Most quantum computing announcements revolve around the processor. A company unveils more qubits, lower error rates, faster gates or a new architecture that promises to overcome the limitations of previous designs. That is understandable. The quantum processor is where the science happens, and processor performance remains central to the industry’s progress.

But it can also leave us with a misleading picture of what is being built.

A quantum processor is not a quantum computer, any more than a CPU or GPU is a data center. The processor may be the heart of the machine, but a heart sitting on a laboratory table is not a functioning body.

A useful, fault-tolerant quantum computer will require processors, error correction, interconnects, refrigeration, wiring, control electronics, classical computing and software to operate as one integrated system. Progress in one layer cannot compensate indefinitely for bottlenecks in the others.

That is why IBM’s latest quantum announcement matters.

IBM has connected two modular cryogenic cells and operated them as a shared ultra-cold environment. It is not the sort of announcement that lends itself to a headline about a record number of qubits or an imminent quantum revolution. It is, however, another necessary link in the system IBM says will become its Starling fault-tolerant quantum computer in 2029.

Building a Better Quantum Refrigerator

IBM’s superconducting quantum processors must operate at temperatures extremely close to absolute zero. Heat and environmental noise can disrupt the fragile quantum states on which computations depend, so the processors are housed inside specialized refrigeration systems known as cryostats.

Today, those cryostats are typically self-contained cylindrical structures built around an individual processor. That approach has allowed IBM and others to develop increasingly capable machines, but it presents serious problems when the objective shifts from operating one processor to connecting many processors as parts of a larger fault-tolerant system.

There is only so much room inside a conventional cryostat. As the number of qubits rises, so does the need for wiring, control electronics, readout equipment and chip-to-chip connections. Longer connections can introduce more noise. Additional electronics generate heat. At some point, simply making the refrigerator larger stops being a workable scaling strategy.

IBM’s answer is a box-shaped modular cryogenic cell. Each cell contains its own vacuum chamber, dilution-refrigeration equipment and thermal shielding. The cells can be positioned next to one another and joined through a protected cryogenic tunnel. That tunnel allows quantum and classical connections to pass between cells while maintaining the extremely low temperature required by the processors.

IBM has now connected two of these cells and operated them as one cryogenic environment. The combined structure stands more than eight feet tall and eight feet wide. IBM says it reached 4 kelvin—the temperature of liquid helium—in less than five days before cooling below 15 millikelvin. That is more than 180 times colder than deep space.

Each module provides approximately 0.53 square meters of wiring area and 2.75 cubic meters of vacuum-chamber volume. According to IBM, that creates up to 12 times more wiring capacity than its commonly used existing quantum systems. Future modules are intended to house thousands of qubits, while the architecture is ultimately designed to connect hundreds of quantum chips. IBM explains the architecture in more detail here.

Those numbers are impressive, but the larger point is not that IBM has built a very large and exceptionally cold refrigerator. It is that IBM is designing the physical infrastructure around a future in which multiple quantum processors must communicate and operate as one machine.

Scaling Out the Quantum Computer

IBM plans to connect processors using what it calls L-couplers. In plain English, these are cryogenic cables designed to carry quantum information between separate chips across distances of approximately one meter.

That is easier to describe than it is to accomplish. Quantum information is extraordinarily sensitive to noise, heat and signal loss. Connecting quantum processors is not the equivalent of plugging ordinary servers into an Ethernet switch.

Nevertheless, the architectural principle will be familiar to anyone who has followed the evolution of classical computing. There are practical limits to how much can be placed on one chip. Continued growth eventually depends on connecting multiple processors and distributing work across a larger system.

The same pattern is visible in AI infrastructure. The GPU receives most of the attention, but an AI system’s performance depends on memory, advanced packaging, networking, storage, power, cooling and software. A shortage or performance limitation in any one of those layers can constrain the entire system.

Quantum computing is reaching that systems stage much earlier in its commercial development. The cooling equipment, wiring and interconnects are not accessories surrounding the quantum computer. They are parts of the computer itself.

Modularity also allows IBM to improve those parts separately. A processor, control system or section of cryogenic infrastructure could potentially be tested, replaced or upgraded without redesigning the entire installation. Instead of building a single fixed laboratory apparatus, IBM is trying to create a platform that can evolve one subsystem at a time.

Another Section of Track

This milestone needs to be kept in perspective. IBM has demonstrated that two cryogenic modules can be connected and cooled together. It has not yet demonstrated quantum processors performing useful computations across them.

IBM says it plans to install Nighthawk processors in the modules later this year to expand operational testing. Its roadmap calls for L-couplers to connect multiple processors into a system with at least 1,000 programmable qubits in 2027. Starling remains scheduled for 2029. Those are IBM’s stated targets, not guaranteed outcomes. The company’s announcement distinguishes the current infrastructure milestone from the processor deployments still to come.

IBM did not unveil the train this week. It demonstrated another section of track without which the train cannot reach its destination.

In our recent Techstrong Special Report, All Aboard the Quantum Train, we argued that quantum computing should not be understood as a single race toward one decisive breakthrough. It is a long, interconnected engineering and commercial journey. Hardware, error correction, control systems, classical computing, software, infrastructure and applications all have to move forward.

IBM’s modular cryogenic milestone makes that argument tangible. The processor is only one car on the train.

The quantum market is becoming increasingly crowded with newcomers pursuing different qubit technologies, architectures and pieces of the stack. Some may produce better individual components than IBM. Some may reach important milestones sooner. There is no reason to assume the eventual quantum market will belong to one company or one technical approach.

IBM’s distinction is that it is attempting to assemble the complete system. That includes processor design, semiconductor fabrication, error correction, interconnects, cryogenic infrastructure, classical controls, Qiskit software, cloud access and an enterprise and research ecosystem.

Its progress can sometimes appear incremental because it often arrives one subsystem at a time. But those increments continue to correspond with a detailed public roadmap leading toward Starling in 2029.

IBM still has enormous technical challenges ahead. Cooling connected modules is not the same as operating reliable logical qubits across multiple processors. A fault-tolerant Starling in 2029 remains an ambitious target rather than a certain destination.

There will also be more dramatic quantum announcements between now and then. Some will promise better qubits, faster gates or earlier advantage. IBM’s progress can look almost pedestrian because it often arrives one subsystem at a time. But quantum computers will not be built from a processor announcement. They will be built by making every part of an extraordinarily complicated system work together.

This week, IBM added another important link—and continued its steady march toward Starling.