Princeton's New Quantum Institute Is Aimed at the Junction Inside Superconducting Qubits
Through the 1950s at Bell Labs, a thermally grown skin of silicon dioxide moved from laboratory curiosity to the working surface of an industry. Carl Frosch and Lincoln Derick showed that such an oxide could mask a silicon wafer during diffusion; Mohamed Atalla investigated the same grown oxide for its ability to quiet — passivate — the silicon surface that had made devices unreliable; and in 1959 Atalla and Dawon Kahng used it to demonstrate the metal-oxide-semiconductor transistor. The oxide was studied deliberately and at length. It still was not the headline: the transistor was. What the oxide did was supply the surface that planar processing and, later, mass manufacturing were built on — a film nobody outside the fabs could name.
On 25 August 2026 the US National Science Foundation announced more than $290 million across eight Quantum Leap Challenge Institutes. The number is what travels: $290 million, read as a position in a race. Read the awards instead of the headline and a different picture appears, and the sharpest example is an institute whose entire five-year program is aimed at a single circuit element and the metal–oxide interface inside it.
Practical takeaway. This round buys fabrication, error correction and sensing: the layers no supplier can sell you yet, and the layers that decide whether anything sold later works. Qubit counts appear nowhere in the eight award titles.
The claim, and the record it is tested against
The claim is the familiar one: a wealthy state writes a large check and the number is read as a position in a race, denominated in machines and qubits. It is a claim worth testing against the instrument, because the instrument is public. NSF's own release gives the structure — eight institutes at roughly $28 million to $37.5 million each over five years, three newly formed and five renewed, drawing on 36 higher-education institutions across 19 states, with the Department of Energy national laboratories, NIST and the Department of War named as federal collaborators and more than 30 companies attached. Brian Stone, performing the duties of NSF director, framed the round as “a next step for us in understanding the quantum world we live in”. That is the vocabulary of research, and it survives intact into the award titles.
The award list carries the real answer. FTQSAA, Harvard-led with MIT and UCLA, takes fault-tolerant systems, architectures and applications. PRACTIQAL at Yale takes the physics and engineering of practical quantum error correction. HQAN at Illinois takes hybrid architectures and networks; CIQC at Berkeley takes computation; RQS at Maryland takes robust simulation; QuBBE at Chicago takes quantum sensing for biophysics and bioengineering; Q-SEnSE at Colorado takes entangled science and engineering. Every one of those is a description of a bottleneck. None is a description of a machine.
What $27.9 million buys: three layers and an oxide
The clearest case is MARQUIS — Manufacturable and Resilient superconducting Quantum Information Systems — which Princeton will lead with $27.9 million over five years, directed by Nathalie de Leon with Valla Fatemi of Cornell as deputy director, across nine institutions including MIT, Stanford, UC Santa Barbara, Dartmouth, Michigan State, Iowa and NY CREATES.
The institute's subject is one component. A Josephson junction is a three-layer superconductor–insulator–superconductor stack: two metallic superconducting electrodes separated by an oxidized insulating barrier only a few atoms thick, across which Cooper pairs tunnel. It is the element that makes a superconducting qubit a qubit rather than an ordinary circuit. Princeton's stated programme is to find new ways to fabricate that junction and to reinvent its materials system outright — which is a live question because, on Princeton's account, virtually every superconducting qubit in the world is built from the same aluminum and aluminum oxide sandwich, an approach essentially unchanged for more than 25 years. The industry has scaled the number of junctions by orders of magnitude without changing what a junction is made of.
Michel Devoret, a senior investigator on the institute, has described the decades-long effort to make better superconducting qubits as a “graveyard” of ideas for aspiring physicists and engineers. That is an unusual sentence to find in a press release announcing a new institute, and it is the most informative line in it: the difficulty is stated up front, by someone inside the programme, in the same document that announces five years of funding for it. The Princeton group has some standing to try: its own qubits, built on reworked component technology, have been reported to perform 15 times better than leading industry chips.
The seven others answer the same way
Look along the list and the pattern holds. Colorado's Q-SEnSE renewal sits at the top of the round's range at $37.5 million, with 39 researchers across 16 institutions under Jun Ye, and the applications it names are next-generation atomic clocks, the world's most stable lasers, optical frequency combs for air-quality monitoring, and biomedical sensors that detect disease in human breath. Chicago's QuBBE puts quantum sensing inside biology. Neither is a computer. Both are instruments, and instruments are where quantum technology has actually been reaching users — a pattern visible in the market as well, as when the A-share market's first quantum measurement listing turned out to be a scientific-instruments business rather than a computing company.
This is what makes the round read as industrial policy. A state funding qubit counts is buying a number that a vendor can also produce; a state funding junction fabrication, error-correction engineering and clock-grade sensing is buying the parts of the stack that no vendor can currently sell, because they do not yet exist in sellable form. It is the same distinction Quentir has traced before in defense procurement, where the question was whether the money buys the discovery or the copies of it. Here the money goes somewhere else again: into the manufacturing method that would make copies possible at all. And the unglamorous layer keeps turning out to be the binding one: the constraint in IBM's modular cryogenics milestone was not computation either, but how many days it takes to get cold.
Two states, one week, opposite ends of the curve
The comparison that makes this round interesting arrived the same week from the other direction. At its 43rd Foundation Day, India's Centre for Development of Telematics unveiled fourteen indigenous quantum-secure products — quantum key distribution systems over fiber, a near-infrared single-photon detector, and a family of post-quantum encryptors built on the NIST lattice standards — which C-DOT's chief executive, Rajkumar Upadhyay, characterized as production-grade and as having already booked revenue, and which ministers tied publicly to India's 6G ambitions. The trade report records an unveiling and a commercialization claim; it does not establish deployment, and nothing here should be read as saying those fourteen products are installed anywhere.
