IBM Designs Modular Cryogenic System for Scalable Quantum Computing

IBM, Google, Rigetti, and IQM Quantum Computers use superconducting approaches for their quantum computing systems. These systems must operate at ultra-cold temperatures near absolute zero, generally between 10 and 20 millikelvins. Meeting that requirement creates engineering challenges involving cooling-system cost, physical size, and scalability.

In superconducting quantum computers, qubits are vulnerable to external disturbances, including light, noise, and interactions with other qubits. Heat creates atomic vibrations, known as thermal noise, that can break apart the delicate electron pairs responsible for the quantum state. When that happens, data can be lost before a calculation finishes.

IBM has introduced a modular cryogenic cooling subsystem that the company says can reach 15 millikelvins. The system is part of IBM's preparation for Starling, its planned first fault-tolerant quantum system, expected in 2029. Starling is designed to support as many as 200 logical qubits and up to 100 million quantum calculations, 20,000 times more than IBM quantum computers can perform today. The system is also expected to include advances in error correction and processor design.

A shared ultra-cold environment

The new cryogenic infrastructure consists of two modules that together are more than eight feet tall and eight feet wide. When connected, they reach 4 kelvins, approximately the temperature of liquid helium, in fewer than five days. They then reach a final temperature of 15 millikelvins, which is more than 100 times colder than outer space.

According to Jerry Chow, IBM Fellow and chief technology officer for quantum-centric supercomputing at IBM, the system is more than a cooling platform. It is an architectural step toward scaling quantum computers. Chow and Oliver Dial, IBM Fellow and vice president of quantum systems, presented the infrastructure.

The modular architecture creates a shared ultra-cold environment where multiple quantum chips can be connected. It also provides space for the high-density wiring required by larger systems.

“When we talk about scaling, it is not just about building bigger and bigger chips at the processor level,” Chow said. “It is really about all the infrastructure and the supporting pieces around it as well in the system.”

The design is intended to support faster development cycles. By building the cooling system, processors, controls, and other components together, IBM scientists can test and refine future quantum systems more quickly.

Built for shipping and maintenance

The cryogenic system can be separated into modules, which makes it easier to ship to customer sites and reassemble. Dial said the design could support systems that are “arbitrarily large.” IBM expects Starling to use approximately a dozen connected modules.

The modules include doors for installation and maintenance. Metal electromagnetic-interference gaskets around the doors help block radio waves and other electromagnetic radiation that could disturb the quantum processors. Rubber O-rings create vacuum seals.

A heat shield made from layers of Mylar, which researchers refer to as “super insulation,” helps limit heat transfer. Inside the system, a series of metal shields provides progressively lower temperatures, while a dilution refrigerator supplies the cryogenic cooling.

The architecture also reserves substantial space for wiring. IBM says the system has approximately 12 times more wiring area than its Quantum System 1. Wiring density will be important for installing more complex quantum processors.

“As you go lower and lower in the fridge, it gets colder and colder until at the very bottom, you reach the point where quantum processors will be installed,” Dial said.

Components and L-couplers

IBM is sourcing many components, including the cryogenic cooling engines, from established vendors. One system operating in IBM's laboratories in Poughkeepsie, New York, uses equipment from Bluefors, a Helsinki-based Finnish company that produces cryogenic systems for quantum computing and other applications.

Using established components can accelerate engineering and simplify future maintenance, servicing, and upgrades, Chow said.

The modular layout also reduces the distance that IBM's L-couplers must span between processors. IBM first tested the L-couplers on Flamingo, a proof-of-concept multichip quantum processor introduced in 2024.

L-couplers connect quantum chips and enable information sharing between them. Conventional on-chip couplers operate over short distances to create the entanglement needed for two-qubit gates. L-couplers extend that capability through an aluminum superconducting cable up to approximately one meter long.

The longer connections allow processors to be placed farther apart, creating more room for wiring and making it easier to test, debug, and replace individual processors. They also allow a quantum computer to span multiple connected modules.

“We're talking about building an extremely dense interconnected machine,” Dial said, describing the expected use of L-couplers in future systems.

Planned demonstrations

Chow said IBM plans to conduct a demonstration later this year using one Nighthawk quantum processing unit in each of the two cryogenic units. Nighthawk is a 120-qubit QPU introduced last year.

By the time Starling is available, Dial expects the modules to support approximately 2,000 physical qubits, equivalent to about four QPUs.

IBM's cooling infrastructure reflects the company's current focus in quantum computing, Chow said. The underlying science for fault-tolerant quantum computing has been established, while reaching that goal now depends on many engineering improvements across processors, software, controls, infrastructure, and error correction.

“It's not about a single breakthrough to get to fault tolerance,” Chow said. “It's really about thousands of these little engineering feats that we're demonstrating all across our entire ecosystem.”