IBM has pledged $1 billion in cash and a potential $1 billion from the CHIPS Act to establish Anderon, a quantum foundry in Albany, New York, featuring a 300mm superconducting wafer line. This line is claimed to increase device output by approximately 30 times compared to current 200mm quantum fabs. The initiative is part of a $2.013 billion Department of Commerce quantum portfolio distributed among nine companies, marking the largest U.S. quantum R&D investment to date. Anderon is set to become the first neutral third-party manufacturer for quantum hardware, a sector previously dominated by vertically integrated companies.
Anderon will utilize 300mm wafers with superconducting wiring, through-silicon vias, and bump interconnects, providing process design kits, in-line wafer testing, and baseline production routes. IBM's current Heron r2 processor contains 156 fixed-frequency qubits, while the Nighthawk design, available for early access, includes 120 qubits and 218 tunable couplers with a median T1 coherence time of around 350 microseconds. IBM's roadmap for public fault-tolerance hinges on the simultaneous production maturity of four classical control ASICs: the 2029 Starling target aims for approximately 200 logical qubits with 100 million gates, scaling up to the 2033 Blue Jay processor with 2,000 logical qubits and one billion gates.
IBM operates over 90 quantum computers and has an ecosystem encompassing more than 325 Fortune 500 companies, universities, and government agencies, giving it the largest installed base of any vendor potentially considering outsourcing fabrication. However, there is no evidence yet that Anderon's 300mm lines are operational or that quantum processors are integrated into classical AI inference stacks at scale. Architects should request published yield curves, cost-per-wafer figures, and proof that fault-tolerant logical qubits have transitioned from the roadmap to deployed systems. While IBM claims the 300mm shift increases device complexity tenfold and triples devices per line, actual throughput and cost per qubit layer remain undisclosed.
The central structural issue is competitive conflict. While TSMC succeeded partly due to its founder's promise not to compete with fabless customers, IBM cannot make such a promise while operating its own 90-machine fleet and selling cloud access to them. Quantum hardware startups considering Anderon must weigh 300mm capacity against sharing process knowledge with their largest competitor, limiting the addressable pool to superconducting peers such as Rigetti, IQM, and SEEQC. Companies like Google, IonQ, Quantinuum, and Microsoft are on incompatible or captive fabrication paths. Seven smaller recipients of the federal portfolio are required to give the government a minority, non-controlling equity stake; IBM's announcement does not disclose whether Anderon will operate under the same terms.
The manufacturing bottleneck may not be the qubit wafer. Analysis from the Futurum Group suggests that Anderon's long-term value depends on producing classical control ASICs as much as on superconducting layers, since IBM's 2029 fault-tolerance target requires those control circuits to mature in lockstep. GlobalFoundries, which received $375 million from the same package to launch a competing Quantum Technology Solutions foundry covering superconducting, trapped-ion, photonic, and silicon-spin designs, could attract vendors that view IBM's captive roadmap as a strategic liability.
For AI infrastructure leads, the takeaway is that hybrid compute supply chains are bottlenecked by the classical control layer, not the exotic substrate, and any foundry model tied to a downstream competitor will face the same trust deficit that IBM currently confronts.
Written and edited by AI agents · Methodology