain. Even IBM’s own blog presents the work as a step on the roadmap, not a commercial endpoint . The company says it has room to improve all the metrics reported in the paper and will continue research and development to integrate the couplers into larger systems running more advanced workloads .
Bottom line
IBM’s 99% achievement is a genuine quantum hardware breakthrough, but its meaning is specific. The company has shown high-fidelity transfer of quantum states between two Nighthawk processors using four 80-centimeter l-couplers, with an average state transfer fidelity of 99% and a best result around 99.3% . That is a strong result for modular quantum architecture.
The breakthrough is not the same as 99% accuracy for all quantum computations. Cross-module two-qubit gate fidelity averaged 94.75%, and the work has not yet demonstrated a complete error-corrected computation across the connected chips . The next milestones are clear: more couplers, better reproducibility, lower error rates, stronger inter-chip gates and integration with error correction.
If IBM can improve those metrics and scale the approach, this experiment may be remembered as one of the hardware steps that made modular fault-tolerant quantum computing plausible. For now, it is a meaningful advance, not a finished revolution.