A Quantum Computer Just Held Hands With a Diamond 1,000 Times a Second. The Fine Print Matters.

Last week, a quantum computer in Maryland did something its designers have been chasing for years: it shared a quantum link with a completely different kind of quantum device, more than a thousand times per second. The announcement came from IonQ on October 9, backed by a technical preprint posted on October 7, and the numbers are real. The experiment generated an average of 1,032 entangled pairs per second between a trapped ion and a silicon-vacancy center in diamond, at 87.9 percent Bell-pair fidelity.

That rate is more than four times faster than the previous trapped-ion record. It is also not peer-reviewed yet, was measured in a laboratory link between two devices rather than a network, and still sits one to two orders of magnitude below what distributed quantum computing is thought to require. Both facts are true at once, and the honest version of this story needs all of them.

A match between two species

The pairing itself is the interesting part. The computing qubit is a single barium-138 ion, a charged atom held in place by electric fields, chosen because trapped ions offer long coherence times and high-fidelity logic gates. The memory is a silicon-vacancy center, an atomic defect in a diamond crystal embedded in a tiny optical cavity, chosen because it couples to light unusually efficiently. They are different machines speaking different physics, joined by a photonic interface.

To connect them, the experiment has the ion emit a single photon at 493.5 nanometers, converts its wavelength to 737.45 nanometers to match the silicon-vacancy transition, reflects it off the diamond cavity, and detects it. The detection event "heralds" success, meaning the control system knows exactly when a linked pair exists instead of having to guess. Attempts repeat every 896 nanoseconds, about 1.116 million attempts per second, with roughly 731 attempts needed on average for each successful pair. Success is rare per attempt (about 0.137 percent); the rate is high because the attempts are relentless.

One entangled pair per millisecond sounds exotic. In engineering terms, it is a pipeline number, and pipelines are what scale.

The honest accounting

Start with what is genuinely impressive. Generating entanglement at kilohertz rates across heterogeneous hardware is a real engineering achievement. Each pair arrives roughly once per millisecond, which is close to the planned cycle time for IonQ's trapped-ion processors. If a processor runs on millisecond cycles and remote links deliver a fresh pair every millisecond, the link no longer leaves the processor waiting. That alignment matters, because separate quantum processors must hold entangled pairs before they can exchange quantum information.

Now the caveats, which the sources supply. First, the 1,000-per-second figure measures pair generation before readout. The experiment spent 664 microseconds reading each successful event, so including readout overhead, wall-clock throughput drops to 612 pairs per second. Still fast, but a different number. Second, the paper is a preprint posted to arXiv (2610.10705); it has not yet been peer-reviewed. Third, fidelity. The measured 87.9 percent Bell-pair fidelity is lower than the above-94-percent fidelity bound reported by the earlier 250-per-second record from the Monroe group at Duke, and that earlier result linked two identical ions. This comparison is not apples to apples: the new experiment trades some fidelity for a much higher rate across unlike systems, and the previous record remains the better apples-to-apples benchmark for ion-to-ion links. Fourth, the demonstration is a laboratory link between two devices. IonQ describes no network of machines, and neither the company release nor the preprint describes error-corrected computation.

The road to the quantum data center

IonQ's framing is instructive: classical data centers scaled by networking specialized processors, memory, and networks together, and the company argues quantum systems will scale much the same way. The result sits within DARPA's Heterogeneous Architectures for Quantum (HARQ) program, which funds exactly this kind of cross-technology interconnect research. On the commercial side, IonQ has reported sales of silicon-vacancy quantum memory systems to the University of Maryland and to SDT in South Korea, and the company says the new paper supplies performance data missing from those commercial announcements.

The preprint's own authors set the distance remaining. Distributed quantum computation, they write, needs entanglement rates of 10 to 100 kilohertz, ten to a hundred times faster than what they measured. (Why the rate matters: our explainer on quantum error correction.) They project more than 5 kilohertz from better ion-to-fiber coupling, lower signal-chain loss, and a different spin-photon gate, but that projection is not measured performance. Beyond rate, fault-tolerant distributed computing also needs higher pair fidelity, entanglement swapping, actual storage of heralded states in the longer-lived nuclear spin, and processor-level tests in one system. In this demonstration, the experiment entangled the silicon-vacancy's electron spin; the silicon-29 nuclear spin, which offers longer-lived storage, was initialized but not used in the reported link. The single-photon reflection protocol scales favorably, with success rising linearly with collection efficiency, which is part of why the authors are optimistic.

None of this diminishes the result. It situates it. A kilohertz heterogeneous link is one solved piece of a puzzle with several unsolved pieces still on the table.

What it means in perspective

What happened here is better described as plumbing than as a revolution. The most useful quantum computer will probably not be one machine of one type; it will be an assembly of specialized parts, computing qubits of one species, memory of another, connected by photonic links. This week, for the first time, that link ran at processor-relevant speed between two species of qubit. The quantum data center is not here. But the wires are starting to hold.