IonQ Quantum Interconnect Hits 1,000 Entanglements Per Second

A faster link between quantum machines

IonQ has reported a photonic interconnect that generates more than 1,000 entanglement events per second, a rate of over 1 kHz, between a trapped-ion qubit and a solid-state quantum memory.2 The company announced the result on October 9, 2026, and says it beats the previous trapped-ion interconnect record by more than four times.16 That earlier record came from the Duke University group led by IonQ co-founder Chris Monroe, which also collaborated on the new paper.2 The company's own release cites the earlier benchmark as O'Reilly et al., published in Physical Review Letters in 2024.2

The experiment pairs two different kinds of quantum hardware. A trapped-ion qubit, which holds its quantum state for a long time, is linked by light to a silicon-vacancy (SiV) memory, a solid-state device that couples efficiently to photons.1 IonQ describes the combination as getting the strengths of both, and claims it is the fastest qubit-to-qubit interconnect rate on any platform.2 Results from tests of an end-to-end hardware link are in an arXiv preprint.25

The company framed the result around its business strategy. CEO Niccolo de Masi compared quantum scaling to the way classical data centers grew by networking specialized processors, memory and communications. He said the result clears a key interconnect bottleneck on the way to "networked quantum data centers."2 Chief Scientist Monroe said that moving qubits with photons will be needed in any large-scale quantum computer.1

Why the connection rate matters

Quantum information cannot be copied and sent the way classical bits are. Entanglement between separate systems is the resource that lets distant qubits work together, which makes it the key input for both quantum networking and distributed computing.1 The difficulty has been producing entanglement fast and reliably enough to be useful. A slow link makes a multi-chip machine wait on its connections.1

In my reading, this makes the result more important than a typical lab record. IonQ's long-term plan depends on photonic links. The company's accelerated roadmap calls for an interconnected two-chip module with about 20,000 physical qubits by 2028, supporting about 1,600 error-corrected logical qubits. It then aims for more than 2 million physical qubits across many linked modules by 2030.15 The 2025 purchase of Lightsynq, whose photonic quantum-memory interconnects were described as raising ion-ion entanglement rates by up to 50 times, was meant to make that networking possible.15 The new demonstration is the first strong public evidence that the memory-buffered approach works on real hardware at a meaningful speed.

Mihir Bhaskar, IonQ's SVP and GM of Quantum Technologies at SkyWater, said the result shows a photonic interconnect "need not be a bottleneck." He added that the technology can connect to almost any qubit type.4 IonQ expects the architecture to extend to neutral-atom systems and to superconducting processors fitted with transducers that convert microwave signals into light.1 That fits DARPA's Heterogeneous Architectures for Quantum (HARQ) program, which funds high-speed interconnects that work across several qubit technologies and which IonQ is part of.26

Where the coverage agrees, and where it doesn't

The coverage agrees on the main facts: the 1 kHz rate, the more-than-fourfold gain, the ion-plus-SiV design and the HARQ connection.136 Much of it repeats the company's press release closely.34 There is one notable inconsistency. One outlet describes the rate as beating IonQ's own previous record.5 The company's release instead credits the prior benchmark to Monroe's Duke lab.2 Given Monroe's dual role, the difference is small in practice, but the Duke attribution is the accurate one.

The more useful split is between coverage that reads like a press release and coverage that adds caveats. The most careful account says that a high entanglement rate alone does not make large-scale distributed quantum computing commercially ready. It notes that connection fidelity, network coordination, error correction and multi-module integration all still have to be solved.1 That is the right way to read the result. The announcement focuses on speed, and none of the coverage reports how accurate the generated entanglement is. For any link meant to carry error-corrected logical qubits, accuracy matters as much as speed.

