The Quantum 2012 Moment

Programming note: ARPU will be off this week, back on Monday.

Shovels for the Next Gold Rush

If you are a global bank, you sometimes find it useful to tell your shareholders that you are on the cutting edge of technology. You want to suggest that you are using the most advanced physics known to man to squeeze an extra basis point out of a bond trade.

And so, recently, HSBC announced a "world first": algorithmic trading on a quantum computer. It was a great headline. The only problem, as the Financial Times reported, was that even HSBC's own researchers couldn't explain why the quantum computer achieved better results, and actual quantum physicists dismissed it as a "zombie claim"—a result that looks alive but has no actual scientific soul.

Usually, when a bank claims to be using magic physics to trade bonds and the scientists say "that's not how physics works," the story ends there. But in the quantum computing market, the fake headlines might actually be a sign that the real industry is finally starting.

The argument, put forward by IBM's head of research Jay Gambetta, is that quantum computing is currently having its "2012 Moment."

In 2012, a neural network named AlexNet crushed the competition at an image recognition contest. Crucially, the researchers didn't achieve this using a massive supercomputer cluster; they used two Nvidia GTX 580 graphics cards—consumer hardware designed to render realistic explosions in Call of Duty. This was the moment the industry realized that the parallel processing power used for video games was actually perfect for the matrix math of AI. It transformed Nvidia from a toy maker for teenagers into the engine room of the AI economy.

That sparked the Big Bang of the modern deep learning era. Thirteen years later, those two gaming cards have evolved into a $1.4 trillion data center build-out.

The quantum industry is betting that history is about to rhyme. The machines are still noisy and error-prone. They are not yet fault-tolerant. But the financial machinery required to scale them is being assembled with surprising speed.

The Jensen Pivot

The most important signal here is not a breakthrough in the lab, but a change of heart in the boardroom.

In January 2025, Nvidia CEO Jensen Huang famously dismissed quantum computing as being "15 to 30 years away" from being useful. In the tech world, "30 years" is a polite way of saying "never."

But by June, Huang was on stage in Paris declaring that the industry had reached an "inflection point."

Did the laws of quantum mechanics change in five months? Probably not. But Nvidia's product roadmap did. Huang unveiled CUDA-Q, a platform for hybrid quantum-classical computing.

This is the classic Nvidia playbook. Huang isn't trying to win the race to build the perfect quantum processing unit (QPU). He doesn't care if the winning qubit is made of superconducting loops or trapped ions. He just wants to build the bridge that connects those exotic, noisy machines to his very profitable GPUs.

The vision Huang laid out is one where every supercomputer eventually has a QPU attached to it, like a sidecar on a very expensive motorcycle. The QPU handles the weird physics simulations; the GPU handles everything else. It reframes quantum from a replacement technology into a high-margin accessory for the AI factory. When the person selling the shovels for the current gold rush starts building a connector for the next type of shovel, it is usually a sign that the digging is about to start.

The Financial Noise

If the physics is still "noisy," the capital markets are coming through loud and clear.

2025 has become the year of the quantum land grab. Pure-play quantum stocks like D-Wave (up over 600%) and Rigetti (up over 300%) are trading at valuations that exist entirely in the future. D-Wave is currently priced at over 300 times its next-twelve-months revenue; Rigetti is at 500 times. Meanwhile, IonQ is on an acquisition spree, buying up companies like ID Quantique and Oxford Ionics to secure networking patents and supply chains.

This is the financial footprint of a technology that is moving from "science project" to "valuation category." It is an asset class in the sense that you can buy and sell the stocks, even if the underlying assets don't actually perform a function yet. Industry consultants describe this phase as "readiness rather than deployment," which is a polite way of saying that companies are buying the machines as very expensive training wheels for their science teams, rather than because they have a problem that requires a quantum solution today.

And so you have a market where:

1. Banks are publishing press releases about quantum bond trades that scientists don't understand.
2. Startups are raising billions to build fault-tolerant machines that don't exist yet.
3. Nvidia is building the software plumbing to ensure that, whoever wins, the data still flows through a Blackwell chip.

The risk, of course, is that the AlexNet parallel is wrong. In 2012, the researchers didn't have to invent new physics; they just had to realize that gaming chips were good at math. Quantum computing is a much harder bet. It requires building fault-tolerant systems that work at near absolute zero and don't collapse if someone sneezes in the next room.

If those machines never arrive, 2025 will be remembered as the moment the industry mastered the zombie claim: a way to keep raising billions for a technology that is always ten years away. For now, Nvidia and the banks are happy to get on the headline. Whether they ever get the computer is a different question.

More on Quantum Computing:

• Next-Level Quantum Computers Will Almost Be Useful (IEEE Spectrum)

On Our Radar

Our Intelligence Desk connects the dots across functions—from GTM to Operations—and delivers intelligence tailored for specific roles. Learn more about our bespoke streams.

The TSMC Center of Gravity

• The Headline: TSMC is poised to report a record 27% jump in quarterly profit, as insatiable AI demand cements its status as the non-negotiable supplier for the entire industry. (Reuters)
• ARPU's Take: TSMC has become the sole, indispensable infrastructure layer of the AI revolution. They are functionally indifferent to who wins the AI model or GPU war—be it Nvidia, Apple, or AMD—because every contender must pay their toll at the foundry gate.
• The Investment Implication: These results solidify TSMC's status as the ultimate picks and shovels investment of the AI boom. While the application and model layers are a high-risk, high-reward bet on which company will win, TSMC represents a non-speculative claim on the entire industry's aggregate capital expenditure. Their earnings are a direct proxy for the total amount of money being poured into the AI arms race, making them the most durable and least volatile way to capture value from the entire trend.

P.S. We're building a handful of bespoke intelligence streams for members. If you're facing a specific intelligence challenge, drop us a line here.

You received this message because you are subscribed to ARPU newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.