//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>
IBM announced this morning that it has connected and cooled the first two modules of a new cryogenic architecture designed to support multiple quantum processors, marking another step on its roadmap toward a fault-tolerant quantum computer in 2029.
But the milestone also illustrates how the quantum scaling problem is changing. Building larger systems is no longer simply a matter of putting more qubits on a chip. The surrounding infrastructure—cooling, wiring, control electronics, processor-to-processor links, and system reliability—has to scale with them.
The two connected modules, operating at IBM’s Poughkeepsie, New York, facility, can cool to below 15 millikelvins, the company said. Their box-shaped design allows them to sit alongside one another, creating a shared cryogenic environment with considerably more room for wiring and processor interconnects than IBM’s earlier systems. According to IBM, each module offers up to 12 times more wiring space than its most widely used quantum systems.
At a pre-announcement press event on Monday, IBM emphasized that the system is operational cryogenic hardware rather than a concept design. But quantum processors are not yet operating across the two modules. The company plans to install one Nighthawk processor in each module later this year and demonstrate the connection between them.
By Arduino Team 08.19.2026
By Raphi Zadicario, Product Manager and Chief Architect, Lumissil Microsystems 08.17.2026
By Krizelle Paulene Apostol , Software Systems Engineer, Jamila Macagba , Senior Software Systems Engineer, and Maggie Maralit , Software Systems Design Engineering Manager 08.13.2026
Scaling the refrigerator
The need for a larger cryogenic architecture reflects a broader problem: Scaling a quantum computer means finding room not only for qubits, but also for the wiring, connectors, shielding, control hardware, and other infrastructure surrounding them.
Janne Lehtinen
In an exclusive interview with EE Times, Janne Lehtinen, CTO and co-founder of SemiQon, independently identified wiring and control as among the biggest barriers to scaling quantum hardware.
According to Lehtinen, today’s systems can require roughly 2.5 cables per qubit, he said. Those are often coaxial cables accompanied by attenuators, amplifiers, couplers, and connectors. “You are going to end up having millions of connectors in your setup, which are not that reliable,” he said. “There are too many things that can go wrong in this kind of mechanical assembly.”
He said a brute-force extension of that architecture becomes increasingly impractical somewhere in the thousands or tens of thousands of qubits.
IBM’s new system gives the company considerably more physical room to work with and enables more chip-to-chip connections within and between modules. But more space does not eliminate the wiring problem.
An inside view of two of IBM’s modular cryogenic systems, designed to support fault-tolerant quantum computing in a scalable manner (Source: IBM)
Oliver Dial, VP of Quantum Systems at IBM, wrote EE Times in a follow-up exchange that the company believes its current plans can support thousands of qubits per module. And because the wiring bays can be changed without rebuilding the entire cryogenic system, IBM hopes the architecture will also give it room to improve wiring density later.
Longer term, simply making room for more cables may not be enough. Lehtinen argues that large quantum computers will require more control electronics to move into the cryogenic environment and closer to the quantum processor. “You cannot put millions of cables in your fridge,” he said.
SemiQon is developing cryogenic electronics for both silicon spin and superconducting qubits. Silicon spin qubits can operate at somewhat higher temperatures than superconducting qubits, Lehtinen said, giving engineers more cooling power to work with and making it easier to integrate control electronics close to the processor.
SemiQon chiplets and motherboard (Source: SemiQon)
IBM is moving in the same general direction. Dial wrote that IBM is actively researching cryo-CMOS, particularly for its tunable couplers.
That suggests the larger refrigerator is not necessarily the final answer to the wiring problem. It gives IBM room to scale its current architecture while it works on technologies that could eventually reduce the amount of room-temperature wiring required.
Modularity creates an interconnect problem
Splitting a quantum computer across processors creates another challenge: The processors have to communicate closely enough to behave like one machine.
IBM plans to do that with its L-coupler technology, which carries microwave photons across an aluminum superconducting cable between processors. The company said it has demonstrated a two-qubit operation across an L-coupler with 99.3% fidelity and is targeting 99.9%.
Two of IBM’s modular cryogenic systems, side by side, with the doors closed (Source: IBM)
Lehtinen, who had not been briefed on IBM’s announcement, said links between separate processors ultimately need performance close to that of two-qubit gates within a processor. “99.9 sounds very reasonable,” he said, although the precise requirement will depend partly on algorithms and how frequently those links are used.
Dial wrote that the main limitation on IBM’s current L-coupler fidelity is the coherence of the qubits connected to it. The planned Nighthawk demonstration will test not only inter-processor fidelity and state transfer, but also whether linking the processors reduces the performance of the individual processors.
In an exclusive interview with EE Times, Charles Foley, CEO of quantum networking company memQ, said that once processors are connected, interconnects become a different engineering problem from classical networking. “You’re dealing with an uncertain probability that degrades over time,” Foley said.
Quantum links are constrained not simply by bandwidth, he said, but by photon loss, fidelity, coherence time, and the rate at which entanglement can be established.
Charles (Chuck) Foley
For Foley, the longer-term test is whether quantum interconnects can move beyond bespoke demonstrations and become engineered subsystems with defined specifications for parameters such as wavelength, encoding, coherence, loss, and entanglement rate.
That’s also an important distinction for IBM. Making the cryogenic environment modular solves one physical scaling problem, but it puts more pressure on the links between processors to perform reliably enough that the modules behave as one computer.
Reliability becomes the next test
IBM itself acknowledges that scaling the cryogenic environment does not remove the other engineering challenges. At the pre-announcement press event, IBM identified control electronics and system reliability among the biggest remaining obstacles on the road to fault-tolerant computing.
Dial wrote afterward that connectors and room-temperature control electronics concern him particularly because both scale at one or more channels per qubit. That closely echoes Lehtinen’s concern about the sheer number of cables and connectors required by today’s control architectures.
Modularity does not yet mean easy maintenance either. Dial wrote that replacing anything cryogenic in the new system would still require warming and cooling the system as a whole. IBM is studying operational redundancy, including a possible “fail-in-place” strategy.
That’s important to Foley, who pointed out how priorities will change when quantum computers evolve from bespoke research systems into commercial products. “In order to have utility, scale, and commercial feasibility, they have to be serviceable and maintainable,” he said.
IBM has now demonstrated that the cryogenic environment itself can be modularized. The harder question is whether everything inside and between those modules—wiring, controls, processors, connectors, and quantum links—can scale with it.
Read also:
AI Hardware’s Next Frontier Is Integration
GlobalFoundries’ Growth Makes the Case for a U.S. Photonics Buildout
CEA-Leti Pushes Stacking Roadmap as AI Runs into Memory and Power Limits
AI AND BIG DATA, CHIPLET, CRYOGENIC TECHNOLOGIES, QUANTUM, SEMICONDUCTORS