Nobel Prize Winner: Nobody Sees What's Coming After AI

ELI5/TLDR

John Martinis won the 2025 Nobel Prize for showing, back in 1985, that the weird rules of quantum physics work not just inside atoms but inside electrical circuits big enough to hold in your hand. That discovery is the seed of every quantum computer being built today. His timeline: in five to ten years, a quantum computer big enough to crack the encryption holding up Bitcoin, banking, and most of the internet will exist. Governments and Google have been preparing for years. Most people haven’t.

The Full Story

What he actually discovered

The pitch most people hear about quantum is that an electron can be in two places at once. The Martinis version is more concrete. Throw a tennis ball at a wall, it bounces. Throw an electron at the equivalent wall and every so often the electron just shows up on the other side, having passed through without going around. This is “tunneling,” and physicists already knew it happened inside atoms.

What Martinis showed in 1985 was that this also happens inside an electrical circuit “about that big” — a thing you can solder together on a bench. Voltages and currents in macroscopic hardware were obeying quantum rules. That sounds technical and is. The consequence is not.

That’s the moment quantum stopped being just theory. He proved it works in machines.

If quantum effects only live inside atoms, you can study them but you can’t engineer them. If they live inside circuits, you can build with them. Forty years later, those circuits are quantum computers.

What a quantum computer is actually for

Martinis is uninterested in the science fiction framing. The two examples he keeps returning to are unsexy and useful: simulating molecules and designing materials.

Today, when an architect designs a kitchen or an engineer designs a circuit, they build it virtually first. Cheaper, faster, fewer mistakes. You cannot do this with molecules — the equations describing how electrons in a drug actually behave around a target protein are too hard for normal computers to solve. A quantum computer would. Improve a pharma company’s understanding of how a drug works by even a few percent and the dollars get large quickly.

Healthcare, finance, materials, chemistry. He waves at all of it. The honest answer is nobody knows yet which industry gets hit hardest, because the hardware isn’t good enough to find out.

The hardware bet

Most people in quantum are working on algorithms — the software side — because software is cheap. Martinis runs a company doing the opposite. He is building the actual chips. He frames it as “the exact wrong thing” with a smirk, then explains the bet.

I like to think about Nvidia… their knowledge of designing the computers and putting together the systems have made them extremely valuable.

The analogy is GPUs. For decades, the obvious career move was writing software, not making the silicon underneath. Then Nvidia made the silicon underneath, AI happened, and Nvidia became the most valuable company on earth. Martinis thinks quantum has the same shape — whoever solves the hardware bottleneck owns a disproportionate slice of the trillion-dollar pie.

His specific pitch: today’s superconducting qubits (the leading approach, much of which he invented) are built with what he calls “artisanal” lab fabrication. He wants to make them in actual semiconductor fabs, with actual semiconductor tooling, so they can scale from a few hundred qubits to a million. A million is roughly the number you need before “error-corrected, general-purpose quantum computer” stops being a slide deck and starts being a machine.

The crypto problem

The headline-grabby part of the conversation. A few days before this interview, Google published a paper saying a quantum computer could break Bitcoin’s encryption in nine minutes using twenty times less hardware than previously believed.

Martinis is careful here — he defers to his CEO on the details — but the picture is roughly this. Older Bitcoin uses a weaker encryption scheme. Newer Bitcoin uses a stiffer one. If you hold old Bitcoin, you can move it to a stronger encryption today and be safe. The problem is that a lot of Bitcoin is “kind of unclaimed” — coins held in old wallets nobody touches, including, presumably, Satoshi’s stash. That money becomes hackable the day a sufficiently large quantum computer turns on.

So that is a market for a quantum computer.

Said deadpan. He’s already in conversations with the US Treasury about it, because if a private company or a criminal group gets there first, the implications are not small. The wider point: it isn’t just Bitcoin. The encryption (RSA) holding up banking, messaging, government data, all of it, sits on the same kind of math. The whole internet needs to migrate to “post-quantum cryptography” inside the same five-to-ten-year window.

The good news: this isn’t a surprise to anyone in charge. The math showing quantum computers could break RSA was published by Peter Shor in 1994. NIST (the US standards body) has been running a public competition for quantum-safe algorithms for over a decade. The replacements exist. You can download them. The only question is whether the world swaps them in fast enough.

Timelines: five to ten, not three to five

Google’s CEO says three to five years. IBM, like Martinis, says five to ten. Martinis has a clean explanation for the gap: a CEO making a public prediction has every reason to lowball it.

Because if you’re a CEO and you’re making a prediction, you want to have a lower number because you don’t want to be caught without that.

