SAXON Q’s 128-Qubit Diamond Quantum Computer: Real Fabrication Advance, Misleading Qubit Count

July 21, 2026 – On July 21, SAXON Q announced commercial availability of the SXQ128 and SXQ512, room-temperature diamond quantum computers with headline counts of 128 and 512 qubits. The systems fit in a standard server rack and run off a wall outlet. Within days, commentators were comparing SAXON Q’s qubit count to Google’s 105-qubit Willow processor.

I took a closer look at the numbers. The comparison to Willow is misleading, and several claims circulating on social media contain factual errors. But the announcement also contains a genuine fabrication advance that deserves accurate framing.

Sixteen Eight-Qubit Cores

The SXQ128 distributes its 128 physical qubits across 16 processing cores, with SAXON Q describing eight “fully entangled” qubits per core. The SXQ512, shipping from Q2 2027, will use 32 cores of 16 qubits each. XenoSpectrum’s analysis drew the critical distinction: the announced scale and the scale at which coherent computation actually occurs are two separate things.

What connects the cores? SAXON Q’s own QC2026 whitepaper answers this more candidly than the press release. It describes independently operated cores, parallel job execution, and circuit cutting where subcircuits run on individual cores and results are recombined classically. It also identifies the quantum-mechanical connection between cores as, in its words, “a key open issue,” and acknowledges that circuit cutting “introduces significant classical computation overhead and does not scale well for highly entangled problems.”

That is an honest engineering assessment, and I respect it. But it means the SXQ128’s disclosed native quantum width is eight qubits per core. The 128 figure describes aggregate system capacity across classically coordinated cores. If you need to run a quantum algorithm requiring more than eight entangled qubits, this architecture cannot do it on the published evidence.

The Willow Comparison

Google’s Willow is a 105-qubit superconducting processor in a single connected register. According to Google’s published benchmarks, its mean CZ gate error is 0.33% (roughly 99.67% fidelity), with iSWAP errors at 0.14% (99.86% fidelity). Willow demonstrated below-threshold surface-code error correction: increasing code distance from 3 to 5 to 7 reduced the logical error rate rather than amplifying it. Entangling-gate error is the constraint that determines how deep a useful circuit can run.

Comparing 128 aggregate qubits across sixteen classically coordinated 8-qubit cores to 105 qubits in a single connected register is a category error. The two numbers measure different things: Willow’s 105 describes the width of a coherent processor, while SXQ128’s 128 describes total capacity across independent cores.

What SAXON Q Has Published (and What It Hasn’t)

The whitepaper contains a specification table that the press release omitted. For the Gen3 QC2026 Dual Core (SAXON Q’s predecessor to the SXQ128), it lists single-qubit gate fidelity “up to 0.9992” and two-qubit gate fidelity of 0.97. That two-qubit number (97%) is the figure that matters for circuit depth, and it is absent from every press release and news article about the SXQ128. The 99.92% fidelity in the announcement appears to be the same single-qubit randomized-benchmarking result for one nuclear-spin qubit from the Gen3 system, presented without context about the operation type, register coverage, or the substantially lower two-qubit performance.

For comparison, DLR’s acceptance testing of SAXON Q’s earlier 4-qubit system in July 2024 used a lower-bound methodology: all four qubits tested across hundreds of thousands of gates, measuring worst-case performance. The 99.92% figure for the new systems is a peak value for one qubit type. These metrics are not comparable.

SAXON Q also claims 6 to 10 times better energy efficiency than GPU-based approaches, without disclosing the compared workload, accuracy target, or system boundary.

The Genuine Advance

The comparison to Willow may be misleading, but I would be wrong to dismiss the announcement entirely. There is a real fabrication result underneath the marketing.

SAXON Q builds qubits using nitrogen-vacancy (NV) centers in diamond: a defect in the carbon lattice where one carbon atom is replaced by nitrogen and a neighboring slot left empty. The trapped electron’s spin can be controlled optically at room temperature. NV centers have been a known qubit platform for over two decades, but conventional ion implantation techniques convert only a few percent of implanted nitrogen into usable NV qubits.

