On September 8, 2026, Fujitsu announced a prototype quantum computer that uses tin-vacancy (SnV) centers in diamond. The company said it built a single-module system that integrates SnV centers into an on-chip photonic circuit capable of operating quantum gates. The device runs at about -271.6°C, with qubits controlled through Fujitsu's cloud platform. The significance of this achievement lies not in qubit count, but in the fact that the materials, circuitry, and control systems needed to link separate modules optically have been consolidated into a single unit.
At the same time, this device does not feature 1,000 logical qubits. Optical connections between multiple modules remain a task for the next development phase. Separating the current demonstration from Fujitsu's fiscal year 2035 target reveals the distance still to be covered before reaching a large-scale, fault-tolerant quantum computer.
A Single-Module Demonstration, and Three Unmet Milestones
Fujitsu's explanatory materials explicitly describe the diamond module integrated chip inside the cryostat as "one module." What was demonstrated was an operation that calibrated the frequencies of electron and nuclear spins and confirmed Rabi oscillations when microwaves were applied to the electron spin. The pathway that converts quantum circuits into control sequences of light, microwave, and RF waves, and operates qubits from the Fujitsu Hybrid Quantum Computing Platform, was also verified.
Breaking down the development stages by approach and timeline yields the following:
| Timeline | Target | Fujitsu's Stated Achievement/Goal | Status |
|---|---|---|---|
| September 2026 | SnV diamond spin | Single module integrated on-chip; quantum gate operation and cloud-based control confirmed | Demonstrated |
| 2027 | SnV diamond spin | Prototype of multiple modules connected via light | Development goal |
| FY2030 | Superconducting | 250 logical qubits with over 10,000 physical qubits | Development goal |
| FY2035 | Integration of superconducting and diamond spin under study | 1,000 logical qubits | Overall goal |
What operated this time was a single module. Optical connection of multiple modules is a 2027 development goal; 250 logical qubits is a fiscal year 2030 target; and 1,000 logical qubits is part of the overall roadmap goal for fiscal year 2035.
Physical qubits refer to the actual elements that manipulate the states of electrons or atomic nuclei. Logical qubits are units that distribute information across many physical qubits, enabling error detection and correction. For this prototype, Fujitsu has not disclosed the total number of physical qubits or gate fidelity. Coherence time and the results of any executed algorithm are also unknown. Therefore, the four stages in the table cannot be compared as a single continuous performance metric.
The phrase "world's first" also has specific scope. Fujitsu's footnote states that this is the world's first quantum computer of its kind to integrate SnV centers onto a chip, based on the company's own research as of September 2026. It does not mean this is the world's first diamond spin quantum computer overall, nor the world's first with logical qubits.
Why Tin Was Chosen: To Make Light Easier to Extract
The diamond spin approach uses color centers formed by impurities and vacancies in artificial diamond. Electron spins serve as the communication link that connects to the outside world via photons, while nearby carbon-13 nuclear spins act as the data-holding role that retains quantum states for longer periods. Rather than making a single chip indefinitely larger, Fujitsu's concept is to connect these small quantum modules using light.
The nitrogen-vacancy (NV) center, commonly used until now, consists of a nitrogen atom and an adjacent vacancy in diamond. In the SnV center adopted this time, a tin atom sits between two vacancies. Because of its inversion symmetry, it is less susceptible to charge noise and is considered a promising candidate for aligning light frequencies even in nanofabricated environments. Entangling distant qubits requires bringing photons arriving from both sides into an indistinguishable state, and this stability helps achieve that.
Fujitsu describes SnV's light emission efficiency as roughly 10 times that of NV. What increased is the efficiency of extracting photons used for qubit readout and optical connections. This is not a measurement showing that quantum computation speed, qubit count, or gate fidelity improved tenfold. Since comparison conditions and error margins are not detailed in the explanatory materials, the roughly 10-fold figure should be treated as the company's own materials evaluation.
Even guiding photons outside the chip presents material processing challenges. Fujitsu bonded high-quality diamond, into which tin ions had been implanted, onto a substrate made of alumina and oxide-coated silicon. Diamond that was originally several hundred micrometers thick was thinned to several hundred nanometers and coupled with a low-loss alumina waveguide for visible light. This design avoids the constraint of building an entire optical circuit from high-quality diamond alone, instead bonding together materials suited to their respective strengths.
