Ask when quantum computing will become useful, and you get different answers, different timelines, and different applications. That’s because the answer depends on what “useful” means.

Breaking encryption, simulating chemistry, and solving business problems are different jobs. Each one needs a different level of quantum computing power. No quantum computer today can handle the biggest of these tasks. Even so, governments are already setting deadlines to prepare for them, through NIST’s post-quantum cryptography transition deadlines.

This article walks through what “useful” actually requires, how far today’s machines are from it, and what to watch next.

Want to see how today’s machines measure up against these benchmarks? Browse our Quantum Sector Landscape Scorecards to compare seven companies on qubit count, accuracy, stability and roadmap.

Why There Is No Agreed-Upon Date

There is no agreed-upon date because there is no agreed-upon task. An investor who hears “useful by year X” should first ask: useful for what? A quantum computer becomes useful when it can perform a specific task accurately and efficiently. These two dimensions are important:

Accurately means it gets the right answer, reliably, not just some of the time.
Efficiently means it gets there in a practical amount of time, and does better than the ordinary computers already available.

It does not have to solve every problem to be valuable… which cuts both ways. This means quantum computers could become useful for smaller, specific tasks sooner, even if they are still far from solving larger problems with greater commercial value. Many Proof-of-Concepts are designed with this in mind.

Think of a golf cart, a pickup truck, and a freight train. All three are vehicles. Each is built for a different job, and nobody would value a golf cart maker on freight rail economics. Quantum computer applications are the same way.

Three applications come up most often:

Breaking today’s encryption. Mostly a security and policy question. Its main market effect today is the cost of defending against it.
Simulating complex chemistry. A potential research and industrial market, and the most demanding benchmark in this piece.
Speeding up certain business calculations. The broadest category, with no single published benchmark to measure against.

The first 2 have something investors rarely get in this sector: published estimates of the hardware each would take. That gives us a fixed yardstick to measure today’s machines against. We are ignoring the third application in this article, as it requires a great deal longer to dissect.

Why does “physical” versus “logical” matter so much? Read Physical Qubits vs. Logical Qubits: Why the Headline Number Rarely Tells the Whole Story

The First Test: Breaking Encryption

Quantum computers could eventually break today’s encryption, and better algorithms have sharply cut the estimated hardware needed to do it. But current systems are still far from that point, even as NIST prepares organizations for new security standards.

Why encryption is a target

Much of today’s internet security rests on math problems that ordinary computers cannot solve in any practical amount of time. The protection does not come from the problem being impossible, but from the problem taking far too long to solve. A powerful enough quantum computer could change that. It could solve some of these problems much faster, including the math behind RSA encryption.

So the real question is not whether it is possible in theory. It is how powerful the machine needs to be.

Google researcher Craig Gidney estimates that breaking a 2048-bit RSA key would take fewer than 1 million noisy physical qubits and less than a week. Is “less than a week” the metric? It depends on the need. Very few companies have a plan for a machine this size. IonQ’s roadmap targets at least 2 million physical qubits by 2030. PsiQuantum has said it aims to house a computer of up to 1 million qubits, but its public timeline has been loose. IBM’s roadmap points to a machine called Starling in 2029. It targets 200 logical qubits (error-corrected qubits, which are better than physical ones). IBM has not given a physical qubit count. Quantinuum targets “hundreds of logical qubits” with its Apollo system, now expected by 2029. For contrast, the largest chip in our scorecard today is IBM’s, at 120 physical qubits.

However, these targets change. In 2019, the estimate was about 20 million qubits. This number is therefore fluid and expected to decrease as QPUs increase in capacity.

NIST’s October 2026 requirements, however, and not fluid, and sets 2 dates for when they expect this to become a threat. NIST has set two dates for retiring today’s most common encryption, including RSA.

After 2030, NIST labels these methods “deprecated.” That means they can still be used, but NIST warns they carry risk. Each organization has to decide for itself whether to keep using them.

After 2035, NIST labels them “disallowed.” That means they are no longer allowed under NIST’s standards.

A June 2026 executive order and White House memo require federal agencies to update their most important systems with quantum-safe security. The deadlines are 2030 for protecting private data and 2031 for verifying digital signatures.

Companies selling to the federal government may also need to comply by the end of 2030, although it is not final yet. Meanwhile, the NSA aims to make military and intelligence systems quantum-safe by 2035.

It’s important to note that these deadlines arrive before anyone has shown a quantum computer that can break these systems. Organizations have to move to new security standards first, without waiting for the threat to appear.

The Second Use Case: Simulating Chemistry

Quantum computers could help simulate complex molecules, the kind that ordinary computers struggle to model accurately. Example use cases are:

Molecular energy. Calculating how much energy a molecule holds in different arrangements. This tells chemists which forms are stable and which are not.
Reaction pathways. Tracing the steps a reaction goes through, including short-lived in-between stages that are hard to observe in a lab. Catalysts, which speed reactions up, are a major target. FeMoco falls in this group.
Materials. Modeling how electrons behave inside solids. Battery materials and superconductors are common examples.
Drug binding. Estimating how strongly a drug candidate attaches to the protein it is meant to act on.

