Nobel Prize in Physics Winner: The Quantum Leap That Changed Everything - John Martinis
Summary
Martinis’s Nobel-linked experiment became valuable not merely as proof, but as the foundation for a superconducting-quantum-computing field. His 1985/86 circuit showed that a macroscopic electrical object containing billions of electrons could exhibit discrete quantum behavior; Google later scaled the approach to a 53-qubit “quantum supremacy” demonstration in 2019. His test for genuine breakthroughs: whether they lead to “other experiments and other papers and other inventions.”
The commercial gap is brutally numerical: today’s controllable superconducting systems have roughly 50–100 qubits, while a general-purpose machine may require about 1 million. Qubits may retain their state for perhaps 100, 1,000 or a few thousand operations before noise erases their memory, making ongoing error correction essential. “A million is a good round number,” Martinis says; useful machines must become both much bigger and much cleaner.
Martinis’s company aims to do something meaningful in roughly 8–10 years, while conceding that the industry has been forecasting ten years “for a while now.” Existing machines run real algorithms and publishable experiments, but claims of computational difficulty do not yet make them economically useful. His blunt assessment—“there’s more hype than reality”—keeps the forecast conditional on solving fabrication and noise bottlenecks.
The scaling thesis is industrial fabrication, not another laboratory-only qubit milestone. Martinis’s company is working with Applied Materials, Synopsys, Hewlett Packard Enterprise and theory startups to move toward modern 300 mm semiconductor processes, aiming for GPU-like cost and quality. If that manufacturing jump works, he believes “we can scale up very rapidly.”
Martinis sees Chinese teams as technically near-parity and worries that Chinese authorities may restrict publication until comparable Western results are public. He says Chinese teams reproducing Google’s supremacy work “know what they’re doing”; after Google published a materially improved result, China quickly indicated similar performance. His proposed manufacturing advantage is access to advanced fabrication tools and processes that, he says, are not available in China, including certain 300 mm equipment.
AI may help quantum computing, but Martinis rejects it as a substitute for clean hardware and clear control. He sees roles in modeling, error-correction decoding and hybrid quantum-AI algorithms, yet remains “a little bit old school”: noisy fabrication cannot be rescued into great performance by software. The near-term focus is process engineering, materials, controls and error correction—not an assumed AI shortcut.
Deep dive
1. A garage builder turned Leggett’s question into an experiment
Martinis traces his experimental instinct to San Pedro, where his fireman father lacked a high-school education but was “a very smart person” constantly building garage projects. Physics clicked because the hands-on world had mathematics behind it; at UC Berkeley, adviser John Clarke introduced him to quantum behavior in electrical devices.
Anthony Leggett’s load-bearing question was, “Do macroscopic objects behave quantum mechanically?” The proposed object was an electrical circuit containing billions of electrons and atoms, whose collective current and voltage could reveal whether quantum laws extended beyond the microscopic constituents for which the theory had been developed.
Friedberg framed microscopic objects as probability distributions rather than points following predetermined paths. Martinis sharpened the idea with the atom: classical attraction would collapse electron and nucleus together, but the electron is “fuzzy,” an extended wave whose allowed standing-wave frequencies give atoms size and characteristic spectra.
Leggett’s point about Schrödinger’s-cat paradox was empirical: treating the cat as potentially alive and dead assumes that a macroscopic object can occupy a quantum-superposition state, yet no experiment had established that. Martinis saw an unusually deep thesis problem—test that assumption directly in a circuit rather than debate the paradox philosophically.
2. A Josephson circuit made macroscopic quantum behavior measurable
Quantum tunneling was the original proposed test. A particle’s wave mostly reflects from a barrier, but a small part can appear across it—“just like walking through walls.” The effect already matters in memory circuits, where electrons leak through insulators only 10–20 atoms thick, and in magnetic memories that depend on tunnel junctions.
A human body will effectively never tunnel through a wall because too many atoms would have to be in the right positions and momenta at once. The electrical circuit changed the odds: operating near 5 GHz meant it could attempt the transition “five billion times a second,” giving an otherwise rare macroscopic event enough opportunities to become observable.
In a superconductor, Martinis explained, “all the electrons condense into one state” and move collectively without random scattering. Friedberg’s childhood demonstration—cooling a superconducting disc with liquid nitrogen and levitating a magnet—made that persistence tangible; MRI magnets similarly sustain a superconducting current and magnetic field for very long periods.
The Josephson junction placed an insulating barrier between two superconductors, allowing Cooper pairs to tunnel together without loss. Combined with a capacitor, its nonlinear kinetic inductance formed an LC resonator. The decisive observation was discrete energy frequencies, analogous to a sodium lamp’s specific yellow line—direct evidence that a macroscopic circuit obeyed quantum mechanics.
