QCNEWS
The 8 inch wafer line builds the full superconducting chip stack on site — the qubit making layers, from the heart of each quantum bit through 3D chip stacking packaging. It's the manufacturing backbone behind India's quantum push.
QpiAI's new Bengaluru foundry doesn't yet ship a 10,000-qubit quantum processor. It does ship something quieter and more durable: an 8-inch wafer line, on Indian soil, that runs the full superconducting QPU stack from electron-beam lithography through Josephson-junction fabrication to 3D flip-chip assembly and cryogenic packaging (Quantum Computing Report). Phase 2 of the company's 70,000-square-foot Jakkur R&D center is open. The 10,000-qubit number is a 2027 plan, not a product.
The distinction matters because it is the central reader-payoff gap in the press coverage. Today's fabricated chips at the Bengaluru line are small. QVidya is an 8-qubit transmon, a superconducting circuit where pairs of electrons flow without resistance, used for early algorithm work. Yukti is a 9-qubit fluxonium, a related superconducting design with higher coherence but harder to scale, and it is being used as a surface-code and logical-qubit test bed. Indus is a 25-qubit transmon integrated with hybrid classical HPC, and was the chip that won QpiAI a slot under India's National Quantum Mission. Kaveri is a 64-qubit transmon with proprietary low-loss flip-chip interconnects. The next production milestone is Ganges, at 128 physical qubits (QpiAI corporate site).
That 128-qubit ceiling is the foundry's current flip-chip cap, not a benchmark against the global field. IBM's Heron production chip runs at 1,121 qubits. Google's Willow and Quantinuum's H-series push the higher-fidelity end. QpiAI is not claiming to have caught any of them. The claim is narrower and more specific: India now has a domestic Class 100 and Class 1,000 cleanroom line doing the entire superconducting device lifecycle in one place, with R&D pathways into photonic and semiconductor spin qubits alongside the production superconducting work, and TechCrunch has framed QpiAI as India's chosen vehicle for a global push.
The capital path matches the scope. QpiAI reports $20 million to $25 million in capex to date, with a planned $10 million to $15 million Phase 3 expansion to support the 10,000-qubit target by 2027 (QpiAI corporate site). Businessworld's separate $35 million figure sits inside the same band once currency and timing differences are reconciled. The combined roughly $30 million to $40 million range is a credible packaging-and-integration investment for a superconducting startup. It is not a leading-edge silicon fab budget. The right comparison point is PsiQuantum's much larger capex or IBM's longer multi-decade build-out, not a TSMC-class megafab.
The downstream play is the Quantum Supremacy Centre, QpiAI's planned hybrid data center linking fault-tolerant QPUs with AI clusters. Four international sites are on the roadmap, with a 10-acre India campus referenced (QpiAI corporate site). The "supremacy" framing is the company's marketing term; no published quantum-advantage benchmark has been tied to a QSC site yet, and the data centers are not a deployed service. What is on the record is government backing through the National Quantum Mission startup cohort and a DST/PIB press release for the QpiAI-Indus program.
The substantive news is the manufacturing backbone, not the qubit count. A 128-qubit fab cap with a 2027 roadmap to 10,000 is concrete enough to pressure-test: Ganges has to land first, Phase 3 capex has to close, and the foundry has to keep yield and fidelity competitive against global peers. The reader should hold the foundry tier and the qubit ladder in separate hands and let the 2027 deadline do the work.