Making one computer from separate modules
Researchers at the University of Oxford reported a distributed quantum-computing experiment in Nature on 5 February 2025. They linked two trapped-ion processor modules with optical fibre, generated entanglement between them and used that shared resource to perform quantum gates on circuit qubits held in different devices. The team then ran a small version of Grover's search algorithm using several of these non-local gates.
The experiment addresses a central engineering problem. A useful fault-tolerant quantum computer may need very many high-quality qubits, yet packing every qubit, control beam and connection into one trap becomes increasingly difficult. A modular machine could instead connect smaller processors, much as conventional computing systems combine nodes. Quantum modules need links that preserve superposition and entanglement, however, not just cables that copy ordinary bits.
How the optical link performed a gate
Each module contained dedicated network qubits and circuit qubits. Photons emitted by network ions travelled through optical fibres to a central measurement station. A particular pair of detector clicks heralded that the distant network qubits had become entangled. Local operations and classical feed-forward then consumed that entanglement to implement a controlled-Z gate between circuit qubits in different modules, a process called quantum gate teleportation.
No ion or useful object was transported. What was transferred was the operation's quantum correlation, using pre-shared entanglement and classical information. The protocol was deterministic once the required entangled pair had been successfully heralded, and it could be repeated while protecting information stored in the circuit qubits.
The researchers measured about 86 percent fidelity for the teleported controlled-Z gate. The remote entangled state itself reached about 96.9 percent fidelity, and it took an average of roughly 103 milliseconds to generate a successful link in this setup. Those separate numbers show where performance is lost: photon collection and remote entanglement are difficult, while local memory and control also contribute errors.
A real algorithm, at proof-of-principle scale
To demonstrate more than an isolated gate, the team distributed Grover's search across the two modules. The implementation returned the target with a 71 percent success rate. They also assembled distributed iSWAP and SWAP circuits using multiple teleported gates, showing that the link can support a broader family of two-qubit operations rather than one specially chosen demonstration.
Calling this a distributed quantum algorithm describes the architecture, not its practical power. The search problem was tiny and offers no useful speed advantage over an ordinary computer. The measured gate fidelity is also well below what a large fault-tolerant calculation would require without substantial error correction.
Why modularity could still be valuable
Within a small ion trap, qubits can interact with high connectivity and excellent control. Optical photons are better suited to travelling between traps. Assigning those roles to different physical systems lets engineers optimise a local processor and a network interface separately. Additional modules could, in principle, expand a machine without forcing every qubit into a single increasingly crowded device.
The phrase in principle carries much of the challenge. A larger network needs much faster and more efficient photon collection, stable optical paths, high-fidelity remote gates, routing among many nodes and error-correction protocols that tolerate link failures. Waiting for heralded entanglement also requires long-lived memory qubits whose stored states survive repeated network attempts.
The evidence and the boundary of the claim
The published measurements establish that separate matter-qubit processors can execute a circuit containing multiple non-local gates through a photonic connection. They do not establish a quantum internet, a fault-tolerant computer or an easy path to millions of qubits. Distance was about two metres in the laboratory, and every added module will create new control and networking demands.
Even with those limits, the result is a useful systems demonstration. It brings entanglement generation, memory protection, gate teleportation and algorithm execution into one repeatable experiment. Future work can now measure improvements against a complete modular workflow rather than against isolated network components.
