IonQ's photonic link entangles a trapped ion with a diamond memory over 1,000 times a second

The University of Maryland bought the first commercial version of IonQ's quantum memory and interconnect system, announced in April 2026 ^[8]. SDT, a South Korean company, bought another, announced in September ^[9]. Those buyers are the first users of this result. Whoever signed their purchase orders will soon ask the people running the systems what a faster link does for their own experiments.

The pitch is wide. IonQ chairman and CEO Niccolo de Masi compared the problem to how conventional data centers grew, with specialized processors, memory and networking letting computing scale past a single machine ^[5]. Interesting Engineering's report adds networked sensing, where linked devices coordinate measurements across locations ^[18]. What IonQ actually did is narrower. It joined a trapped-ion qubit and a silicon-vacancy qubit in a solid-state memory with light over a photonic interconnect ^[2]. Then it tested the full connection on real hardware and wrote the results up in a technical paper ^[4].

More than 1,000 entanglements a second works out to at least one entangled pair per millisecond on average ^[13]. The report argues that link speed matters because slow connections could cap how fast separate processing units run joint operations, even when each unit's qubits work well ^[11]. Given IonQ's rate, it concludes, optical communication need not become a major bottleneck for distributed quantum computing ^[12].

The account does not report a fidelity for those pairs, or the figure for the earlier trapped-ion record IonQ says it beat ^[15]. A team planning a networked computer needs both numbers. The report itself names the next challenge as turning faster entanglement into reliable operations across more complex networks ^[10]. Chris Monroe, IonQ's co-founder and chief scientist, said photonic links will play an essential role in large-scale quantum computing ^[6].

So the evidence backs half of the case for scaling by linking machines, the half that says the link can be fast. The other half is whether work done across the link comes out right often enough to compute on.

I'd sort any interconnect claim on two axes. One is entanglement rate, which IonQ has now published. The other is the success rate of operations run across the link, which it has not. A result that scores on rate alone, as this one does by IonQ's own account ^[1][10], justifies a pilot where the buyer measures reliability on its own bench, and stops short of an architecture plan that puts dates on joining separate modules into one machine. I'd recommend treating 1,000 pairs a second as a ceiling on throughput and asking IonQ for operation success figures before writing a modular roadmap around it. The cost of waiting is that the two early buyers will have their own reliability data before anyone else does ^[14].