When Do We Actually Need Quantum Computing?
A striking conclusion of our recent work is this:
Only a small minority—on the order of 5–15%—of problems commonly described as “quantum computing applications” actually require quantum mechanics in a foundational sense.
That number might surprise people, so it’s worth being clear about what it does and does not mean. It does not say that quantum computing is unimportant or ineffective. It says something more precise: most problems currently grouped under the “quantum” label already admit classical ways of identifying and amplifying correct answers. Quantum mechanics becomes indispensable only when no such classical notion of correctness exists at all.
The mistake in much of the public discussion is to equate difficulty with quantum necessity. Large search spaces, exponential possibilities, and hard optimization problems feel “quantum” — but size alone is not the right criterion. The real question is whether correctness can be physically marked before the full answer is known.
Marking, Amplification, and Commitment
To make this distinction precise, we separate computation into three stages that are often conflated:
- Marking — creating a physical asymmetry correlated with correctness
- Amplification — concentrating probability or weight onto what has been marked
- Commitment — irreversibly producing a definite outcome
These stages correspond to distinct physical operations with different costs. The central result of our work is simple but far-reaching:
Quantum mechanics is sometimes required to create a mark — but it is not required to amplify or read one out.
Once a mark exists, amplification follows from geometry and irreversibility. Classical systems already do this extremely well: belief propagation in error-correcting codes, constraint propagation in scheduling, and probabilistic inference in Bayesian models are all examples of the same underlying mechanism. Quantum systems can realize this process too — but they are not uniquely required for it.
Why the 5–15% Matters
When this distinction is applied systematically across problems commonly cited as quantum applications — optimization benchmarks, machine-learning pilots, logistics, chemistry workflows, cryptography, simulation, and sampling — a clear pattern emerges. Most admit classical marking mechanisms: energy functions, constraints, likelihoods, or scores that tell you when you are “getting warmer.” Those problems may still be hard, but they do not require quantum mechanics to recover answers once marking exists.
The genuinely quantum-required cases form a much narrower class: tasks where correctness itself is quantum-defined, or where no classical process can produce partial evidence correlated with the answer. Quantum simulation, quantum error correction, entanglement verification, and cryptographic algorithms like Shor’s fall squarely in this category.
The 5–15% figure is therefore not a rhetorical flourish. It is an order-of-magnitude statement about where quantum mechanics is physically indispensable, as opposed to merely useful or interesting.
Why This Clarification Strengthens Quantum Computing
Quantum computing is an extraordinary scientific achievement — but it was never destined to be a general-purpose replacement for classical computing. Like particle accelerators or space telescopes, it is powerful precisely because it is specialized.
Clarifying that only a small fraction of problems truly require quantum mechanics does not weaken the case for quantum computing. It strengthens it. It allows the field to focus on the domains where quantum resources are irreplaceable, rather than carrying the burden of solving every hard problem in sight.
The right question is no longer “Is this problem hard?”
It’s “Can correctness be marked classically — or does it live in the quantum domain?”
That distinction is what turns the 5–15% from a provocation into a guide.