What Can Quantum Computers Do Today? An Honest Answer

Not much that ordinary computers cannot already do, at least in the cases we have reported and checked. No quantum computer in our coverage has yet run a useful calculation out of reach of regular machines. What quantum technology can do today is narrower: quantum sensors are being tested at sea as a backup to GPS, quantum simulators are helping physicists study problems that are very hard to compute, and quantum computers are getting steadier and easier to check.

It matters beyond the lab: to ship crews who cannot rely on satellite signals, to chemists checking a new drug candidate, and to anyone who pays for electricity lost in wires. Here are the quantum stories we have published so far, grouped by what they show.

Quantum sensors: the closest to everyday use

Ships that lose or distrust satellite signals may one day steer by gravity instead. In a sea trial off Australia, the company Q-CTRL used a quantum gravimeter and matched its gravity readings against a map. Over an 83 kilometer coastal route with satellite positioning withheld, the final position error fell from about 26 kilometers to about 4 kilometers. It is a company preprint, not yet peer reviewed, partly funded by the UK Royal Navy, and the run lasted about six hours. Read how a quantum gravity sensor helped a ship find its way without GPS.

Keeping quantum computers steady long enough to matter

The calculations that could one day help design medicines and materials may need to run for days or even months, but qubits drift out of tune and the usual fix is to stop and recalibrate. Google Quantum AI and Google DeepMind reported in Nature that a reinforcement learning agent adjusted more than 1,000 control settings on the Willow processor while error correction kept running, using the machine’s own error alarms as feedback. Against drift injected on purpose, stability improved up to 3.5-fold. The chip was storing information, not running a useful calculation. See how a quantum chip learned to retune itself while it works.

Has a quantum computer beaten a regular one? Checking the claims

Before anyone relies on a quantum answer, outsiders have to be able to check it. This summer an IBM Research and University of Chicago team posted a 70-qubit experiment that came with its own certified accuracy estimate. Researchers at the Singapore University of Technology and Design and NVIDIA used 256 GPUs to compute the exact odds of every output IBM had published. They found a weak spot in the circuit’s shape, but the accuracy estimate held up. IBM redesigned the test as a 64-qubit circuit, with the challengers as coauthors. Both papers are preprints, and the task is a benchmark, not a useful calculation. Follow the summer a quantum advantage claim got checked in public.

Standing in for physics that ordinary computers struggle with

Some copper-based materials carry electricity with no resistance at surprisingly warm temperatures, and physicists have argued about why for decades. A team led by Markus Greiner at Harvard built a model of these materials from ultracold lithium atoms in a grid of laser light and mapped the pseudogap, a strange metallic state at the center of the argument. Published in Nature, it used an analog simulator built for this kind of problem, not a general-purpose computer, and it is not a measurement in a real material. Read how cold atoms mapped a decades old superconductor puzzle.

A Duke-led team used a chain of 13 trapped ytterbium ions to simulate a simplified version of how the strong force, stretched like a string, snaps and makes new particle pairs. The pairs appeared first near the string’s ends, not where the textbook picture puts them. The work, in Nature Physics, used a one-dimensional toy model. At 13 ions ordinary computers can still check the answer, and the team’s checks agreed. It offers understanding, not a product. See how 13 trapped ions acted out matter springing from a stretched field of force.

First steps toward chemistry and medicine

Chemists making a candidate medicine often use NMR to confirm they made the right molecule. Researchers from HQS Quantum Simulations and Quantinuum used 42 qubits on Quantinuum’s H2 trapped-ion computer to simulate the NMR spectrum of a hard benchmark molecule, and its main features matched the classical reference. The authors claim no quantum advantage: ordinary computers can still solve this molecule, and the resolution, about 0.07 ppm, is too coarse for the roughly 0.01 ppm detail chemists care about. It is a company preprint. Read how a quantum computer took a first pass at checking the molecules behind medicines.

How close is any of this to everyday life?

Nothing in our quantum coverage is in everyday use yet.

Getting close: gravity navigation at sea. Tested on a real ship; the authors say multi-week tests, sharper maps and a smaller system are still needed.
Years away: self-tuning error correction on Willow. A step toward long, uninterrupted runs, shown so far in a memory experiment.
Further still: the IBM benchmark, the NMR simulation, the pseudogap map and the string breaking simulation. Each adds to science or to trust in the machines, but none is a calculation anyone needed yet.

Questions people ask

Is quantum computing real yet?

Yes, the machines are real and running experiments, but in the cases we have reported none has yet done a useful job that ordinary computers cannot. In each, the authors describe their work as a step: a memory experiment, a benchmark, or a test molecule that classical computers can still solve.

Has quantum advantage been proven?

Not in a settled way in the stories we have checked. IBM and University of Chicago researchers claimed it on a specially designed sampling benchmark, then outside researchers computed the exact odds of the original run’s published outputs on 256 GPUs. IBM’s revised, harder test is estimated to be infeasible for current tensor-network and stabilizer-based classical methods, but it is a preprint, not peer reviewed, and the task is a benchmark rather than a useful calculation.

What is quantum error correction?

Qubits make mistakes, so error correction protects the quantum information inside the machine by constantly checking for the telltale signs of errors. Those checks produce alarms called detection events. In Google’s Willow work, a learning agent used those alarms as feedback to keep more than 1,000 control settings in tune while error correction kept running. The goal is a machine that can run for days without stopping to recalibrate.

Can quantum sensors replace GPS?

Not replace, but they could back it up. In a sea trial off Australia, Q-CTRL’s quantum gravimeter helped a ship navigate by matching Earth’s gravity to a map, with satellite positioning withheld. The final position error fell from about 26 kilometers to about 4 kilometers over a roughly six hour run. It is a company preprint, not yet peer reviewed, and longer tests and a smaller system are still needed.

Will quantum computers help make new medicines?

Possibly one day, but not yet. In one preprint, Quantinuum and HQS used 42 qubits to simulate the NMR spectrum chemists use to check a molecule’s structure, and the main features matched the classical answer. The authors say ordinary computers can still do this job and the result is not sharp enough for routine use. Long calculations for medicines and materials will also need machines that can run for days.

The honest picture is a field doing careful, checkable work: a ship that found its way by gravity, and challengers invited onto a paper after finding its weak spot. To see when one of these steps moves closer to a chemistry bench or a ship’s bridge, follow new stories on our quantum page.