Most quantum computing platforms need something extreme to keep qubits alive: a dilution refrigerator cooling to millikelvin temperatures, an ultra-high vacuum chamber, or a network of lasers holding trapped atoms in place. At SaxonQ, we’ve taken a different path. Our qubits live inside an artificial diamond chip, and they work in ambient conditions — room temperature, in air, with no vacuum and no cryogenics. That single design choice changes what a quantum computer can be: not just a machine in a data center, but something mobile enough to roll on wheels, and eventually small enough to sit on a chip alongside its own control electronics.

What makes a diamond qubit

The core of our technology is the nitrogen-vacancy (NV) center: a point defect in a diamond lattice where one carbon atom is replaced by a nitrogen atom, sitting next to a missing carbon atom. When negatively charged, the electronic structure of this defect gives us our first qubit — an electron spin with a ground-state splitting in the gigahertz range. Around each NV center, a diamond’s naturally occurring carbon-13 nuclei — about 1.1% of all carbon atoms — provide additional nuclear-spin qubits we can address individually, typically six or more per NV center.

Getting a high yield of usable NV centers has historically been the bottleneck. Simply implanting nitrogen into diamond converts only about 1–10% of implanted atoms into working NV centers. Six years ago, we discovered that co-doping with sulfur changes that dramatically: sulfur donates electrons that charge the NV center correctly, prevents damaging vacancy clusters, and captures stray hydrogen that would otherwise disable the defect. That process, which we’ve since patented and continued to refine, now delivers yields of 85% or higher — and last year, independent density-functional-theory work confirmed the microscopic mechanism we’d proposed.

From single qubits to working algorithms

High yield is only useful if the resulting qubits perform well. Our randomized benchmarking shows gate fidelities above 99.9% for electron-spin gates, and 95–97% for two-qubit operations in the nuclear subspaces — all at room temperature. To test whether this translates into real computation, we ran Grover’s search algorithm across all eight possible three-qubit search targets. On average, our system found the marked state with 77% probability, compared to the 37% you’d expect from random classical guessing. That’s a clear, measurable quantum advantage on real hardware, not just a simulation.

Our roadmap scales in two directions at once. Within a single NV center and its surrounding nuclear spins, we get eight qubits — what we call a “qubyte.” Pairing two NV centers, entangled via a dipole-dipole interaction gate at a separation of roughly 10 nanometers, gives 16 total qubits. From there, we’re working toward a 4×4 grid of NV centers on a single chip: 16 electronic qubits, each controlling roughly six or seven nuclear spins, for on the order of 100 entangled qubits in one register.

The other axis is parallelism. Rather than chasing only bigger individual registers, our systems run multiple cores side by side — our SXQ128 and SXQ512 systems already run 8 and 16 entangled qubits per core, respectively, across 32 or more independent cores. Multiple cores don’t give you the quadratic or exponential advantage of a bigger algorithm on one register, but they let you run the same workload many times faster in parallel, split a data set across cores, or simply serve multiple users without anyone waiting in line.

None of this works without precise, synchronized control. Each NV center core needs one microwave channel to drive the electron-spin qubit and one radio-frequency channel to address its nuclear spins — and because different nuclei around a given NV center resonate at slightly different frequencies (due to their random positions in the lattice), a single RF channel can address many nuclear qubits at once through frequency multiplexing. Quantum Machines’ control hardware sits at the center of this: a fully loaded system gives us the microwave and RF channels we need to drive 32 cores of 16 qubits each — 512 qubits under coordinated control from one platform. As we scale further, integrating this control electronics directly onto the diamond chip itself is the next major engineering milestone, and one we’re pursuing together.

Why room temperature matters

Room-temperature, ambient operation isn’t just a convenience — it opens deployment scenarios no other qubit modality can reach: quantum computing at the edge, embedded in robots or autonomous vehicles, in precision medicine settings, or eventually in handheld devices. It also matters for energy. A dilution refrigerator-based system can draw roughly two orders of magnitude more power than the GPUs it’s meant to compete with; our systems target roughly a tenth of that. The first realistic milestone for quantum advantage may not be beating classical computing by orders of magnitude — it may simply be matching it, using far less power.

Scaling diamond quantum computing to the markets that matter — automotive systems, AI data centers, aerospace, robotics — depends on three things happening together: more qubits and higher fidelities, a smaller physical footprint, and lower energy per gate operation. We believe the physics doesn’t stand in our way here; what remains is largely an engineering challenge, and one with a clear roadmap in front of it. Thank you to everyone who joined our recent seminar and asked such thoughtful questions — they’ve already shaped some of what we’re prioritizing next.