Quantum Amplitude Estimation

Overview

Quantum amplitude estimation estimates the probability that a quantum process produces a particular outcome, which is the amplitude squared of the corresponding state, to a precision that improves in proportion to 1 over the number of runs. A classical Monte Carlo method improves only as 1 over the square root of the number of samples. For a given accuracy, amplitude estimation therefore needs quadratically fewer repetitions.

That quadratic gain is the basis of many proposals for quantum finance, including option pricing and risk analysis, since those problems are classically solved by Monte Carlo sampling. Amplitude estimation is the quantum replacement for the sampling loop.

The problem it solves

Estimating the probability of an event or the expected value of a function, traditionally done via repeated random sampling (Monte Carlo methods).

History and background

Brassard, Høyer, Mosca, and Tapp posted the algorithm in 2000, building on Grover's search and Kitaev's phase estimation. It generalizes quantum counting. The paper was published in 2002 in a volume of the American Mathematical Society's Contemporary Mathematics series, and it is now a standard reference.

For two decades the algorithm needed phase estimation and deep circuits. In 2019 and 2020, groups including Suzuki and colleagues, Grinko and colleagues, and Aaronson and Rall proposed versions without the quantum Fourier transform that use only Grover iterations and classical post-processing, making the method more suitable for noisy hardware. Applications to finance were developed by Rebentrost, Gupt, and Bromley in 2018 and by Woerner and Egger in 2019, and resource studies such as Chakrabarti and colleagues in 2021 estimated the hardware needed for useful derivative pricing.

How it works

Combines amplitude amplification with phase estimation to extract a probability estimate with quadratically fewer samples than classical Monte Carlo.

Suppose a quantum circuit A prepares a state in which the good outcome has probability a, which equals the sine squared of theta. The Grover operator Q, built from A, rotates the state by an angle of 2 theta in the plane spanned by the good and bad components. The number we want, a, is hidden in this angle.

Estimating the angle is a phase estimation problem for Q. Using t control qubits and applying Q up to 2 to the t minus 1 times in total gives an estimate of theta, and therefore of a, with an error of about 1 over 2 to the t. The cost, in uses of A, is about 2 to the t, so the error scales as 1 over M for M uses, which is the Heisenberg limit.

A classical sampler repeats the process M times and counts the fraction of good outcomes, with statistical error 1 over the square root of M. That is the source of the quadratic gap. Newer versions get the same scaling by running Q a carefully chosen number of times and fitting the outcomes with maximum likelihood.

Step by step

1. Define a quantum circuit A that prepares a state with the good subspace marked by a flag qubit and a target amplitude a to be estimated.
2. Construct the Grover operator Q from A and the reflections about the good subspace and about the initial state.
3. In the canonical version, add t counting qubits, apply Hadamards, and apply controlled powers of Q.
4. Apply the inverse quantum Fourier transform and measure the counting register to obtain an integer m.
5. Compute the estimate of a as the sine squared of pi times m divided by 2 to the t.
6. In the iterative versions, instead run Q^k for several values of k, measure the flag qubit many times at each, and combine the results by maximum likelihood to get a.

A worked example

Suppose a financial model gives a payoff that, after scaling, equals the probability a of measuring a flag qubit as 1. A classical Monte Carlo simulation to estimate a to within 0.001 needs about a million samples, since the error shrinks like 1 over the square root of the sample count.

Amplitude estimation reaches an error of 0.001 with on the order of a thousand applications of the Grover operator. Each application is a deep circuit that includes the full model, so the real comparison depends on how fast each circuit run can be repeated on hardware.

Complexity and resources

What hardware it needs

• Fault-tolerant hardware able to run deep circuits with many repetitions of the model.
• Fast logical gates: published resource estimates for option pricing require logical clock rates far above what is available today.
• Efficient loading of probability distributions into quantum states.

Applications

• Option pricing, credit risk, and value at risk calculations in finance.
• Monte Carlo integration for physics and engineering simulations.
• Estimating expectation values in quantum chemistry measurement.
• Estimating means and moments in statistics and machine learning.

Limitations and caveats

• The quadratic speedup is small enough that overheads from error correction and state preparation can cancel it at realistic problem sizes.
• The circuit A must implement the full model coherently, including random-number-like distributions, which is costly.
• Noise degrades the Heisenberg scaling, which limits the benefit on near-term devices.

Common misconceptions

• It does not speed up Monte Carlo by an exponential factor. The advantage is quadratic.
• It is not usable on current hardware for finance problems of practical size.

How it compares

• Quantum counting is the special case where A creates a uniform superposition.
• Classical Monte Carlo is simpler and runs on current computers, but needs quadratically more samples.
• Quantum mean estimation extends the idea to estimating the mean of a general random variable.

Real-world impact

A leading candidate for near-term financial applications like risk analysis and options pricing, where Monte Carlo methods are already heavily used classically.

Why it matters

Amplitude estimation is the main route by which quantum computing could affect finance, risk management, and scientific simulation, which are all heavy users of Monte Carlo methods. Banks have invested in resource studies precisely because the quantum version has a clean provable speedup.

It is also a good example of how a theoretical gain must be weighed against constants. Resource estimates have shown that the hardware needed for a benefit is demanding, which sets realistic expectations for the timeline.

Frequently asked questions

Keep learning

• Related algorithm: Quantum Counting
• Related algorithm: Quantum Monte Carlo Integration
• Related algorithm: Quantum Mean Estimation
• Related algorithm: Quantum Amplitude Amplification

This page is written for educational purposes. Complexity statements describe the best known results at the time of review (September 30, 2026), and research in this area moves quickly, so check recent papers before relying on a specific bound.

← Back to Algorithms Database