IBM joined and cooled two modular cryogenic systems into a single operating environment for the first time, a physical milestone on the path to its fault-tolerant Quantum Starling system targeted for 2029. The demonstration support’s Futurum’s quantum foundry report thesis that refrigeration capacity, alongside packaging, is the gating input for fault-tolerant quantum computing. IBM is converting that input into a manufactured product that it controls.
What Is Covered in This Article:
- IBM’s first two modular cryogenic systems, joined and cooled below 15 millikelvin
- Cell specifications: 0.53 square meters of wiring area, 2.75 cubic meters of vacuum volume, 4 kelvin in under 5 days
- Nighthawk installation later in 2026 and the 1,000-programmable-qubit linkage target for 2027
- Futurum’s quantum foundry report thesis: cryogenics and packaging as the binding constraint on quantum scaling
- Competitive positioning versus Google, Quantinuum, IonQ, PsiQuantum, and Microsoft
The News: IBM announced on August 19 that it has joined and cooled two modular cryogenic systems into a single operating environment, a milestone toward IBM Quantum Starling, the fault-tolerant quantum computer the company plans to deliver in 2029. The combined structure, assembled in Poughkeepsie, New York, stands more than 8 feet tall and 8 feet wide and reached temperatures below 15 millikelvin, which IBM describes as more than 180 times colder than deep space. The system cooled to 4 kelvin in under 5 days, and each box-shaped aluminum module offers up to 12 times more wiring space inside its vacuum enclosure than IBM’s most widely deployed quantum systems. IBM plans to install IBM Quantum Nighthawk processors in the modules later in 2026 and to link multiple processors with at least 1,000 programmable qubits in 2027.
“Bringing fault-tolerant quantum computers to industries depends on several fundamental advances,” said Jay Gambetta, Director of IBM Research and IBM Fellow.
IBM Links Modular Cryogenic Systems. Is the Fridge Quantum’s Real Bottleneck?
Analyst Take: IBM’s connection of its first modular cryogenic systems delivers the evidence Futurum asked for in its June quantum foundry report, which argued that the true determinant of quantum market leadership will be securing capacity in the downstream cryogenic and packaging supply chain rather than wafer fabrication alone. Every superconducting roadmap through the end of the decade runs through a dilution refrigerator, and until this week those refrigerators were bespoke cylindrical instruments built by a two-vendor supply base. IBM’s response is to redesign the refrigerator as a repeatable manufactured cell with a box-shaped aluminum module with 0.53 square meters of wiring area and 2.75 cubic meters of vacuum volume that connects side-by-side with its neighbors and shares a single cold environment. Two of those cells now operate together below 15 millikelvin in Poughkeepsie. IBM now treats cryogenics the way foundries treat lithography cells, as standardized production infrastructure whose replication rate sets the ceiling on everything downstream. The company that spent 2026 assembling a $10 billion roadmap, a dedicated 300mm foundry, and $1 billion in CHIPS Act incentives has extended vertical integration into the coldest link of the chain.
Modular Cryogenic Systems Convert Quantum’s Scarcest Input Into a Product Line
Futurum’s foundry report quantified why refrigeration deserves this attention: dilution refrigerator lead times run 6 to 9 months at Bluefors and Oxford Instruments, and the helium-3 that powers their coldest stage costs $1,900 to $2,600 per liter while representing roughly 0.0001% of the world’s helium supply. Those constraints bind every superconducting program at once, because the supplier base is shared. IBM knows this dependency firsthand. IBM Quantum System Two runs on the KIDE cryogenic platform co-developed with Bluefors, and in 2022 the company built Goldeneye, a 1.7-cubic-meter experimental “super-fridge”, to test whether one giant cryostat could house a fault-tolerant machine.
The new architecture reads as the verdict on both experiments. Each cell exceeds Goldeneye’s entire experimental volume. The design is meant for repeat manufacture, with capacity growing by adding cells. IBM reports the joined modules reached 4 kelvin in under 5 days, a figure that, if it survives production replication, makes cryogenic commissioning a schedulable step in system assembly. A quantum foundry that fabricates chips 30 times faster on 300mm wafers, per IBM’s internal estimates, would simply pile up inventory without refrigeration capacity to install it in. Modular cells close that loop.
