Quantum computers have a housing problem. The most powerful machines on Earth, IBM’s and Google’s, live inside dilution refrigerators that cool their processors to roughly 10 to 20 millikelvin, colder than deep space, in facilities that cost millions to build and run. A German startup is selling a different vision: a quantum computer that fits in a standard server rack, plugs into a wall outlet, and runs at room temperature, built from defects in diamond.
Saxon Q, a spin-out of Leipzig University, announced on July 21 that it is taking orders for two systems, the SXQ128 and SXQ512, which it calls the first diamond-based nitrogen-vacancy (NV) quantum computers to exceed 10 qubits. The company, founded in 2021 by physicists Marius Grundmann and Jan-Berend Meijer, says the 128-qubit machine can ship within three months and the 512-qubit version from the second quarter of 2027. Prices are undisclosed. Independent analysts have pointed at the fine print in the claims.
How diamond qubits work
The qubit in an NV-diamond system is a defect: a nitrogen atom replacing a carbon atom in the diamond lattice, next to a vacancy. The spin of the trapped electron, plus nearby nuclear spins, forms the qubit, initialized and read out with light and manipulated with microwaves. Because diamond’s stiffness leaves too little thermal energy at room temperature to generate the lattice vibrations that normally destroy quantum coherence, NV qubits maintain millisecond-scale coherence in a normal office environment, without cryogenics, vacuum, or cleanroom.
That is the platform’s real advantage: deployment. A quantum computer that can live in a server room on premises enables edge use cases, including robotics, automotive, and defense, without cloud latency, at a fraction of the size, weight, and power of a superconducting machine. The company claims six to ten times the energy efficiency of GPU clusters, a figure whose methodology is undisclosed.
The qubit count, examined
The headline numbers deserve scrutiny. The SXQ128’s 128 qubits are arranged as 16 cores of 8 fully entangled qubits each; the SXQ512 as 32 cores of 16. The company confirmed on the record that only qubits within a given core can be entangled; the cores run in parallel, coordinated classically, not quantum-connected. So 128 is an aggregate count of physical qubits, not an entangled register width, and one analyst described the comparison to Google’s 105-qubit Willow as a category error.
The claim of being the first to exceed 10 qubits also depends on what is being counted. Delft researchers demonstrated a 10-qubit diamond register in a research setting in 2022, at cryogenic temperature; Saxon Q’s claim is defensible as a commercial-first, not a literal ever-first. Fidelity figures (up to 99.92 percent single-gate) come from previous-generation four-qubit hardware, not the new systems, and no SXQ128 or SXQ512 unit had been delivered or independently benchmarked as of late July.
The company’s verified track record is real but modest: a four-qubit system accepted by the German Aerospace Center (DLR) after passing gate-fidelity thresholds above 95 percent single-qubit and 90 percent two-qubit, and a four-qubit system running continuously at Fraunhofer IWU in Dresden since June 2025. An eight-qubit single-core system with two coupled NV centers was due for delivery to a paying client before the end of the third quarter of 2026, and a demo SXQ128 was still being set up.
What the skeptics say
Independent analysis has been pointed. The fabrication advance (a sulfur co-implantation process claimed to convert more than 85 percent of implanted nitrogen atoms into functional NV centers, with three-to-ten-nanometer placement precision) is widely credited; Meijer’s group published a 75.3 percent conversion yield using sulfur-assisted charge engineering in 2019. But analysts note the fidelity methodology is undisclosed, two-qubit fidelity on prior hardware (97 percent) trails superconducting rivals like Willow’s roughly 99.7 percent, and no demonstrated workload has beaten a classical computer.
The realistic near-term value, several analyses conclude, is on-premises deployment of small entangled registers for applications where room-temperature operation matters more than raw power, with the caveat that circuit cutting between cores carries significant classical overhead. Error correction and 10,000-plus-qubit machines are on the company’s roadmap after 2030, not on the order form. For now, Saxon Q is selling something real: a working four-qubit-class machine that runs off a wall socket, wrapped in a 128-qubit spec sheet that the field will spend the next year testing.
Sources
- Saxon Q press release and newsroom, July 21, 2026
- HPCwire, July 21, 2026; The Quantum Insider, July 21, 2026; Quantum Computing Report, July 21, 2026
- Supercomputing News fact-check, July 27, 2026
- PostQuantum analysis, July 22, 2026; XenoSpectrum analysis, July 22, 2026
- Abobeih et al., Nature 606:884-889 (2022), DOI: 10.1038/s41586-022-04819-6
- Live Science, August 6, 2026