From 100-Qubit Chips to Full Systems, Xianggan Pushes Quantum Compute as a Service

Xianggan Technology, a Chinese quantum computing company spun out of the Beijing Academy of Quantum Information Sciences, says its superconducting chips have reached the 100-qubit scale, with chip lifetimes above 100 microseconds, average single-qubit gate fidelity above 99.9% and two-qubit gate fidelity above 99.5%. Its goal for this year is to deliver a complete 100-qubit system and, in parallel, build a 1,000-qubit quantum cloud platform that turns quantum processing into a callable service.

The pitch rests on a familiar wall. Molecular simulation and large-scale combinatorial optimization scale exponentially with problem size: a binary problem with 100 variables has roughly two to the power of 100 possible combinations, far beyond what classical machines can enumerate, no matter how many GPUs are stacked.

Qubits, which can sit in superpositions of 0 and 1 and entangle with one another, expand the computational space with every added qubit. That is why quantum processing units are being described as a third kind of compute after CPUs and GPUs, and why Xianggan's real task is engineering rather than physics.

On the chip side, the company uses a superconducting route built around tantalum circuits, sapphire and high-resistivity silicon substrates, and ultra-high-vacuum magnetron sputtering to grow low-defect tantalum films. It says the full fabrication flow no longer depends on a single platform, and it is building its own line so that design, process iteration, tape-out and packaging can happen in-house.

Above the chip sits the machine. Xianggan's current systems support independent control and readout of 10 to 1,000 qubits, integrating chip, control electronics and software, and can be deployed on-premises or accessed through the cloud. Its control electronics have passed system-level validation on processors above 100 qubits, with single-qubit gate fidelity above 99.99%, two-qubit fidelity above 99.8% and readout fidelity above 99%.

The business started with control electronics. Revenue from that line grew from just over 3 million yuan in 2024 to 31 million yuan in 2025, with customers including institutes under the Chinese Academy of Sciences, Tsinghua University and the Greater Bay Area Quantum Science Center. Those orders funded the system work.

The planned cloud platform is meant to lower the entry barrier. Users will be able to drag and drop quantum circuits in a graphical interface or submit jobs through an API, and the company plans interfaces for AI agents plus natural-language task submission. CEO Jin Yirong compares today's quantum programming to writing assembly by hand.

Applications under exploration include finance, where the company lists 15 use cases from risk control and anti-money-laundering to portfolio optimization and derivatives pricing, plus hybrid quantum-classical work with supercomputers and weather modelling. Xianggan has also been co-building the Quafu quantum cloud platform with the Beijing academy and sponsoring a real-hardware challenge for algorithm developers.

The honest caveat is noise and engineering maturity. Jin argues the practical near-term move is to use AI for calibration, parameter optimization and fault diagnosis so that quantum hardware improves faster, and only then to feed quantum capacity back into AI. Turning quantum compute into something as callable as cloud capacity is less a physics problem than a simultaneous engineering problem across chip, control, system and platform. The 100-qubit delivery deadline and the 1,000-qubit platform are the two milestones to watch.