Molecules are so complex that ordinary computers struggle to simulate how they behave. Researchers can approximate them, but the approximations often miss what matters.

Quantum computers could eventually help. They could give researchers a clearer picture of these molecules and the chemical reactions they take part in. FeMoco is a useful test case. It is part of an enzyme involved in converting nitrogen into forms living organisms can use. It is also a molecule with a published, detailed estimate of what a quantum simulation would cost.

Better simulations could help researchers understand how these molecules behave and how they take part in chemical reactions. Current estimates suggest that would take more than 2,100 reliable logical qubits and billions of operations. Today’s demonstrated systems are far below those requirements.

A published estimate says simulating FeMoco would require:

2,142 to 2,196 logical qubits. These are qubits protected through error correction.
5.3B to 32B Toffoli gates. These are a type of operation used in quantum calculations.
Around 4 million physical qubits at a 0.1% error rate, finishing the task in under 4 days.
Or about 1 million physical qubits if the error rate improves to 0.01%.

Quantum computing companies have published roadmaps for building larger and more reliable machines. These targets are not directly comparable, and none has been demonstrated yet. Commercial value is a separate challenge on top of that.

Company	Published target	Year	What still needs to be proven
IBM	200 logical qubits and 100M quantum operations	2029	Delivering the promised scale and operation count
Quantinuum	Hundreds of logical qubits with error rates below 10⁻⁶	Before 2030	Reaching that scale while holding the target error rate
IonQ	800 logical qubits	2027	Independent verification of logical qubit performance and error rates
IonQ	1,600 logical qubits	2028	Showing the projected logical qubits are reliable enough
IonQ	80,000 logical qubits	2030	Scaling to that size with verified logical performance
DARPA	Verify whether a useful quantum computer can be built	By 2033	Whether companies can demonstrate practical, scalable systems

Roadmaps are promises. Our Quantum Sector Landscape Scorecards test them against what each company has shown so far, and our Company Profile Deep Dive Reports cover each company in full for members.

The chemistry benchmark gives a useful reference point of about 2,142 logical qubits.

IBM’s 2029 target of 200 logical qubits is about 1/10 of that.
IonQ’s 2028 target of 1,600 logical qubits is about 3/4 of that.
Quantinuum’s demonstrated 48 error-corrected logical qubits is about 1/45 of that.

These comparisons only cover qubit counts, but say nothing about whether the qubits would be reliable. Most of these figures are also projections, while Quantinuum’s 48 is a reported result. Therefore, from our reviewed of published roadmaps, we are estimating it will be at least 2030 before we can run chemistry experiments on quantum computers. However, there are a lot of estimates and assumptions baked into that target, which means that a lot can happen that postpones this until 2033.

Where Things Stand

There is no single date when quantum computing becomes useful. Today’s systems cannot yet handle the major encryption and chemistry benchmarks covered here. 2028 may be an important checkpoint, but practical usefulness will depend on demonstrated performance, not company promises.

No quantum computer built today can perform either of the 2 major benchmark tasks at the required scale. The gaps are large:

Encryption: today’s largest gate-model systems are thousands of times short of the 1 million qubit reference point.
Chemistry: the best reported error-corrected result is about 45 times short on logical qubits alone, before counting the billions of operations required.

What We’re Watching

Better algorithms. The estimated qubits needed to break RSA-2048 fell from around 20 million in 2019 to fewer than 1 million in 2025. That is about a 20x reduction using better methods. More gains like this would shrink the requirement without any new hardware.

Better error correction. More reliable logical qubits could reduce the physical hardware required. The FeMoco estimate shows the effect of this. Improving the error rate from 0.1% to 0.01% cuts the physical qubits needed from about 4 million to about 1 million.

Why does a tiny change in error rate matter so much? Read 0.09% is the Future of Quantum Computing.

Faster hardware development. If IonQ’s projected logical qubits are independently verified, its roadmap would bring its qubit count close to the chemistry benchmark around 2028. That would be ahead of IBM’s 2029 target. The key word is “if.” The projection only counts once outsiders confirm the qubits perform as claimed.

However, there are factors that could slow down the timeline to usefulness.

Higher error rates than expected. Every estimate in this article assumes a specific error rate. Miss it, and the hardware needed goes up.
Difficulty connecting and controlling larger numbers of qubits. This is the calibration problem from earlier. It gets harder as machines grow.
Roadmap delays, or results that cannot be repeated at larger scale. A result that works once in a lab is not the same as one that holds up in a bigger system.

See why this gets harder as machines grow in Calibration and the Future of Quantum.

There is no single year when quantum computing suddenly becomes useful for everything.

Smaller, specific applications may become useful sooner. However, this progress does not mean quantum computers are close to solving larger, more complex problems, which are the justification for the quantum computing revolution to begin with.

Keep reading. Compare seven companies in our Company Profiles Deep-Dive Reports. Get plain-English explainers and breaking industry news in News & Insights. For company-by-company analysis, visit our members-only portal at members.coherence.report.

When Will Quantum Computing Be Useful?