Quantum's TSMC Moment Comes With an Ownership Question

Within 12 days this September, four announcements redrew the map of quantum hardware. Quantum Foundry Copenhagen unveiled plans for a 5,300-square-metre chip facility, GlobalFoundries finalised a US CHIPS award of up to US$375 million for its quantum foundry business, SkyWater signed a multi-year superconducting manufacturing agreement with Qolab, and Anderon, IBM’s new quantum foundry, finalised a US$1 billion CHIPS award, matched by a further US$1 billion from its parent. Each of these announcements concerns who will manufacture the chips quantum computers depend on, a question the industry has increasingly begun to ask in recent times.

Quantum chips have historically come from bespoke cleanrooms run by the teams that designed them, where success meant a device that worked. Scaling quantum processing units (QPUs) means producing wafers repeatedly, at predictable quality and acceptable yields, within useful timescales. The classical semiconductor industry resolved a version of this problem when specialist foundries such as TSMC let designers concentrate on architecture while manufacturers pooled capital and process expertise. Quantum is unlikely to follow the same path, at least for now, since its qubit types demand different processes, fabrication choices shape performance directly, volumes remain small, and two of the most capable new foundries belong to companies that also build quantum computers. A layered arrangement looks more likely, with qubit-critical steps kept in-house and specialist foundries serving particular modalities, which makes who fabricates, tests, and qualifies the devices as strategically vital as which architecture wins.

Quantum Foundries Are Taking Several Shapes at Once

Several ownership models are emerging simultaneously, complicating any tidy narrative about a single 'TSMC for quantum'. GlobalFoundries comes from conventional chipmaking with no quantum computer of its own. It is developing platforms across superconducting, trapped-ion, photonic, topological, and spin qubits, which lets it grow with the market without picking a winner. Anderon runs a 300-millimetre pure-play quantum foundry, while SkyWater, acquired by IonQ in July, continues as a merchant supplier. In Europe, Quantum Foundry Copenhagen, owned by the Novo Nordisk Foundation, will sell wafer fabrication to vendors worldwide from 2027. Meanwhile, Imec’s €50 million (US$58 million) European SPINS pilot line is developing multi-project wafers (single wafers carrying several customers' designs) for spin qubit developers. Each addresses the same problem, that few hardware companies can efficiently build expensive fabs and specialist know-how alone, which raises the question which structures customers will trust with strategically important manufacturing.

When Your Foundry Belongs to Another Quantum Company

Anderon demonstrates the tension between serving a parent and supplying a market. The foundry promises customers it will never compete with their products, echoing the commitment on which TSMC built its foundry reputation. Yet Anderon is owned by IBM, a leading superconducting quantum computer developer, and IBM Research Director Jay Gambetta has said IBM’s quantum roadmap requires the scalable manufacturing foundation Anderon provides. SkyWater presents a related case, having signed Qolab as lead customer for its new SC250 superconducting platform even though its parent, IonQ, builds trapped-ion machines. Nothing suggests either foundry has or would compromise a customer. However, merchant status does not guarantee ownership independence, which obliges the industry to define what neutrality truly requires in practice, from IP protection and information separation to capacity guarantees and equal access.

Provided neutrality can be defined and demonstrated, the case for using these foundries is considerable, since one of the scarcest inputs in quantum fabrication is know-how. Specialist tools, proven processes, cryogenic testing, and engineers who deliver repeatable yield remain hard to find. Defects and process variability during qubit manufacturing can limit coherence (how long a qubit retains its information), so process control translates directly into performance. Large foundries can spread those capabilities across customers and standardise processes more quickly over time, as SkyWater's dedicated tool upgrades and locked-in high volume production costs for Qolab illustrate. That does not produce inevitable winners, given that different quantum architectures demand different manufacturing processes. But Anderon or SkyWater could become default suppliers in process niches where credible alternatives remain scarce. Such gaps are real, as a government-commissioned 2024 Royal Academy of Engineering review found that the UK then lacked commercial superconducting fabrication capability for quantum applications. The UK has since begun responding: the Engineering and Physical Sciences Research Council (EPSRC) has awarded £1.5 million to a University of Glasgow project with Quantcore, a Glasgow university spinout, to develop niobium circuit fabrication at its existing nanofabrication centre.

Ownership will matter less where protections around customer IP, process data, and capacity allocation are strong, and considerably more where they are less certain.

Why Some Developers Keep Fabrication In-House

Foundries only become useful if customers trust them with critical manufacturing, which depends on IP protection, process isolation, turnaround times, and run-to-run reproducibility. Keeping fabrication in-house carries its own costs, tying up capital, specialist staff, and management attention in facilities that smaller hardware companies struggle to keep fully utilised, and committing them to process development that a foundry could spread across many customers. Some developers have nonetheless decided that outsourcing is not yet worth it. Silicon Quantum Computing, a Firgun Ventures portfolio company, fabricates its atomically precise chips in Sydney, Australia and can design, produce, and test a new one in under a week, a feedback loop the company links directly to device quality. Rigetti makes its superconducting chips at its own Fab-1, and its US$100 million CHIPS award, finalised in September, partly funds fabrication capability for high-connectivity architectures. For both, the calculation weighs a fab's cost against the value of retaining process knowledge and controlling fabrication closely enough to optimise the qubits themselves.

That calculation will shift as external foundries improve, though in-house fabrication should stay valuable wherever manufacturing know-how is tightly coupled to technical advantage. Portability matters too, as few developers want to depend on a process that cannot be reproduced elsewhere. Architectural diversity complicates the comparison with classical chips, since the various quantum hardware systems need different processes, pointing towards shared facilities hosting several specialised platforms. Hybrid arrangements are already appearing. SQC retains its atomically precise qubit fabrication in-house while partnering with SkyWater to supply complementary superconducting resonators and tailored silicon wafers for integrating its processors into larger quantum-classical systems. With foundry economics hinging on utilisation as much as capability, the resulting industry may resemble the classical fabless model in some segments and look quite different elsewhere.

Quantum Manufacturing Becomes an Investable Layer

Foundries also shift where value sits, as a company no longer needs to design the winning computer to share in quantum's growth. Manufacturing, advanced packaging, cryogenic testing, and qualification can each support businesses serving several developers. These businesses offer exposure to rising hardware activity without a binary architecture bet, provided they achieve the volume, repeatability, and perceived neutrality rivals require. Public and philanthropic capital reinforces that notion, as the Anderon and GlobalFoundries awards belong to a roughly US$2 billion CHIPS quantum package across nine companies, while Copenhagen and SPINS show Europe treating manufacturing as strategic infrastructure. The likely outcome is a distributed network of foundries rather than one global champion, giving investors a more resilient, if fragmented, supply base to back.

From Building the Best Qubit to Manufacturing It at Scale

Quantum computing has spent years debating who will build the best qubit, and its next phase has begun to ask who can manufacture good qubits repeatedly, economically, for a large number of customers. The manufacturing layer remains at an early stage, with Copenhagen's facility not yet live, while Anderon and SkyWater test whether developer-owned foundries can become trusted infrastructure for everyone else. How that test resolves will set the terms on which every other hardware developer decides whether to build, buy, or share its manufacturing. The quantum supply chain will ultimately be shaped by who owns the factory and who is willing to use it.