IBM Links Two 120-Qubit Chips Using Superconducting Cables

IBM researchers successfully connected two 120-qubit quantum processors using four superconducting cables, marking a significant step toward modular quantum computing. The study demonstrates that separate chips can share information with 99 percent fidelity, providing a blueprint for building larger systems that exceed the physical capacity of a single processor chip.

Architecture for Modular Quantum Systems

The researchers utilized two IBM Nighthawk R1 superconducting processors for this experiment. Each chip contains 120 qubits, which are the fundamental units of quantum information. In traditional designs, scaling up requires cramming more qubits onto one piece of silicon. This approach eventually hits physical and thermal limits.

By using a modular strategy, developers can instead link several smaller processors. This allows the computer to function as a unified machine while maintaining manageable chip sizes. The team used four aluminum coaxial cables to bridge the gap between the two processors. These cables are roughly 80 centimeters long and function as microwave links.

These links operate inside the ultracold environment required for superconducting qubits. The distance of nearly a meter is necessary to accommodate the cooling equipment and wiring needed for high-performance hardware. This design keeps the processors far enough apart for physical maintenance while keeping them logically connected.

Earlier experiments in the field typically linked processors with only a few qubits or used a single connection point. This project is the first to show a multichannel interconnect between chips of this scale. The team focused on specific qubits near the connection points to verify the technology’s effectiveness.

The cables are attached perpendicularly to the surface of the processor. This vertical orientation is a change from older designs that connected cables parallel to the chip. The vertical approach is more efficient because it allows for a higher density of connections. It also makes the hardware easier to assemble during manufacturing.

Reliability is a major factor in modular design. The team used indium-to-indium pressure bonds to secure the cables at room temperature. These bonds survived three cycles of warming the system up and cooling it back down to near absolute zero. This durability proves the hardware can handle the regular maintenance cycles of a working data center.

Performance and Information Transfer

The primary goal of the experiment was to move quantum states between chips without losing data. Quantum information is notoriously fragile and easily destroyed by environmental noise. The researchers used network benchmarking to measure how well the links performed during these transfers.

All four superconducting links achieved state-transfer fidelities of at least 98.86 percent. The highest recorded fidelity reached 99.29 percent. Fidelity indicates how accurately the quantum state on the receiving chip matches the state originally sent from the source. These high scores suggest the physical cables are highly effective.

Moving a state is only the first step. To perform actual calculations, the system must execute gates between qubits located on different chips. The researchers performed controlled-Z gates, which are essential for creating complex quantum logic. These operations are more difficult because they require multiple steps and controls.

Distant gate operations combined two state transfers with several local gates. Because more steps are involved, there is more opportunity for error. The average fidelity for these inter-chip gates was 94.75 percent. The best performance for a single gate reached 96.42 percent. This confirms that logic can be executed across the modular boundary.

The team also successfully generated entanglement between the two chips. Entanglement is a state where two qubits become linked regardless of the distance between them. They created entangled pairs and even larger groups of four and six entangled qubits distributed across the chips.

These multi-qubit states prove the connections support more than just simple point-to-point transfers. A four-qubit state across the chips reached 77.23 percent fidelity before any error correction. These results stayed above the mathematical threshold required to prove genuine entanglement existed. The ability to coordinate multiple qubits across chips is vital for scaling algorithms.

Overcoming Challenges and Future Improvements

Despite the success of the links, the researchers identified several areas where the technology needs to evolve. One major issue involved magnetic interference within the cables. When the aluminum cables cooled down to become superconductors, they trapped magnetic fields that hindered performance.

To solve this, the team added two layers of magnetic shielding around the cables. They found that the ability of the cable to hold energy dropped if the magnetic field was too high during the cooling process. While the shielding worked for this experiment, the researchers noted that a more permanent solution is needed for commercial systems.

The study used an error model to find the source of most performance issues. About 77 percent of the errors came from the qubits themselves rather than the cables. Specifically, the qubits lost their stable phase relationship, a problem known as dephasing. This means the cables are actually performing better than the chips they connect.

The researchers pointed out that they skipped certain fabrication steps to speed up the testing process. These missing steps are known to improve qubit stability. Therefore, they expect that future versions using higher-quality chips will see even better results. Their models suggest that improved qubits could push state-transfer fidelity toward 99.7 percent.

Another planned improvement involves the way the chips are laid out. Currently, the connection elements are separate from the main computing area. Future designs will likely integrate these links directly into the computing grid. This change would remove the need for extra internal steps, making the entire process faster and more accurate.

The transition from four links to hundreds of links is the next major hurdle. Manufacturing these systems at scale requires high levels of precision and repeatability. Every connection must be perfectly aligned and shielded to prevent interference between the channels. The IBM team’s work provides the initial evidence that such a feat is physically possible.

Building an error-corrected quantum computer requires thousands or even millions of qubits. Since no single chip can currently hold that many, modularity is the most likely path forward. This experiment proves that the “quantum glue” holding these modules together is strong enough to maintain the necessary quantum properties for complex work.

The researchers from the Thomas J. Watson Research Center have established a foundation for this scaling. By solving the mechanical and quantum challenges of chip-to-chip communication, they have moved modular quantum computing from a theoretical concept to a functional reality. The focus now shifts to refining these components for industrial-grade applications.