Chips

A quantum computer is not a faster laptop. It is a different kind of gamble

Quantum computers matter for certain problems because they use quantum states differently from bits. That does not make them better at everything, or useful for your spreadsheet.

June 4, 2026Fang Yu

A useful way in is to ask what the machine is bad at. It cannot hold data reliably, it cannot run for long without errors, and it does nothing whatsoever for the ordinary work a laptop handles. What it has instead is unusual: a quantum computer is built from quantum states that can be manipulated in ways ordinary bits cannot. That gives special algorithms a path to advantage on special problems. It also makes the machines fragile, noisy and hard to scale. If a classical computer is a reliable clerk moving through instructions, a quantum computer is a delicate instrument that sets up interference so wrong answers cancel and useful structure becomes more visible. That is powerful. It is not universal speed.

The first correction: it is not a faster CPU

A normal computer stores information as bits, zeros and ones. A quantum computer uses qubits, which are physical systems described by quantum states. That difference is not a speed setting. It is a different way to encode and transform information.

For many everyday tasks, a quantum computer would be the wrong tool. Email, spreadsheets, web browsing and most application code do not become magically better. Quantum advantage appears only when a problem has structure a quantum algorithm can use.

What superposition actually buys you

Superposition means a qubit can be in a combination of states before measurement. Popular explanations often turn that into it tries all answers at once. That is misleading because measurement gives one outcome, not a complete list of all possibilities.

The trick is interference. Quantum operations are designed so unwanted paths cancel and useful patterns become more likely. The art is setting up the problem so the final measurement carries information you could not get efficiently another way.

Entanglement links qubits in ways classical bits cannot copy

Entanglement creates correlations between qubits that cannot be explained as simple independent states. It is one reason quantum systems can represent certain relationships compactly.

Entanglement is also fragile. The same sensitivity that makes quantum behavior powerful makes it easy for noise and the environment to ruin the computation. Hardware is a fight against unwanted interaction.

The useful problems are narrow but important

Quantum computers may help with simulating quantum systems, certain optimization tasks, materials science, chemistry and some cryptographic problems. These are not small niches if they work, but they are not everything.

The most natural fit is simulating nature at the quantum level. Classical computers struggle because quantum systems grow complex quickly. A controlled quantum system may model another quantum system more naturally.

What will quantum computers be used for?

The honest list is short, and each item has a different amount of evidence behind it.

Chemistry and materials come first. Molecules obey quantum rules, so a quantum machine can model them more naturally than a classical one. The textbook target is FeMoco, the metal cluster inside nitrogenase, the enzyme bacteria use to turn nitrogen from the air into ammonia; a 2017 paper by Reiher, Wiebe, Svore, Wecker and Troyer showed how a quantum computer could work out its reaction mechanism, a question classical methods have not settled. Better catalysts, battery materials and drug molecules are the same kind of problem. None of it has been done yet at a useful scale.

Physics simulation is second, and the only area with a verified result so far. In October 2025 Google reported in Nature that its Willow chip ran an algorithm that measures how quantum information spreads (an OTOC) 13,000 times faster than the best classical method on one of the world's fastest supercomputers, with a result that can be checked. It was a physics experiment, not a product.

Breaking today's public-key encryption is third, and the one that makes headlines. Shor's algorithm would do it, which is why RSA and elliptic-curve cryptography are already being replaced; what that means for coins is covered in will quantum computing break Bitcoin. Optimisation, logistics, finance and machine learning come last. They fill many pitches, but no quantum speed-up on a real business problem has been demonstrated, so claims there deserve the most doubt.

What will they be able to do by 2030?

Watch the vendors' dated promises, because they can be checked when the date arrives. IBM says Starling, planned for 2029, will run 100 million quantum operations on 200 logical qubits, and Blue Jay, targeted for 2033, a billion operations on 2,000. Quantinuum says it will deliver a fully fault-tolerant machine, Apollo, by 2029. DARPA's Quantum Benchmarking Initiative is testing whether any approach can reach utility scale, meaning the computing is worth more than it costs, by 2033. If those machines arrive on schedule, the chemistry problems above are what they will be pointed at first. If they slip, the roadmaps will show it before the press releases do.

Will there ever be a quantum laptop?

Not one that replaces yours. The leading machines need a dilution refrigerator colder than outer space, racks of control electronics, or vacuum chambers and lasers, and email, browsing and spreadsheets gain nothing from qubits. Researchers do not rule out small quantum devices in principle, but nothing on any roadmap would make your daily computing faster.

The only thing sold under that name is an educational box. SpinQ markets portable machines based on nuclear magnetic resonance, with two or three qubits, as a "quantum laptop"; its two-qubit Gemini Mini weighs 14 kg and was reported at about $8,100 when it launched in 2022. It is a teaching tool and cannot beat a phone at anything.

If you want to run something on a real quantum computer, use the cloud. IBM's free Open Plan gives up to 10 minutes of processor time in every 28-day window. Amazon Braket charges $0.30 per task plus a fee per shot, from $0.000425 on a Rigetti machine to $0.08 on IonQ's Forte, so 1,000 shots on Forte cost about $80. Ten minutes is plenty for the small circuits a beginner writes.

Error correction is the mountain between demo and machine

Qubits are noisy. Calculations fail unless errors are detected and corrected. Quantum error correction uses many physical qubits to create a smaller number of more reliable logical qubits.

This is why raw qubit count alone is not enough. A machine with many noisy qubits may be less useful than a smaller system with better fidelity and error control. The scoreboard is quality, connectivity, control and correction together.