In 2019, Meijer and colleagues (Lühmann, John, Wunderlich, Meijer, and Pezzagna) published a 75.3% NV-centre creation yield using sulfur-assisted charge engineering in Nature Communications, roughly a tenfold improvement. SAXON Q now says its proprietary production process exceeds 85%, but that higher number remains company-reported. The peer-reviewed result is 75.3%. Both numbers are creation yields: the proportion of implanted nitrogen that forms NV centers. Neither number directly measures the end-to-end yield of addressable, coherent, gate-qualified qubits on a finished processor.

The practical deployment story is also real. SAXON Q delivered a 4-qubit system to DLR’s Innovation Center in 2023. After extensive testing against thresholds of >95% single-qubit and >90% two-qubit gate fidelity, DLR accepted the system in July 2024, and it has been available via QCI Connect since August 2025. A second 4-qubit system was deployed at Fraunhofer IWU in Dresden in June 2025. Both run at room temperature with no cryogenic infrastructure. Compared with superconducting systems like Willow, which require dilution refrigerators at roughly 15 millikelvin, that deployment simplicity is a genuine advantage.

Three months before the SXQ128 announcement, SAXON Q’s most advanced public system was the QC2026 Dual Core shown at Hannover Messe in April 2026: two parallel 5-qubit cores. The jump from 10 qubits across two cores to 128 qubits across sixteen cores, with no new third-party benchmarking, is where my caution concentrates.

State of the Art

Some context on where NV-center computing stands in peer-reviewed literature. The largest fully controlled NV spin register is the 10-qubit system demonstrated by Bradley, Taminiau, and colleagues at QuTech/Delft in 2019. That register used one NV electron spin plus a nitrogen nuclear spin and eight carbon-13 nuclear spins. The team demonstrated pairwise entanglement across all 45 qubit pairs and genuine multipartite entanglement of up to seven qubits. It operated at 3.7 K, well into the cryogenic range.

The best published two-qubit gate between separate NV electron spins achieved 96% fidelity under ambient conditions (Joas et al., Phys. Rev. X, May 2025), with the authors noting the difficulty of expanding such systems beyond 20 qubits. Quantum Brilliance, SAXON Q’s closest NV-center competitor, has deployed two-qubit QPUs at Oak Ridge National Laboratory and Fraunhofer IAF, with a roadmap toward 25-100 qubit systems.

An independently benchmarked 8-qubit room-temperature core with published entangling-gate distributions would be a significant result. SAXON Q’s materials do not yet establish that benchmark.

What to Watch For

Three things will determine whether this announcement represents a commercial milestone or premature marketing.

First, third-party benchmarking of the SXQ128 at the 8-qubit core level. I want to see full-register two-qubit gate fidelity distributions (all qubits, worst case), coherence times, and circuit depth benchmarks from an independent evaluator. DLR’s methodology for the 4-qubit system provides the model.

Second, clarification of what “fully entangled qubits” means operationally: all-to-all gate connectivity, demonstrated pairwise entangling operations, or genuine eight-partite entangled states. These are materially different claims.

Third, independent reproduction of the yield improvement. The peer-reviewed 75.3% creation yield is already important. If the company-claimed 85%+ figure can be verified and the resulting NV centers meet coherence and gate-performance requirements at scale, it shifts the economics of NV-center fabrication industry-wide.

Bottom Line

SAXON Q has credible founders, a peer-reviewed fabrication advance, and a deployment form factor that superconducting quantum computers cannot match. A 4-qubit system passed DLR’s acceptance testing and runs at a German research institution with no cryogenic infrastructure.

The “128-qubit quantum computer” framing runs ahead of the published evidence. SAXON Q discloses sixteen eight-qubit cores coordinated through circuit cutting with classical recombination. Its own whitepaper identifies inter-core quantum connection as an open problem. Until the company publishes full-core entangling-gate data, demonstrates the meaning of “fully entangled” at the 8-qubit scale, and submits to independent benchmarking, the 128 figure describes an aggregate packaging count. Comparing it to Willow’s 105-qubit coherent register should stop.

I will update this article when independent benchmarks become available. The data will tell.