Princeton to Lead $27.9M NSF MARQUIS Institute: Solving Superconducting Quantum Computing's Materials and Manufacturing Bottleneck
> On August 25, 2026, the US National Science Foundation (NSF) announced a new round of Quantum Leap Challenge Institutes totaling $290 million. Princeton University will lead the MARQUIS institute, receiving $27.9 million over five years, targeting the core bottleneck of superconducting quantum processor manufacturing: the Josephson junction. This is not another quantum-advantage paper, but a manufacturing paradigm shift from "hand-crafted in the lab" to "mass-produced in a semiconductor fab."
---
25 Years of Materials Stagnation: The Josephson Junction Problem
To understand MARQUIS's mission, you first need to know the Josephson junction—a sandwich structure only a few atomic layers thick that is the heart of nearly all superconducting qubits.
A Josephson junction stacks three layers: superconductors on both sides with an insulating oxide layer just a few atoms thick in between. Pairs of electrons (Cooper pairs) quantum-tunnel through this insulating barrier, creating a nonlinear inductive element that can be precisely controlled to process quantum information. Its quality directly determines qubit coherence times, fidelity, and scalability.
The problem: for the past 25 years, the entire industry has fabricated Josephson junctions from essentially the same materials technology—aluminum plus aluminum oxide, patterned with polymer stencil masks. This approach works well enough for academic prototypes and small-scale systems, but scaling from dozens to thousands or tens of thousands of qubits requires reinventing the most fundamental manufacturing element.
Princeton professor of electrical and computer engineering and MARQUIS director Nathalie de Leon was blunt: "The community has used essentially the same materials technology for about a quarter of a century. That technology works fine for experimental prototypes and small-scale systems. But to build quantum computers at scientifically useful scale, the most basic element must be reinvented."
---
MARQUIS's Interdisciplinary Scope: When Quantum Physics Meets the Semiconductor Fab
What makes MARQUIS (Manufacturable and Resilient superconducting Quantum Information Systems) unique is that it is not a pure physics institute, but a manufacturing-focused consortium spanning materials science, quantum devices, and semiconductor processing.
Nine research institutions—Princeton (lead, Nathalie de Leon), Cornell (deputy lead Valla Fatemi), MIT, UC Santa Barbara, Stanford, Dartmouth, NY Creates, Michigan State, and University of Iowa—contribute more than twenty laboratories covering the full pipeline from fundamental materials characterization to wafer-scale process integration. The advisory board brings together Google Quantum AI, NVIDIA, Applied Materials, Oxford Instruments, Bluefors, Imec, and MIT Lincoln Laboratory—one of the deepest academic-industry marriages in quantum manufacturing to date.
Cornell physicist Valla Fatemi's team has already achieved key related advances: using krypton gas to deposit tantalum on silicon at 200°C, making the process more compatible with semiconductor manufacturing; and developing a resist-free method for fabricating Josephson junctions using etched silicon trenches, reducing contamination and opening up new materials and processes.
---
Why Must the Josephson Junction Be Reinvented?
The problem with conventional aluminum/aluminum-oxide junctions isn't that they don't work—it's that they can't scale.
First, materials loss. Aluminum oxide dielectric loss limits qubit coherence times. Last year, de Leon's team showed that reimagining junction materials can improve qubit performance to 15 times that of leading industrial chips—the biggest advance in the field in over a decade.
Second, manufacturing compatibility. Superconducting qubits are mostly "hand-crafted" in academic cleanrooms using polymer stencil masks incompatible with mainstream CMOS production lines. Thousands-of-qubit processors require semiconductor-compatible deposition, etching, and patterning processes.
Third, repeatability and yield. Hand-crafted qubits show large parameter variation, making consistency across large arrays difficult. The semiconductor industry spent half a century solving yield and repeatability; quantum computing must graft on that mature expertise.
---
Mid-Scale Test Platforms: A Bridge from Academic Samples to Industrial Production
Another key MARQUIS task is developing test and validation methods for mid-scale quantum processors—a critical step between academic lab-scale systems and future large processors.
These testbeds will standardize research across labs and let experts from other fields (such as semiconductor manufacturing engineers) meaningfully contribute to core challenges. de Leon noted: "In the semiconductor industry, a lot of the best knowledge is hidden behind curtains. The literature is vast, and it's hard for us to sort through it ourselves. So getting a few key experts on board who know the right signposts and ways of thinking is really crucial."
---
NSF's Big Picture: $290 Million, Eight Institutes
MARQUIS is one of eight NSF Quantum Leap Challenge Institutes. This round invests $290 million total, with each institute receiving roughly $28–37 million over five years, spanning quantum computing, sensing, and communication:
- MARQUIS (Princeton-led): superconducting quantum processor manufacturing
- HQAN: hybrid quantum architectures and networks, interconnecting different qubit types
- PRACTIQAL: practical quantum error correction
- CIQC: new quantum algorithms and hardware architectures
- QuBBE: quantum sensing for biophysics and bioengineering, quantum nanoprobes, living-cell measurements
- Q-SEnSE: entanglement science and engineering, precision sensing, atomic clocks
- RQS: robust quantum simulation
- FTQSAA: fault-tolerant quantum system architectures
---
From "Graveyard" to "Factory": Dawn of Superconducting Qubit Industrialization
Nobel laureate Michel Devoret—who first showed how Josephson junctions could create "artificial atoms"—once described superconducting qubit improvement efforts as a "graveyard" for aspiring physicists and engineers. Too many promising ideas have foundered on materials defects, fabrication variability, and engineering complexity.
MARQUIS's launch signals that the US quantum computing strategy is shifting from "chasing qubit counts" to "building manufacturing foundations." It's a late but necessary pivot: without manufacturability, even the best physics designs never leave the lab.
Globally, the quantum computing race has entered deep water. China continues breakthroughs in photonic (Jiuzhang series) and superconducting (Zuchongzhi series) routes; Japan just commissioned its first full-stack neutral-atom quantum computer, Shunkai; IBM and Google keep setting logic-qubit records on the superconducting path. MARQUIS's arrival means the US is attempting to build long-term advantage in quantum manufacturing—a chokepoint technology—not through any single Nature paper, but through systematic engineering capability across materials science, semiconductor processes, and quantum devices.
> Key numbers: $27.9M NSF grant / 5 years (Princeton MARQUIS, 2026-08-25); $290M total NSF investment across eight institutes; nine institutions / 20+ labs participating; de Leon team's 2025 result: 15x qubit performance improvement; aluminum + aluminum-oxide junctions in use for 25+ years; advisory board includes Google Quantum AI / NVIDIA / Applied Materials / Imec.
---
*References: Princeton University press release / NSF announcement / Cornell National Tribune / Quantum Zeitgeist / Quantum Spectator / Reuters / de Leon group 2025 paper*