Duke and IonQ Link 3 Quantum Computers Over a Photonic Network

For years, building a useful quantum computer has felt like trying to scale a skyscraper one floor at a time — every new qubit added makes the system harder to control. A new result from Duke University and IonQ, announced on June 20, 2026, suggests another path: instead of cramming everything into one chip, link smaller quantum machines together over a network.

In a demonstration described in a peer-reviewed paper, researchers from the Duke Quantum Center and IonQ entangled three remote atomic ions located in separate trap modules, connecting them with photonic channels. The team produced a three-qubit Greenberger–Horne–Zeilinger (GHZ) state — a hallmark of genuine multi-party quantum entanglement — without relying on local two-qubit gates between the nodes or on post-selection tricks that throw away unfavorable outcomes.

That distinction matters. Previous experiments have entangled two distant qubits, and others have produced three-party entanglement inside a single processor. Stitching together three physically separate nodes into one entangled state, deterministically, is what a real quantum network needs in order to run distributed algorithms or build a quantum internet.

According to the Duke and IonQ team, the experiment achieved fidelities in the 84% to 88% range for the remote entangled pairs, with the full GHZ state verified through standard quantum state characterization. The performance is high enough to be useful as a building block, and the architecture is designed to scale: each node is a self-contained ion-trap module, and additional modules can be added by extending the photonic interconnects.

Trapped ions are an attractive technology for this kind of work. Individual atoms — in this case ytterbium ions held in vacuum by precisely tuned electric fields — make naturally identical qubits with long coherence times. When an ion emits a photon, that photon can carry quantum information to a partner ion many meters away, where a measurement on the photons "fuses" the two ions into an entangled pair. Repeating that trick across three nodes and weaving the results together is what produces the new GHZ state.

The practical upshot is what scientists call modular quantum computing. Today’s most ambitious processors push to pack more qubits onto a single chip, which gets harder as the qubit count climbs. A modular approach instead links many smaller, well-controlled processors with photons, letting designers add capacity by adding boxes rather than reinventing the chip. It also opens the door to a future quantum network where nodes in different labs, cities, or even countries could share entanglement for secure communication and distributed computation.

For IonQ, the result strengthens its bet on trapped-ion hardware, which competes with superconducting qubits used by other industry leaders. For Duke, it builds on a long line of academic work in quantum networking that has steadily pushed the number of entangled remote qubits upward. Three may sound modest, but in this field each additional node is a major engineering challenge, and the techniques that make three nodes work tend to point the way to four, eight, and beyond.

There is still a long road between this kind of demonstration and a fully practical quantum internet. The team’s near-term goals include raising fidelity further, increasing the rate at which entanglement is generated between nodes, and integrating error-correction protocols that can tolerate the imperfect photons that link the modules.

But the broader message is encouraging. Quantum researchers are no longer just chasing better single processors; they are starting to think about quantum systems the way classical engineers think about data centers, where many machines work together over fast links. With three trapped-ion nodes now confirmed to share genuine entanglement, the blueprint for that future just got a little more concrete.