"The quantum industry has been building qubits the wrong way for 20 years." That's not a dig. That's a $50M startup thesis.
image from Gemini Imagen 4
Nobel laureate John Martinis has founded Qolab to challenge the quantum computing industry's dominant Josephson junction fabrication method, arguing that standard liftoff processing cannot scale to a million qubits. The company is replacing liftoff with deposition and etch—the same semiconductor manufacturing processes used at TSMC, Intel, and Samsung—claiming this approach is the only path to fault-tolerant quantum computing. Qolab's March 2025 paper showed window junction qubits via deposition/etch achieved T1 relaxation times up to 57 microseconds, demonstrating process viability though not record performance.
John Martinis has been telling you quantum computing is five to ten years away for roughly fifteen years. He said it when he was at Google. He said it after he left Google in 2020. He said it at the President of Taiwan's office in February. The man won the Nobel Prize in Physics last year for macroscopic quantum tunneling in superconducting circuits, and he is still telling you five to ten years. At some point you have to respect the consistency.
The difference now is that Martinis is not running someone else's lab. He is CTO of Qolab, a startup he founded in 2022 that is making a very specific bet: that the quantum industry has been building qubits the wrong way, and that semiconductor manufacturing know-how — not boutique qubit craft — is the path to a million physical qubits. The 2025 Nobel laureate is backed by a small team including CEO Alan Ho, former head of product at Google Quantum AI, hardware lead Robert McDermott, and quantum engineer Britton Plourde. Their lab is in Madison, Wisconsin. Their office is in Los Angeles. Their impatience is the product.
The core argument is about fabrication. Every superconducting qubit company currently uses liftoff processing to build Josephson junctions — the critical component that defines a qubit's frequency and stability. You spin on a resist, you deposit metal, you wash away the resist and the metal on top of it, and you are left with a tiny metal island. It works. It has worked for twenty years. Martinis says it does not scale. "No one in the semiconductor industry would think about building a big chip with liftoff," he said on the Quantum Computing Report podcast. "When I mentioned this in Asia from someone who used to work at TSMC, he started laughing out loud."
Qolab is replacing liftoff with deposition and etch — standard semiconductor process steps used to pattern every chip made at TSMC, Intel, or Samsung. Deposition puts a thin film of metal on the wafer. Etch removes everything except the pattern you want. It is how the industry has built chips since the 1970s. It is also how Qolab plans to build qubits at wafer scale.
The company published results in March 2025 showing window junction qubits — made via physical vapor deposition and etch — achieved T1 relaxation times up to 57 microseconds, published on arXiv as paper 2503.11010. T1 is how long a qubit holds its state before decay. Fifty-seven microseconds is not a record. It is a signal that the process is viable. The paper's authors include Martinis and McDermott from Qolab alongside researchers from Applied Materials and the University of Wisconsin-Madison, which gives you some sense of the industrial coalition Qolab is building.
That coalition formalized in November 2025. HPE and seven partners launched the Quantum Scaling Alliance with a target that reads like a press release number rather than a specification: five million physical qubits by 2033. Founding members include Applied Materials, Synopsys, Quantum Machines, Riverlane, and 1QBit. Martinis and Masoud Mohseni of HPE Labs co-lead the effort. The QSA plans pilot deployments in summer 2026. Pilot deployments in quantum computing are roughly analogous to pilot lightings in nuclear fusion: the phrase has been used before with varying degrees of accuracy.
But the timeline attack is where Martinis is most himself. Halfway through this decade, he told EE Times, Google is at roughly one hundred qubits. "If you extrapolate out at that rate, I will certainly be dead by the time they build a general-purpose quantum computer." This is a Nobel laureate saying a competitor's approach will not converge in his lifetime. That is not a technical critique. That is a eulogy. He also said the current industry approach "is going to hit a box canyon at some point." Box canyon is a specific term in optimization — a dead end with no exits. Coming from the man whose lab at Google built the first credible demonstration of quantum supremacy, it lands differently than it would from a marketing deck.
Qolab deployed its first superconducting qubit devices at the Institute for Quantum Computing and Chemistry in December 2025. That is real hardware in the field. It is also a long way from a million qubits.
Here is what you should actually take away from this.
Martinis is right that liftoff is a fabrication bottleneck. Every company currently building superconducting chips at more than a few hundred qubits is fighting the same yield problem — not all the qubits on a chip work, and nobody fully understands why. Deposition and etch is not a secret. It is semiconductor industry 101. The question is whether applying that playbook to superconducting qubits preserves the material properties that make those qubits work in the first place. The 57 microsecond T1 result suggests it can. It does not prove it at scale.
The QSA roadmap to five million qubits by 2033 is a consortium document, not a product spec. Pilot deployments in summer 2026 will tell you more than a press release does. But the direction of travel is plausible in a way that the "five to ten years" forecast has not been for most of the past decade. The difference is that this time the argument is rooted in manufacturing infrastructure that already exists, not in the hope that physicists will figure something out.
If you are a founder or an investor, the Martinis timeline fight is worth tracking because it is happening inside the supply chain, not on a stage. The bet is not about qubit count in the near term. It is about whether the fabrication method used to build those qubits imposes a hard ceiling on how far you can scale. Qolab says it does. The rest of the industry is betting it does not have to. One of those bets will age very badly.
The five-to-ten-year forecast, meanwhile, survives unchanged. At this point it has outlasted three presidential administrations, two separate lab departures, and one Nobel Prize. Whatever else you say about John Martinis, his stamina is not in question.
Sources: Quantum Computing Report (podcast with John Martinis) | EE Times (Martinis QSA interview) | arXiv 2503.11010 (window junction T1 results) | BusinessWire (QSA launch announcement) | NobelPrize.org (Martinis 2025 physics award facts)
Story entered the newsroom
Research completed — 15 sources registered. Martinis says 5-10 years for quantum — unchanged for 15 years. Qolab bets deposition/etch over liftoff for qubit fab. QSA targets 5M qubits by 2033. M
Draft (1024 words)
Reporter revised draft (1051 words)
Approved for publication
Headline selected: The Nobel Laureate Is Impatient, and That's the Product
Published
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Quantum Computing · 18h 58m ago · 2 min read
Quantum Computing · 19h 55m ago · 4 min read