From Refrigeration to Computation
IBM cooled two connected cryogenic cells without quantum processors installed, so this milestone validates plumbing and thermal engineering while leaving every computational question open. IBM’s 12 times wiring figure compares against IBM’s own installed base rather than a competitor’s system. The harder physics arrives with the L-couplers, the meter-scale chip-to-chip interconnects IBM first demonstrated in 2024, which must move quantum information between processors in separate cells at fidelities compatible with error correction. IBM’s Nighthawk processors, at 120 qubits with 218 tunable couplers, install later in 2026, the 2027 target links multiple processors with at least 1,000 programmable qubits, and the Cockatoo chip scheduled for 2027 is tasked with proving entanglement between modules. Further, four custom control ASICs must converge around 2029 for power consumption to stay manageable at up to 3 megawatts per system. Starling remains a 2029 delivery with three sequential proof points still ahead of it, and a cold empty box, however impressive, generates no circuit revenue. IBM’s ecosystem gives it room to absorb schedule risk, with $1.1 billion in quantum contracts signed since 2017 and Qiskit used by nearly 70% of quantum developers, yet the fault tolerance date now depends on interconnect fidelity data that does not exist in public.
Rivals Answer the Cold Differently
Mapped against the field, the announcement sharpens a divide that runs through modality choice. Google shares IBM’s millikelvin problem: its Willow chip, at 105 qubits, demonstrated below-threshold error correction with a distance-7 logical error rate of 1.43×10^-3 per cycle, and scaling that result to useful logical qubit counts will demand exactly the kind of cryogenic real estate IBM just prototyped. Google has published no comparable refrigeration architecture, which leaves it a customer of the same Bluefors and Oxford Instruments supply base whose lead times Futurum flagged. The other leaders opted out of the problem. Quantinuum’s trapped ion roadmap targets the fully fault-tolerant Apollo system by decade’s end, and it does not run through millikelvin dilution refrigerators. IonQ operates its trapped ions near room temperature and spent $1.8 billion acquiring SkyWater to secure fabrication capacity; refrigeration barely figures in its architecture. PsiQuantum’s photonic approach works at roughly 4 kelvin, a regime industrial cryoplants already serve without helium-3 dilution stages. Microsoft’s Majorana 2 remains an early-stage substrate validation. A working cell design removes the scaling penalty superconducting qubits pay for their operating temperature and weakens the strongest structural argument for the warmer modalities. Stumble on cross-module fidelity, though, and the same capital spending becomes the field’s most expensive cautionary tale. Futurum assesses that IBM is the only vendor currently treating refrigeration as a first-class manufactured product, and in a supply chain where fridges arrive slower than chips, that position can compound.
What to Watch:
- Whether Nighthawk processors install in the joined modules later in 2026
- Whether IBM publishes two-qubit gate fidelities across L-couplers spanning separate cells
- Whether IBM links multiple processors with at least 1,000 programmable qubits in 2027, and whether Cockatoo demonstrates entanglement between modules
- Whether IBM discloses cell counts and cryogenic capacity plans for Starling as the Anderon foundry ramps
- Whether Google, Rigetti, or other superconducting vendors respond with modular cryogenic designs of their own
See the complete announcement of the modular cryogenic milestone on the IBM newsroom.
Declaration of generative AI and AI-assisted technologies in the writing process: This content has been generated with the support of artificial intelligence technologies. Due to the fast pace of content creation and the continuous evolution of data and information, The Futurum Group and its analysts strive to ensure the accuracy and factual integrity of the information presented. However, the opinions and interpretations expressed in this content reflect those of the individual author/analyst. The Futurum Group makes no guarantees regarding the completeness, accuracy, or reliability of any information contained herein. Readers are encouraged to verify facts independently and consult relevant sources for further clarification.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Read the full Futurum Group Disclosure.
Other Insights From Futurum:
IBM and Together AI: Did IBM Cloud Just Become a Neocloud?
The Orchestrator: Who’s Conducting Your Enterprise?
Building A Digital-Native Airline: How IBM Powered Riyadh Air’s Launch
Author Information
Brendan is Research Director, Semiconductors, Supply Chain, and Emerging Tech. He advises clients on strategic initiatives and leads the Futurum Semiconductors Practice. He is an experienced tech industry analyst who has guided tech leaders in identifying market opportunities spanning edge processors, generative AI applications, and hyperscale data centers.
Before joining Futurum, Brendan consulted with global AI leaders and served as a Senior Analyst in Emerging Technology Research at PitchBook. At PitchBook, he developed market intelligence tools for AI, highlighted by one of the industry’s most comprehensive AI semiconductor market landscapes encompassing both public and private companies. He has advised Fortune 100 tech giants, growth-stage innovators, global investors, and leading market research firms. Before PitchBook, he led research teams in tech investment banking and market research.
Brendan is based in Seattle, Washington. He has a Bachelor of Arts Degree from Amherst College.
