On July 21, 2026, Germany's SAXON Q began accepting orders for two quantum computers based on nitrogen-vacancy (NV) centers in diamond: the "SXQ128" and "SXQ512." Their total qubit counts reach 128 and 512 respectively, which the company describes as the first diamond-NV-based products to exceed 10 qubits. However, not all qubits are entangled together as a single register. The range that can be fully entangled is 8 qubits per core for the SXQ128 and 16 qubits per core for the SXQ512—meaning the announced scale and the scale at which unified computation actually occurs must be read as two separate things.
The SXQ128 is scheduled for delivery within three months of ordering, while the SXQ512 will begin shipping from Q2 2027. Both are made-to-order systems configured per customer, and pricing has not been disclosed. The previous-generation machine SAXON Q unveiled in April 2026 consisted of two 5-qubit processors running in parallel. This announcement of 128 and 512 qubits, coming just three months later, represents a major leap—but actual delivery of the new machines and third-party performance verification still lie ahead.
How to Count 128 and 512
The point requiring the most caution in the new product specifications is that the total qubit count does not match the range of entanglement within a core. SAXON Q explains that its proprietary multi-core OS coordinates multiple quantum processing cores simultaneously, while the number of qubits that can be fully entangled is specified on a per-core basis.
| Model | Total System | Fully Entangled Range | Delivery Schedule |
|---|---|---|---|
| SXQ128 | 128 qubits | 8 qubits per core | Within 3 months of order |
| SXQ512 | 512 qubits | 16 qubits per core | From Q2 2027 |
This distinction affects how far a quantum algorithm can be loaded onto a single core. Even if multiple cores can be used in parallel, this is not the same as a configuration where 128 or 512 qubits share a single quantum state. SAXON Q's announcement does not clarify how work is divided between cores, whether quantum operations spanning cores are possible, or how much overhead this creates during execution. Therefore, simply comparing total qubit counts against superconducting or ion-trap processors does not constitute a meaningful comparison of computational capability.
The company cites variational algorithms, quantum chemistry simulation, and quantum amplitude estimation as applications for the SXQ128. The SXQ512, with more NV-center-based modules and greater per-core capacity, is aimed at heavier research and applied computing tasks. However, no results have been presented showing that classical computers have been outperformed in these applications. Error correction and logical qubits remain on the development roadmap; the 128 and 512 figures announced here represent totals of physical qubits.
Room-Temperature Operation of NV Centers and Co-Implantation with Sulfur
An NV center is a defect created by replacing a carbon atom in a diamond crystal with a nitrogen atom and forming a vacancy at an adjacent site. SAXON Q controls the electron spin of the NV center and the surrounding nuclear spins as qubits. This approach allows quantum states to be handled at room temperature, without cooling to near absolute zero using a dilution refrigerator. The 4-qubit machines delivered to DLR and Fraunhofer IWU also operate in ordinary room environments.
The manufacturing challenge lies in creating targeted NV centers from implanted nitrogen with high yield and stabilizing them in a negatively charged state. SAXON Q uses a proprietary process that co-implants nitrogen and sulfur into the diamond. A 2024 technical presentation abstract explained that this could raise conversion yield to nearly 90%, compared to the conventional sub-10% rate. In this latest announcement, the company claims to have achieved over 85%, compared to the typical 1–10% range.
However, the proportion of successfully created NV centers and the precision obtained during actual computation are separate metrics. SAXON Q states that the new machines achieve fidelity of up to 99.92%, explaining this as fewer than one error per 1,000 operations, but it has not disclosed whether this applies to single-qubit or two-qubit gates, how many qubits were measured, or the measurement procedure used.
Room-temperature operation makes it easier to minimize footprint, cooling infrastructure, and operating power. The new products reportedly fit into standard server racks and can run continuously on ordinary power supplies. The company also claims 6 to 10 times better energy efficiency compared to GPU-based approaches. However, since the compared workloads, computation times, and system boundaries have not been disclosed, the operational cost advantage cannot currently be confirmed numerically.
The Distance from a 4-Qubit Demonstration to an Order-Ready Product
SAXON Q's track record that has been closely verified by third parties is the "SQ-RT with Princess QPU," accepted by the German Aerospace Center (DLR) in July 2024. This device consists of one NV electron spin qubit and three nuclear spin qubits. DLR set thresholds of over 95% for single-qubit gates and over 90% for two-qubit gates, targeting all four qubits. Verification used dozens of types of tests and hundreds of thousands of gate operations, establishing a lower bound for the entire register.
At the time, DLR outlined the next stage as 8 qubits formed by coupling two NV centers, followed eventually by up to 32 qubits. Fraunhofer IWU also brought a SAXON Q 4-qubit machine online on June 11, 2025, for manufacturing-related research. Furthermore, in April 2026, SAXON Q unveiled the "QC2026 DUAL CORE," which independently controls two 5-qubit cores running in parallel. Up to this point, a clear path can be confirmed: operating a small number of qubits as a product in a room-temperature environment.
On the other hand, the 99.92% fidelity figure for the SXQ128 and SXQ512 is a "maximum value," which differs in nature from the lower bound for the entire register that DLR adopted. Because the measurement targets and procedures do not align, one cannot conclude that precision improved from the 90% range to 99.92%. SAXON Q's current technical webpage also lists a standard 80-qubit multi-core configuration, yet the company's machine listed in DLR's computing resource catalog remains at 4 qubits. What will determine the value of the new products is whether the announced totals can be reproduced on actual hardware and whether third parties can measure operational precision both within and across cores.
The Performance Data That Should Accompany Shipment
The start of order-taking marks a turning point where the diamond NV approach attempts to move from small research-scale machines toward server-rack-ready products. According to SAXON Q, both models can be expanded by swapping diamond chips or adding cores. Combined with the absence of cooling infrastructure requirements, this could make it easier for companies and research institutions to adopt smaller configurations first. Support for Qiskit, OpenQASM, and a proprietary quantum gate language is also practical, as it allows testing from existing development environments.
However, the information needed to make a purchasing decision is not yet all in place. Beyond the undisclosed pricing, the names of customers who will adopt the new machines have also not been revealed. On the performance side, what's needed is the single-qubit and two-qubit gate fidelity across the entire 8- or 16-qubit core, along with effective performance when computation is split across cores. For energy comparisons, the time required to solve the same problem at the same precision, along with power consumption including peripheral equipment, is essential.
The delivery timelines—within three months of ordering for the SXQ128 and from Q2 2027 for the SXQ512—will serve as the first checkpoints. If DLR-style third-party testing on delivered machines can measure both full-core and multi-core operation, diamond quantum computers can advance from being "compact machines that run at room temperature" to computing platforms that maintain performance even as they scale up.