His own number isn’t a forecast so much as a warning shot. He says it partly because he believes it, partly to make sure people start switching cryptography now rather than later.

The Nobel and the lesson underneath

Toward the end the interview gets more personal. Martinis describes his career as a chain of projects that “didn’t work out in some way,” each setback forcing a rethink that turned out, eventually, to be better than the original plan. The biggest one: he led Google’s quantum team, ran the famous “quantum supremacy” experiment in 2019, and then, in his words, was “not Googly enough” and had to leave a project he had been working on since graduate school.

By leaving Google, I was then free to think very creatively… and through thinking freely and starting it with the co-founders, we were able to come up with a bunch of ideas how to make it better.

The Nobel call came at 2:30 a.m. His wife saw it, let him sleep, and chased the morning’s reporters out of the bedroom so he could rest. He woke up at 6 a.m., opened his laptop, and saw his face on a screen with the king and queen of Sweden. “Sounds like a fairy tale.” “It’s a fairy tale.”

Key Takeaways

Martinis’s 1985 discovery: quantum tunneling, previously seen only in atoms, also happens inside macroscopic electrical circuits. This is the physical foundation that makes superconducting quantum computers possible.
December 2024: Google’s quantum chip ran a calculation in minutes that would take the best classical supercomputer longer than the age of the universe.
Useful near-term applications are mostly chemistry-shaped: drug discovery, materials science, molecular simulation. The framing is “design with molecules the way you design with CAD.”
The error-correction threshold is roughly a million physical qubits. Today’s machines have hundreds. That gap is the entire game.
Two camps in the field: software/algorithms (cheap, crowded) and hardware (expensive, lonely). Martinis’s bet is that hardware wins the way Nvidia won.
His specific technical bet: move from “artisanal” lab fabrication of qubits to industrial semiconductor fabs, so qubit production can scale.
Bitcoin’s older encryption is breakable by a sufficiently large quantum computer. Newer encryption is stiffer. Old, unclaimed Bitcoin (Satoshi-era wallets) is the most exposed pile of money — coins that cannot be re-encrypted because nobody holds the keys.
It is not just crypto. RSA encryption underpins almost all internet security. The entire internet needs to migrate to quantum-safe cryptography inside the 5–10-year window.
The replacements already exist. NIST has run a decade-long program building “post-quantum cryptography” standards. Google already runs some traffic on quantum-resistant protocols.
Peter Shor’s 1994 algorithm is the original proof that quantum computers can break RSA. The threat has been known for thirty years; only the timeline has shortened.
Timeline disagreement is structural: vendor CEOs (Google) say 3–5 years; independent experts (Martinis, IBM) say 5–10. CEOs underbid because they don’t want to be caught flat-footed.
Peter Thiel’s “definite vs indefinite optimist” frame: Martinis self-identifies as a definite optimist — he wants a specific thing to build, not a general bet on the future.
Career texture: Martinis’s biggest jumps came after being pushed out of projects, including Google after the 2019 quantum supremacy experiment. The forced reset enabled the rethink that founded his current company.

Claude’s Take

This is a Davos sit-down with a very real scientist, framed for a YouTube growth audience. The framing is doing a lot of work. The hook (“nobody sees what’s coming after AI”) is a bit overcooked — the 5-to-10-year timeline for cryptographically relevant quantum computers has been the consensus among serious physicists for years, and post-quantum cryptography has been a public NIST project since 2016. This is not breaking news.

What it does offer is a calm, technically credible voice making the case that quantum is finally entering its hardware-execution phase, the same way deep learning did around 2012. Martinis is the right person to listen to on this. He invented the modern superconducting qubit, ran Google’s effort, and is now staking a company on the unsexy, capital-intensive part of the problem. His Nvidia analogy is doing real work. The dull-sounding bet — make better silicon — is also where the trillion-dollar outcome historically hides.

The crypto angle is the part most worth taking seriously. Not because Bitcoin will collapse next Tuesday, but because “harvest now, decrypt later” is already an acknowledged threat: state actors are recording encrypted traffic today on the assumption they can decrypt it once a sufficiently large quantum machine is online. That makes the migration to post-quantum cryptography a real near-term operational problem for banks, messaging platforms, and governments, even if the actual quantum machine is a decade away.

Score 6/10. Solid hardware-pilled answers from one of the best people in the field, but the format is talk-show-with-a-product-placement. The “plot” device read takes up a good ninety seconds, the Davos backdrop limits depth, and several of the host’s framings overstate novelty. Watch it for Martinis’s voice on hardware and timelines; skip it if you’ve read any post-quantum cryptography primer in the last three years.