Fujitsu's new quantum chip is built to scale by linking modules with light

Fujitsu, with QuTech (TU Delft) and the University of Tokyo, has fabricated a diamond-spin quantum processing unit prototype that bets against the field's default scaling approach: many small diamond modules linked by light, not one growing monolithic chip (Quantum Computing Report).

Tin-vacancy (SnV) color centers emit roughly 10× more photons than the more common nitrogen-vacancy (NV) centers, the figure Fujitsu's own characterization puts forward. That brightness is what lets the qubits talk directly to photonic integrated circuits (PICs) and skip the microwave-to-optical translator that other modular quantum designs still need. A translator converts a microwave signal from one qubit into a single photon a neighboring chip can detect, then back again, and each round trip costs fidelity. The Fujitsu team's bet is that a tin vacancy in a diamond lattice, with its built-in inversion symmetry, will deliver enough photons on its own to remove that step entirely (Fujitsu press release).

The same inversion symmetry makes SnV centers less sensitive to local electric-field noise than NV centers, because it cancels a class of stray-charge fluctuations that normally limit NV coherence. That stability matters when the emitter sits inside a photonic circuit rather than in a relatively quiet free-space trap. The trade is that SnV centers have to be made deliberately. The build flow is ion implantation of tin into a diamond, surface activation, bonding to an alumina-on-silica (Al2O3/SiO2) carrier, then thinning the diamond from hundreds of micrometers down to a sub-micrometer layer so the optical mode in the alumina waveguide can reach the emitter. Alumina is the waveguide material of choice because it holds low optical loss across the visible wavelengths where SnV's zero-phonon line sits (Fujitsu research page).

The module is designed to repeat. Fujitsu's blueprint routes photons between modules through 50:50 beam splitters and single-photon detectors, treating each diamond chip as a node in a small optical network. Coupled carbon-13 (13C) nuclear spins inside the diamond act as local quantum memory. Fujitsu's published figures put the nuclear-spin coherence time above one minute and the electron-spin coherence time above one second; those are the project's own characterizations, not yet independently verified. Underneath, an FPGA-based (field-programmable gate array) software-defined control stack uses a global controller to orchestrate multi-module sequences and a local processor per module to generate the RF, microwave, and laser pulses that drive each diamond (Quantum Computing Report).

The cryostat holds the prototype at 1.55 Kelvin (-271.6°C), still colder than the cosmic microwave background at about 2.7 K but two orders of magnitude warmer than the ~15 millikelvin regime superconducting quantum processors require. That is not "warm" in any practical sense, but it is a meaningful relaxation in cooling load. Dilution refrigerators that reach 15 millikelvin are custom, slow to recycle, and limit how many qubits a single machine can host. 1.55 K sits well within reach of standard helium-4 cryostats. The 1.55 K figure is the cryostat's operating point; the qubit's effective temperature under operation has not been published (Fujitsu press release).

What has not been demonstrated is the multi-module state transfer the architecture is built around. The beam splitters and single-photon detectors are part of the design, and the optical networking is the explicit scaling story, but the prototype is currently a single diamond chip bonded to a photonic integrated circuit, exposed for remote access through the Fujitsu Hybrid Quantum Computing Platform. The "modular" in "modular architecture" is the bet, not a proven multi-module result. On the available evidence, Fujitsu has built the first node of an optical network and is letting outside researchers try the routing logic in software before the next module is fabricated.

Recent quantum hardware announcements, including IonQ's revised revenue guide, D-Wave's CHIPS Act award, and SkyWater's foundry partnership with Qolab, have trained the press to read any new prototype as a generic "quantum is advancing" datapoint. The Fujitsu prototype is more specific. It is one of the first public hardware bets that the path past quantum's scaling problem runs through many small diamond modules emitting into optical fiber, rather than one larger chip that grows in place.