This bonding technology has a precursor. In 2023, researchers from Fujitsu and QuTech reported directly bonding diamond and sapphire at room temperature, achieving a shear strength exceeding 14 MPa and high transmittance in the visible light range. This prototype advanced the process further—bonding and thinning materials, then integrating nanoscale diamond containing SnV centers with alumina waveguides into a single optical circuit.
The 1.53°C Difference Matters Less Than What Can Be Reduced Inside the Cryostat
The prototype operates at approximately -271.6°C, or about 1.55K in absolute temperature. This is 1.53°C higher than the approximately -273.13°C (about 0.02K) that Fujitsu cites as a typical value for superconducting approaches. While this may appear to be a small difference on Celsius or Fahrenheit scales, it represents a different temperature regime when measured from absolute zero. However, since the published figures are rounded, it would be inappropriate to treat the ratio in absolute temperature as a precise measure of performance difference.
1.55K is still far from what would colloquially be called "high temperature." Photos and diagrams of the prototype show a cryostat surrounded by lasers, single-photon detectors, and high-frequency equipment. This is neither room-temperature operation nor cryostat-free operation. How much cooling power, cryostat volume, and operating costs are reduced compared to superconducting approaches has not been disclosed.
The temperature margin could offer implementation advantages. This is because wiring and control circuitry that can be placed at the coldest stage of a cryostat face constraints on heat generation and space. In 2024, Fujitsu and QuTech placed a cryo-CMOS circuit operating at 4K within the same cryostat to drive NV qubits. This time as well, anticipating multi-module configurations, Fujitsu has presented a design that allocates processors for overall control and for local control of each module. The value of this temperature margin will be determined by how much wiring count and thermal load can be suppressed as the number of modules increases.
High Fidelity and Remote Gates: Separate Achievements Obtained with NV
Fujitsu cites two research results as evidence supporting the diamond spin approach. Both experimentally demonstrate components necessary for scaling up, but neither constitutes a performance test of this SnV prototype.
The first is a quantum gate reported in 2025 by QuTech and others in Physical Review Applied. The target was a two-qubit system consisting of an NV center's electron spin and nitrogen nuclear spin. Using gate set tomography, they achieved fidelities of up to 99.999(1)% for single-qubit gates and 99.93(5)% for two-qubit gates. While these are high figures, results measured on a two-qubit NV device cannot be directly transferred to the many qubits on an SnV chip.
The second is a 2026 Nature Communications paper, also using NV centers. The research team placed two qubits each in separate cryostats at 3.9K and 3.4K, connected by 2 meters of optical fiber. They created remote entanglement using electron spins and executed unconditional CNOT gates between carbon-13 nuclear spins. Significantly, rather than relying on post-selection of successful cases, they reflected measurement results in real time for subsequent operations.
This remote experiment concretized the principle of "connecting via light." However, what was connected were two NV nodes, whereas this SnV prototype represents a single module. The entanglement generation rate between SnV centers, remote gate fidelity, and connection wait times have not yet been shown. At this stage, the three achievements—high-fidelity local gates, light-based remote gates, and SnV integration—must be read as separate experiments.
The Yield of Optical Connections Will Be Tested in 2027
For the multi-module prototype planned for 2027, the evaluation point will not be how many modules were placed, but whether the optical connections function as a computational resource. Key figures to watch include the probability of extracting and detecting photons, the frequency and fidelity of entanglement generation between distant SnV centers, and how long nuclear spins can retain information while waiting for remote gates. The yield of manufacturing optical waveguides and diamond, as well as the rate at which the emission frequencies of multiple SnV centers can be aligned, will also affect mass producibility.
Furthermore, the performance of error correction combining local gates with optical connections will be a critical evaluation axis for determining whether this can be extended to logical qubits. Fujitsu's stated goal of 250 logical qubits for fiscal year 2030 is part of the superconducting approach's plan; the architectural composition and number of diamond spin modules for the 1,000-logical-qubit machine targeted for fiscal year 2035 have not been disclosed.
This prototype has opened a gateway by mounting SnV onto a photonic circuit and operating it through control software. The next step is not simply adding more cryostats to the photographs. If Fujitsu can simultaneously disclose the success rate, fidelity, and wait time when connecting multiple SnV modules via light, this single-module prototype will finally transform into measurable data for large-scale implementation.