3. One strange result became a computing architecture
The work appeared in Physical Review Letters around 1985 or 1986 and drew substantial attention, including a Scientific American article. Yet Martinis’s retrospective is unsentimental: proving macroscopic quantum mechanics was noteworthy, but observers could still ask, “What is it good for? What are you going to do?”
At a UC Santa Barbara conference, Richard Feynman’s closing talk proposed using quantum mechanics for computation. Martinis admits he did not catch everything, but the crowd mobbing Feynman convinced him this was the “most interesting fundamental question.” Peter Shor’s factoring algorithm in the early 1990s later supplied a concrete problem.
Martinis continued the device physics with Michel Devoret’s group in France, worked at NIST while David Wineland’s group was just down the hall, and conducted experiments involving electron counting and metrology. He then went “all in” during the late 1990s as theory matured and U.S. government funding appeared. At UCSB, his laboratory progressed from basic devices to five- and nine-qubit computers.
Google offered the money and team continuity that academia struggled to sustain for a complex machine. In 2019, Martinis’s group published a 53-qubit quantum-supremacy experiment whose mathematical output was much harder to emulate classically. It was “not practical,” he stresses, but demonstrated the power of a quantum computer and that it worked at meaningful scale.
4. Today’s machines compute, but usefulness remains orders of magnitude away
A superconducting qubit remains conceptually close to the original experiment: a Josephson-junction inductance, a capacitor and an oscillator near 5 GHz. Microwave pulses change its state, readout circuitry measures it, and capacitive coupling links many such elements into an array.
Current superconducting systems have roughly 50–100 fully controlled qubits and can run genuine, complicated algorithms. Neutral atoms are the promising “newcomer on the block,” with big arrays already demonstrated, though Martinis says their gates still need much better control.
The limiting factor is noise. Depending on quality, qubits might complete roughly 100, 1,000 or a few thousand operations before losing their memory. Martinis compares them with dynamic RAM that must be refreshed—but here the refresh is quantum error correction, pushing a general-purpose system toward approximately 1 million qubits.
Asked for a useful-computing timeline, Martinis said his company—and many other groups—aims to do something in the next 8–10 years, while immediately preserving the uncertainty: people have predicted ten years “for a while now.” Present systems are good enough for scientific papers and experimentation, but “they aren’t really big enough to be useful yet.”
5. Fabrication, not AI, is Martinis’s preferred scaling lever
Asked whether AI is accelerating quantum engineering as it might fusion or materials science, Martinis’s answer was deliberately restrained: “There may be” useful modeling applications, error-correction decoding and hybrid quantum-AI algorithms. But poorly fabricated hardware with unclear controls will not deliver great performance, however sophisticated the software becomes.
His company’s thesis is that current fabrication methods contain the key technology bottlenecks. After successive process generations—from the simple 1985 devices through more sophisticated work around 2000 and the 2019 supremacy chip—the next jump is to modern 300 mm semiconductor tools, standard processes and new recipes aimed at GPU-like cost and quality.
China raises the urgency. Martinis says researchers reproducing Google’s supremacy experiment understood the theory and generated “great results”; after Google published another materially improved result, Chinese work soon indicated near-parity. Friedberg said he had heard the same concern: Chinese authorities might prevent publication until a comparable result appears in the Western press.
Martinis sees advanced manufacturing access as a possible defensive advantage because, he says, the 300 mm tools they plan to use are unavailable in China. His consortium combines Applied Materials, Synopsys design tools, Hewlett Packard Enterprise and theory startups. The wager is that industrial engineering can create “a huge leap forward” while helping protect their lead.
6. Impact, not novelty, turned the 1980s result into a Nobel story
Martinis measures the experiment’s importance by what followed: 1,000, perhaps several thousand, researchers now pursue superconducting quantum computers, while companies sell machines or access time on them. The field also spurred advances across materials, fabrication, control systems and measurement—“engineering and physics” that he calls beautiful.
Martinis said he and colleagues, including Michel Devoret, had attended Nobel symposia where organizers assessed the field’s vitality and whether potential leaders could give a strong representative talk. He understood that they were being considered, but emphasized that merely receiving the invitation was already a fantastic honor.
Earlier announcement days left him briefly disappointed, an attitude he came to dislike because merely being considered was extraordinary. This year he forgot about the date; his wife took the 3:00 a.m. call but let him sleep until 5:30 so he would not be grumpy for reporters arriving at 6.
Outside his company, Martinis remains drawn to instrument builders. He highlighted Ben Mazin’s superconducting detectors for exoplanet searches, related to work Martinis helped establish in the 1990s. The attraction is consistent from garage projects to quantum processors: “I like building instruments,” especially when better devices open new fields of observation.