A new 2,682 photon quantum machine, built with the University of Science and Technology of China team behind a light based quantum prototype, runs at room temperature and shares a Chinese telecom's public cloud with supercold quantum chips, though both still target narrow…
China Telecom has put a second kind of quantum computer on its public cloud, one that computes with single particles of light through optical circuits at room temperature rather than at temperatures colder than deep space. The new Tianyan-P2000 photonic machine runs Gaussian boson sampling on 2,682 photons and shares its cloud fabric with the operator's existing cryogenic superconducting cluster. Chinese state media, echoing the announcement, frames the deployment as the world's first multi-tenant cloud service to expose both photonic and superconducting modalities to paying external users.
"Photonic" means the computer manipulates individual photons, single particles of light, sending them through a network of beam splitters and mirrors, rather than encoding information in superconducting circuits that have to be chilled to within a few thousandths of a degree above absolute zero. The distinction matters for two reasons. Photonic machines can run at room temperature and route over ordinary telecom optical fiber, which is why a telecom operator can plausibly host one. Superconducting chips, by contrast, need dilution refrigerators that look like chandeliers and operate at around 10 to 20 millikelvin, colder than the background temperature of deep space.
The new node is not a from-scratch machine. It replicates the optical architecture of Jiuzhang 4.0, the University of Science and Technology of China prototype whose performance was published in Nature in May. Jiuzhang 4.0 ran Gaussian boson sampling, a specialized complexity-theoretic benchmark, on 1,024 squeezed-state inputs across 8,176 optical modes. The Tianyan-P2000 reportedly scales that to 2,682 photons, a configuration built jointly with Jiuzhang (Jinan) Quantum Technology Co. and the Chinese Academy of Sciences Center for Excellence in Quantum Information and Quantum Physics.
The deployment narrative, which the announcement calls the world's first multi-tenant dual-modality quantum advantage service, rests on the cloud architecture as much as on the hardware. China Telecom already hosts the Tianyan-504 cluster of superconducting QPUs on the same Tianyan quantum cloud, reportedly brought online in 2024. Adding the photonic node to the same public platform means a remote customer can, in principle, submit a boson-sampling job to the photonic chip and a superconducting workload to the cryogenic chip through one login, then compare the results.
That multi-tenant framing is the real news, more than the photon count. Up to now, photonic quantum advantage experiments, including USTC's Jiuzhang line, Xanadu's Borealis, and earlier Chinese superconducting demonstrations, have lived in physics laboratories. They have been single-tenant hardware, used by one research group, published in one paper, built from one carefully tuned apparatus. Putting a photonic machine on a telecom's public quantum cloud is a deployment claim, not a physics claim. It says the operator believes it can keep an interferometer calibrated well enough to bill outside users for time on it.
The caveats are real and easy to miss. "Quantum computational advantage," in this context, refers to a complexity-theoretic argument that the photonic machine can sample from a probability distribution that classical computers would struggle to reproduce. It is not a claim that the machine can run Shor's algorithm, simulate molecules faster than a classical supercomputer, or break cryptography. Boson sampling has no known commercial application; the result is a set of photon-detection click patterns that look random to a classical sampler. China Telecom's announcement and subsequent state-media coverage treat "dual-modality quantum advantage" as a feature of the cloud, but the underlying benchmark is the same one USTC published in Nature. The vendor's "world's first" framing should be read as "first commercial cloud to rent out both modalities," not as a broader quantum-supremacy milestone.
The independent Western expert commentary that would normally sanity-check such an announcement is largely absent from the source mix, which leans on China Telecom Quantum Group, China Daily, and CAS-affiliated outlets. That is worth flagging not as a refutation, but as the usual single-vendor disclosure problem that comes with vendor-led quantum milestones. Comparisons to photonic platforms outside China, including Xanadu's Aurora and Borealis, PsiQuantum's utility-scale push, and QuiX's European deployments, would sharpen the picture but are outside the present reference basis.
The interesting question is not whether the Tianyan-P2000 beats a classical supercomputer on boson sampling. The University of Science and Technology of China already showed that, in a peer-reviewed paper. The question is whether a telecom operator can keep a 2,682-photon interferometer stable enough to rent out as a cloud service, and whether customers actually log in to run jobs on it. The Tianyan cloud now provides the place to find out.