Investors also seemed to discount the news. IonQ shares slipped 0.33% to $39.32 in intraday trading on the day of the announcement.89 Two weeks earlier, a real-time error-correction decoder announcement sent the stock up more than 11%.16

Part of a run of fault-tolerance results

The interconnect result is the latest in a series of IonQ announcements in late 2026. On September 22, the company said it had built the industry's first end-to-end real-time quantum error-correction decoder running on a single off-the-shelf CPU.12 In simulated benchmark circuits of up to 408 logical qubits and more than 31.5 million operations, the decoder added as little as 0.02% extra run time.14 Those were simulations, not runs on physical hardware, which matters when comparing it with the interconnect result. IonQ said the decoder validates a core part of its "Walking Cat" fault-tolerance architecture.16 The company laid out that architecture in an April technical report describing a path to 10,000 physical qubits and beyond.1320 Stock-page listings also show IonQ advancing to the final stage of DARPA's Quantum Benchmarking Initiative in early October.2

Taken together, the two results cover two different ways a modular machine can fail. The decoder addresses the classical computing load of error correction. The interconnect addresses the quantum link between modules. A blog post on fault tolerance makes the same point about IonQ's design: buffered photonic links are meant to let logical qubits be spread across modules without losing much fidelity.15 The October result advances the "fast" half of that goal. The "without losing fidelity" half has not been shown publicly.

Roadmap milestones: ambition versus delivery

IonQ's targets are among the most aggressive in the industry. One tracker lists its goals as 100 to 256+ physical qubits and 12 logical qubits in 2026, 10,000 physical and 800 logical qubits in 2027, and 20,000 physical and 1,600 logical qubits in 2028.11 IonQ has fabricated its first integrated 256-qubit Superion QPUs, pre-sold one system in the first quarter of 2026, and targets customer deliveries in 2027.11 Separately, the company aims to have an operational 256-qubit system in the fourth quarter of 2026.19

The track record calls for some caution. IonQ's original SPAC-era roadmap projected about 4,000 physical qubits by 2026 and 32,000 by 2028, targets it is far from meeting on its own. A 2025 short-seller report specifically questioned whether photonic interconnects and modular scaling would work in practice.15 One analysis also notes that IBM gives itself until 2029 to reach a few hundred logical qubits, while IonQ aims for about ten times as many in roughly the same period.15 Rivals' reported results add context. Quantinuum's Helios shipped with 50 error-corrected logical qubits at a 2:1 encoding ratio.11 Infleqtion reported 30 entangled logical qubits in September 2026.11

This is where the interconnect result matters most. Doubts about IonQ's roadmap have centered on photonic networking, and a fourfold jump in entanglement rate on real hardware directly answers part of that criticism. It does not show a networked logical qubit. Whether the 2027 and 2028 targets are met will depend on combining fast links, high-fidelity chips and real-time decoding in one working system.

The post-quantum cryptography angle

IonQ's roadmap also matters for security planning. One analysis notes that the company's 2028 target of about 1,600 logical qubits is close to the roughly 1,000 to 1,400 logical qubits that Gidney et al. estimated in May 2025 would be enough to factor an RSA-2048 key in about a week.15 That would also require very low logical error rates and a stable week-long computation. IonQ has not made public predictions about breaking encryption.15 The same analysis argues that governments and industry should treat the late 2020s as the period when cryptographically relevant quantum computers could arrive, and plan migrations accordingly.15 NIST finalized its post-quantum cryptography standards in 2024, and U.S. federal agencies are working toward a 2030 migration target.17

The interconnect fits into that picture because a 1,600-logical-qubit machine on IonQ's plan would be built from linked modules, not a single chip.15 Faster entanglement between modules is one requirement for that design. This milestone does not change cryptographic risk timelines, but it removes one engineering obstacle that would otherwise have pushed them later.

Early commercial signals

IonQ is already selling the memory and interconnect platform. Its first sale went to the University of Maryland in April, and a second system went to South Korea's SDT in September.2 One report puts the Maryland agreement at $7.5 million, including a silicon-vacancy memory node.6 The SDT deal pairs a Superion 256 quantum computer with an SiV memory module.6 These are small, research-focused purchases, not data-center deployments. Still, they suggest IonQ is building a networking business alongside its main computing roadmap.

The bottom line

This result removes a long-standing objection that photonic links between trapped-ion modules would be too slow to matter. It does not yet show that those links can carry error-corrected computation across chips, and that is the step that would make IonQ's 2027 and 2028 roadmap credible.