Quantum supremacy was a milestone

Experiments showing a quantum machine outperforming a classical computer on a specialized benchmark are scientifically important. They do not mean the machine is now useful for ordinary tasks.

A benchmark can prove a principle while still being chosen because it fits the machine. Practical advantage means solving a valuable problem better than the best available classical methods. That bar is higher.

Cryptography risk is real but not immediate for everyone

A large fault-tolerant quantum computer could threaten some widely used public-key cryptography. That is why post-quantum cryptography is being standardized and adopted before the threat is fully realized.

The right reaction is preparation, not panic. Systems with long-lived secrets should plan migration earlier. Your laptop password is not suddenly gone because a lab improved a qubit.

The hardware paths are still competing

Superconducting circuits, trapped ions, neutral atoms, photonics and other approaches each have tradeoffs. Some are faster, some have longer coherence, some may connect more easily, and all face scaling problems.

The field has not settled on one inevitable architecture. That is normal at this stage. Early computing also had competing hardware ideas before industrial patterns stabilized.

Reading the specs: four numbers, and what each one leaves out

Announcements about quantum hardware are usually built around a single figure. Four numbers matter, and each of them is close to meaningless without the other three.

Qubit count is quoted most often because it is the easiest to understand. On its own it says nothing about whether the machine can finish a useful calculation, only how many pieces it has.

Error rate is how often a single operation goes wrong. Multiply a small per-operation error across the thousands of operations an interesting algorithm needs, and a machine with a great many qubits can still fail to produce a usable answer. This is the number that tends to be buried furthest down the announcement.

Coherence time is how long a qubit holds its state before the environment scrambles it. It sets a ceiling on how long a computation can run before the answer dissolves into noise, and it is a large part of why comparisons between architectures rarely reduce to a single winner.

Connectivity is which qubits can interact with which. When two qubits that need to talk are not neighbours, the machine spends extra operations shuttling information between them, and every one of those operations carries its own error.

Take a comparison built only out of structure: a machine with a thousand qubits but an error rate an order of magnitude higher, set against one with a few dozen qubits and much lower error. Ask which of them finishes the same algorithm, and the answer often runs opposite to the headline.

What is actually inside the box

Photographs of quantum computers usually show a chandelier of gold-coloured plates. That object is mostly not the computer.

Most of what you can see is the refrigerator. Several superconducting designs need temperatures far below anything that occurs in nature, and each gold stage is one step further down, with the actual chip a small thing at the very bottom. Other approaches, trapped ions or photonics among them, look nothing like that, which is one good reason photographs are a poor way to compare architectures.

Around the fridge stands a rack of fairly ordinary electronics producing precisely shaped microwave or laser pulses. Every quantum operation is one of those pulses, so the accuracy of the whole machine rests on something that is, in the end, extremely careful signal generation.

Driving all of it is a completely conventional computer. It compiles the program, schedules the pulses, gathers the results and does the statistics afterwards. A quantum processor is a co-processor: a classical machine asks it a narrow question and then has to interpret whatever comes back.

Holding that picture in mind deflates a good deal of the coverage. Nothing in the architecture suggests a device that replaces the computer on your desk. It suggests a specialised instrument that a classical computer occasionally calls.

What running a job on one actually looks like

The daily experience of using a quantum computer is far less cinematic than the coverage, and knowing its shape makes the limitations obvious.

You do not log in and get a prompt. You describe a circuit, a sequence of operations on a set of qubits, and submit it to a queue, usually through a cloud service. Somewhere in that queue your circuit is compiled into the pulses that particular machine understands, and depending on which qubits can talk to which, what comes out can differ substantially from what you wrote.

Then it is run many times over. One run does not give you the answer; it gives you a single sample from a probability distribution. Practitioners talk about shots, meaning repetitions, often thousands of them, and the result anyone cares about is the shape of the histogram rather than any individual outcome.

Finally the noise has to be dealt with. Some of the spread in that histogram is the physics you were after and some of it is error, and separating the two is a discipline in itself. This is why a demonstration usually arrives as a statistical claim rather than as a computed value.

So the honest summary of the user experience is this: submit a circuit, get back a distribution, then argue about how much of it was signal. That is not a failure of the field. It is what the machine is.

How to read a quantum breakthrough headline

Ask what improved: qubit count, gate fidelity, coherence time, error correction, connectivity, algorithmic performance or a real application. Then ask whether the result scales.

The honest story is neither miracle nor joke. Quantum computing is a serious attempt to build a new class of machine for certain hard problems. It is just not a better laptop.

What ordinary users should do now

Most people do not need to touch quantum computing. Developers and security teams should watch post-quantum cryptography. Students can learn the concepts without assuming an immediate job market explosion.

If your organization stores data that must remain secret for many years, the planning horizon is different. Inventory cryptography, follow standards work and avoid waiting for a crisis headline.

Classical and quantum, side by side

A short checklist before you believe a claim

• Ask what problem the quantum result actually solved.
• Look for logical qubits or error correction, not only physical qubit count.
• Separate benchmark advantage from business use.
• Treat cryptography migration as planning, not panic.

Four claims that sound right and are not

Updated: June 4, 2026. Sections on reading hardware specifications, on what the machine physically is, and on what running a job looks like were added on September 8, 2026. On September 29, 2026 we added what quantum computers are expected to be used for, the dated milestones to watch, whether a quantum laptop exists, and the free and paid ways to run a circuit on real hardware.

Fang Yu · Editor of FutureLens

Fang Yu is the editor of FutureLens, turning published papers, official materials and public explanations into plain-language notes. He is most interested in the gap between a technology's public pitch and the evidence a careful reader can actually check. More about the author