Qiskit Fall Fest 2026: MalaysiaSix layers of progress

IBM released Qiskit Fermions, an open-source toolbox for expressing fermionic operators and circuits before converting them into ordinary qubit circuits. That timing matters: electrons and other fermions obey symmetries that can be lost when a problem is mapped to qubits too early. By postponing the mapping until transpilation, researchers can swap encodings, exploit locality and connect the workflow to Qiskit tools such as ffsim and sample-based quantum diagonalization.

In IBM’s one-dimensional Fermi–Hubbard example, a flow-set encoding plus one ancilla kept two-qubit depth at 12 from four to 100 sites; a straightforward Jordan–Wigner construction reached 407 at 100 sites. This is a compilation comparison, not an experimental speed-up or quantum-advantage claim. Its value is architectural: it gives chemistry and materials researchers a reusable way to test better mappings without hand-building every circuit.

Japan’s Institute for Molecular Science announced that Shunkai, a neutral-atom quantum computer led by Professor Kenji Ohmori’s team, is operational. Hitachi contributed the software stack and Infleqtion the quantum-processing stack. The system traps neutral rubidium atoms with optical tweezers, drives operations with microwave and laser fields, and reads individual atoms through fluorescence imaging.

“Full stack” here means the layers from user input to hardware control and output are integrated; it does not mean fault-tolerant operation. Shunkai starts at about 50 physical qubits, with an expansion to roughly 500 and limited external access planned. A 10,000-physical-qubit, error-corrected machine by March 2031 remains a roadmap goal. Neutral atoms avoid the dilution refrigerators used by superconducting systems, although the atoms still require laser cooling, precision optics and ultrahigh vacuum. No application benchmark or gate-fidelity data accompanied the announcement.

Researchers affiliated with A*STAR, the National University of Singapore and Nanyang Technological University reported a hybrid method for one combinatorial step in molecular docking. They represented candidate protein–ligand contacts as a maximum vertex-weighted clique problem, then used “full-basis encoding” to place up to three binary decision variables on each qubit through the three Bloch-sphere axes.

An 18-variable streptavidin–biotin instance used six qubits, while a 14-variable trypsin–benzamidine instance used five, on IBM’s 156-qubit Heron r2 processor ibm_kingston. The measured solutions recovered the same contact sets as classical solvers. This is useful evidence that richer encodings can reduce scarce qubit requirements, but it is a preprint and the examples were classically tractable. The work neither accelerates an end-to-end drug-discovery pipeline nor demonstrates quantum advantage.

A collaboration involving UNSW Sydney, imec, Diraq and KU Leuven reported the strongest hole-spin qubit performance yet in natural silicon. In a peer-reviewed Nature Communications paper, the team demonstrated single-qubit gate fidelities up to 99.8% and a two-qubit gate quality factor of 240, corresponding to an inferred physical-fidelity limit of 99.7%.

Hole spins are attractive because spin–orbit coupling permits compact, all-electrical control, and silicon fabrication can draw on mature semiconductor processes. They have historically lagged electron-spin qubits because disorder and more complicated spin physics make them harder to control. The experiment used an industrial foundry platform closely related to one already used for electron-spin devices, strengthening the manufacturing case. It is still a small-device result: isotopic purification, lower variability, larger arrays, control wiring and error-correction overhead remain unresolved scaling requirements.

IonQ researchers Min Ye, Andrii Maksymov and Nicolas Delfosse published an end-to-end decoder for a proposed fault-tolerant trapped-ion architecture. Error-correction decoding is a classical task: it must turn a continuous stream of syndrome measurements into corrections quickly enough that the quantum computation does not stall.

Using 12 cores of one Apple M4 Max CPU, the team benchmarked compiled and simulated workloads reaching 408 logical qubits and one million T gates. With assumed 1–5 millisecond syndrome-extraction cycles, the reported timing overhead stayed below 0.3% at a two-qubit error rate of 10⁻⁴ and below 12% at 5×10⁻⁴. The preprint addresses a credible classical-control bottleneck, but it did not decode data from an operating fault-tolerant processor; the conclusions depend on the architecture, error model and hardware timings assumed.

Researchers from the University of Padua, ThinkQuantum and Italy’s Institute of Photonics and Nanotechnology demonstrated quantum key distribution over an 18-kilometre free-space field link at telecommunications wavelengths. Their receiver used a 40-centimetre-class telescope and real-time adaptive optics to correct turbulence before coupling the light into single-mode fibre.

The peer-reviewed experiment generated secure key material at 200 bits per second with a compact analyser and room-temperature detectors. The “intermodal” element is important: practical networks will need free-space channels—for example, between ground stations and satellites—to interoperate with fibre infrastructure. Adaptive optics tackles a real obstacle because atmospheric distortion otherwise makes efficient fibre coupling difficult. This remains a point-to-point field trial, not a deployed network, and its rate, weather tolerance and operating availability would need to improve for many applications.