Stop Waiting for the Quantum Computer

An average new drug costs roughly USD3 billion, takes 15 years, and fails 96% of the time. Julian Tan likes to sit with the cruelest part of that math. A chemist runs a molecular simulation on a classical supercomputer, hits execute and then waits, sometimes for three years, to learn whether the compound was ever worth modeling.

“It’s computationally too expensive,” Tan says. That is a tidy way of saying the universe is bigger than our transistors.

Tan runs IBM’s quantum business across Asia-Pacific, and he has spent the past few years making a specific argument to executives who suspect quantum computing is a decade of press releases away from mattering. His counter is not a slide but a protein.

The cafeteria test

In 2023, IBM parked a quantum computer inside the cafeteria of the Cleveland Clinic, in the middle of the room, next to the trays. Researchers there could simulate a 10-atom molecule and little more. Then IBM, the clinic and Japan’s RIKEN institute began splitting the work. A quantum chip flags the handful of electron configurations that actually matter. A classical supercomputer grinds through the rest. The team went from 10 atoms to 36, then to 303, and by May to a protein complex of more than 12,000 atoms. That is the size at which molecules start to matter for drug discovery, though reaching it is a prerequisite, not the payoff. Nobody has shortened a 15-year pipeline yet.

That division of labor is the whole thesis, and it carries a clunky name: quantum-centric supercomputing. Forget the sci-fi image of a quantum computer in your pocket replacing your laptop. “We’re not going to have a quantum computer in our pockets,” Tan says. In IBM’s framing, the quantum processor is a co-processor, a strange and specialized accelerator that sits beside CPUs and GPUs the way a GPU once slid in beside the CPU. CPUs handle arithmetic. GPUs handle matrices. Quantum handles the intractable middle, the linear algebra and differential equations that describe nature itself.

For a chief data officer, that reframing matters more than any qubit count. You do not rip out your stack. You bolt quantum onto it for the narrow problems where classical brute force detonates, the 10-to-the-60 combinations Tan calls “classically intractable.”

The business case is real. The proof is contested.

The business examples are piling up faster than the skeptics expected. Boeing is using quantum simulation to weave composite wings for maximum strength at minimum weight. Last September, HSBC and IBM reported that a hybrid quantum-classical model predicted whether a corporate bond trade would fill 34% better than the bank’s standard techniques.

Here is where chief data officers should keep one eyebrow raised. The HSBC result came from historical data, not live trading, and the computer scientist Scott Aaronson, a careful quantum skeptic, waved off the advantage claim as a “qombie,” his coinage for a quantum boast that does not survive inspection. Put that objection to Tan, and he does not dodge it. His reply is about motive, not proof. Companies like Boeing, Bosch, Vanguard and HSBC are not chasing physics papers, he says; they want to know “where quantum can help me in my business to outcompete my competitors.”

“The biggest mistake that you make is not being ready when the [quantum computing] technology becomes real.” – Julian Tan @ IBM

IBM says it has run more than 100 such experiments with partners. So the honest reading is that demonstrations are multiplying while verifiable wins over classical computing stay rare. The 34% is real, but whether quantum earned it remains an open argument.

The company also knows it’s not the only major player. Google’s Willow chip pulled off a verifiable quantum advantage, published in Nature in October, running an algorithm 13,000 times faster than a classical supercomputer, with a result you can actually check. But Tan argues that IBM uses the same techniques, and it redesigned its chips, trading raw qubit count for a square lattice that can host error-correcting code, to get there.

The map is also not even fixed for IBM. In July 2026, it agreed to acquire HRL Laboratories, buying its way into silicon-spin qubits, a third qubit modality that sits beside superconducting and trapped ion. And it is not a two-horse race. Quantinuum is another serious contender on a different hardware bet entirely, building trapped-ion machines rather than IBM’s superconducting chips.

Engineering, not science

The louder question is whether IBM’s timelines are marketing. Quantum has a physics problem, not only an engineering one. Decoherence, the tendency of fragile quantum states to collapse, has humbled optimists for 30 years.

Tan’s answer is honest and straightforward: “We don’t view quantum as a science problem,” he says. In IBM’s telling, what remains is integration, not discovery. “Every year we do a green check box,” Tan says of the roadmap, and by his account IBM is “100%” on it “to date.”

The company is betting the balance sheet on that story. It has committed a fresh USD10 billion to quantum over five years, anchored by Anderon, pitched as U.S.’s first pure-play quantum chip foundry, rising in Albany, New York. Half of the foundry’s USD2 billion comes from a proposed CHIPS Act award, the other half from IBM’s own cash. Tan enjoys the etymology. An andiron is the iron bracket that keeps a forge fed with fuel. The foundry, he says, is meant to keep the industry’s furnace supplied with chips.

This is where IBM’s HRL deal comes in. HRL, a Malibu lab jointly owned by General Motors and Boeing (the same Boeing weaving composite wings earlier in this story) brings spin qubits and quantum sensing, and IBM has floated fabricating spin-qubit wafers at Anderon itself. For a chief data officer, the hedge is the real signal. A vendor spreading its bets across three qubit modalities and its own foundry is a sturdier platform to build a decade-long quantum plan on than one lashed to a single strand of physics. The sensing piece matters too, pointing at new data to collect in navigation, life sciences and defense, not just faster ways to crunch the data you already hold.

Singapore has made the same kind of bet at national scale. In February, Prime Minister Lawrence Wong committed SGD37 billion (around USD29 billion) to the Research, Innovation and Enterprise 2030 plan, naming quantum a primary growth area and “an early and deliberate bet.” The city-state has been early for a while. It stood up the Centre for Quantum Technologies at the National University of Singapore in 2007, when the field was mostly chalkboards. By late 2026, Singapore will host Quantinuum’s Helios system, a trapped-ion machine that ranks as the company’s most powerful and the first Helios installed outside the U.S. While it is not an IBM machine, the ambition is, however, not in doubt. And Tan believes that a Singapore bank or manufacturer will still be looking for a working pipeline.

The deadline you can’t defer

That leaves the CISO with the most urgent, least glamorous quantum project of all, which is not computing but defense. “Harvest now, decrypt later,” the practice of hoarding encrypted data today to crack once a capable machine exists, has already dragged deadlines forward. A June executive order gives U.S. federal systems until 2030 to migrate to post-quantum cryptography. If your data still has value in 2030, it is exposed today.

Tan’s closing advice is not to buy a quantum computer but to stop waiting for one. Anyone can run a real IBM machine free for 10 minutes a month, stand up a small team and find the one problem worth the effort. “The biggest mistake that you make,” he says, “is not being ready when the technology becomes real.”

How close is real? Close enough that the spectators are the ones taking the risk.

Image credit: iStockphoto/onimate