Set the two side by side. One state spent the week putting a named product line on the table; the other spent it committing five years of public money to a metal-oxide sandwich and a graveyard of failed ideas. Neither is the correct answer, because they are answers to different questions. A product line answers what a country can offer this year using cryptography that already exists. An institute answers what a country might be able to build in a decade if a materials problem yields. The institute is itself perfectly nameable and dated — MARQUIS, announced 25 August 2026, funded for five years. What it cannot name or date is the thing it exists to produce: whether a better junction arrives, and in what form, is exactly the question the award was written to leave open. That undatable quality is the ordinary reason such work sits with public funders rather than with capital that has to report against a horizon, and it is the reason a gap here is hard to close quickly from any other source.
How Quentir Reads It
Three things follow for anyone who tracks this sector for a living.
First, treat the composition of a funding round as better information than its size. The $290 million figure tells you almost nothing; the eight titles tell you what the technical community, under peer review, believes is actually blocking progress. Read together, those titles answer fabrication, error correction, fault tolerance, modular architectures, networking, simulation and sensing — not raw qubit-count expansion alone. For a superconducting-qubit vendor specifically, a roadmap that promises scale without addressing the junction is promising the easy half.
Second, watch the renewal structure. Five of these eight are continuations, which means the program has now produced a public judgment about which phase-one bets earned another five years. That is a rarer artifact than a launch announcement and it is worth reading when the phase-one reports surface.
Third, note where the humane payoff sits. The applications with the shortest path to an ordinary life — a clock that improves navigation and geodesy, a breath test that finds a disease earlier, a sensor small enough to work inside a living cell — sit in the sensing institutes, not the computing ones. That is the part of the quantum program most likely to touch a patient or a household this decade, and it is the part least represented in the discourse about who is winning.
The archive argument is the honest one for a post like this: every award, instrument, filing and prior thread linked here lives in one place, and following the seam between what states fund and what vendors sell is the work the All-access membership exists to make continuous rather than episodic.
What remains open is whether the political system that wrote this check can hold its nerve. A five-year award on a 25-year-old materials problem, described by one of its own senior investigators as a graveyard, is exactly the kind of commitment that looks indefensible in any single budget year and indispensable in retrospect. The oxide that Bell Labs grew took the better part of a decade to become the surface an industry stood on, and it was never the exciting part. The real test of this round does not arrive with a demonstration next year. It arrives when the five-year award reaches its end, if someone then has to make the case for continuing — and if the junction still does not work better.
Sources: U.S. National Science Foundation, “Eight NSF research institutes to propel U.S. quantum science with $290M investment”, 25 August 2026 (more than $290 million total; eight NSF Quantum Leap Challenge Institutes at roughly $28–$37.5 million each over five years; three newly formed and five renewed; 36 higher-education institutions across 19 states; Department of Energy national laboratories, NIST and the Department of War named as federal collaborators; more than 30 company partners; institute names FTQSAA, HQAN, MARQUIS, PRACTIQAL, CIQC, QuBBE, Q-SEnSE and RQS; quotation from Brian Stone, performing the duties of NSF director). Princeton University, “Princeton to lead major federally funded institute to unlock quantum computing challenges”, 25 August 2026 (MARQUIS — Manufacturable and Resilient superconducting Quantum Information Systems; $27.9 million over five years; director Nathalie de Leon of Princeton and deputy director Valla Fatemi of Cornell; nine institutions including Cornell, MIT, UC Santa Barbara, Stanford, Dartmouth, NY CREATES, Michigan State and the University of Iowa; the Josephson junction described in that release as three metallic layers with an oxidized middle layer a few atoms thick — the middle layer is properly an insulating tunnel barrier, so this article states the device as a superconductor–insulator–superconductor stack across which Cooper pairs tunnel; aluminum and aluminum oxide essentially unchanged for more than 25 years; the “graveyard” characterization attributed to Michel Devoret; the report that Princeton-built qubits worked 15 times better than leading industry chips). University of Colorado Boulder, “$37.5M NSF award aims to define new quantum frontiers, take technology from lab to field”, 25 August 2026 (NSF Q-SEnSE renewal at $37.5 million; 39 researchers across 16 institutions under Jun Ye; next-generation atomic clocks, ultra-stable lasers, optical frequency combs for air-quality monitoring, and breath-based biomedical disease detection). Quantum Computing Report, “India’s C-DOT Unveils 14 Indigenous Quantum Products During 43rd Foundation Day”, August 2026 (fourteen quantum-secure products unveiled at C-DOT’s 43rd Foundation Day, including the Q-AKSHAY quantum key distribution systems, a 900–1700 nm single-photon detector, and post-quantum encryptors on the NIST lattice standards; the production-grade and revenue-booked characterization is attributed to C-DOT chief executive Rajkumar Upadhyay and is a company statement, not an independently verified deployment record; ministerial framing alongside India’s 6G program). Computer History Museum, “Development of Oxide Masking” (Carl Frosch and Lincoln Derick at Bell Laboratories; thermally grown silicon dioxide as a diffusion mask on silicon) and “Metal Oxide Semiconductor (MOS) Transistor Demonstrated” (Mohamed Atalla’s work on thermally grown oxide and silicon surface passivation, and the 1959 MOS transistor demonstrated with Dawon Kahng), used as established semiconductor history and not sourced to any 2026 document. Reported timeline claims about India’s national adoption dates circulating in secondary coverage are mutually inconsistent and are deliberately not cited here. Public sources checked 26 August 2026.
Published intelligence, built to inform your own decisions. Published: August 26, 2026.