Quantum algorithms are the reason the quantum computer exists. Without procedures capable of turning superposition and entanglement into useful answers, the hardware would be merely elegant physics. This page is the map of the territory. It explains what distinguishes a quantum algorithm from a classical one, presents the founding algorithms Deutsch-Jozsa and Grover and the celebrated Shor's algorithm, describes the variational family that dominates the current era and the phase estimation that unifies many of them, and offers an honest guide to which speed-ups are proven and which are only hoped for, up to how to design and test them in quantum emulation. The approach is operational. Each section says what is proven, what is promising and what is marketing, because investment decisions turn on this distinction.
What are quantum algorithms?
Quantum algorithms are computational procedures designed for the circuit architecture of quantum processors. They are sequences of gates that prepare superpositions, encode the problem in the phases and orchestrate interference so that the final measurement returns the solution with high probability. The difference from a classical recipe lies not in the speed of the individual step, but in the structure. A quantum algorithm works on all configurations together, and the designer's skill lies in making the right one emerge. Contrary to a widespread belief, the power does not come from trying all solutions in parallel and reading them one by one, which quantum measurement does not allow, but from making the computational paths interfere so that the wrong ones cancel each other out. The reference academic catalogue, the Quantum Algorithm Zoo, lists hundreds of variants, but the families that really matter for applications can be counted on two hands.
How do they differ from classical algorithms?
In three structural properties. First, the input is often queried in superposition, that is, on many configurations simultaneously, which makes no sense for a CPU. Second, progress is measured not in steps executed but in amplitudes shifted; the quantity to optimise is the probability of the correct outcome. Third, the output is a distribution and not a value, and the statistical repetition of executions is part of the algorithm, not a defect. A practical consequence follows, often overlooked. Comparing a quantum algorithm with a classical one requires counting the total resources, shots included, as discussed in the comparison quantum computer vs classical computer. This is why serious papers always report the gate count and the number of oracle queries rather than execution seconds, because these are the hardware-independent metrics against which quantum algorithms are compared with one another. A paper that claims an advantage in computing time without stating these quantities should be read as promotional material, not as a scientific result.
What does the Deutsch-Jozsa algorithm do and why is it important?
It solves a deliberately artificial problem, namely establishing with a single query whether a binary function is constant or balanced, where the classical worst case requires half of the possible inputs plus one. No one uses it in production, but its value is foundational. It was the first demonstration, in 1992, that interference can extract a global property of a function without evaluating it point by point. It is the ‘hello world’ of quantum algorithms, the first circuit that anyone studying the subject runs, and in emulation it remains the perfect teaching example, because it is small, exact and has every step of the interference visible. Anyone who has inspected it once, gate by gate, understands the principle that underlies all subsequent algorithms.
How does Grover's algorithm work?
Grover addresses unstructured search, that is, finding the single marked item among N possibilities. Classically an average of N/2 attempts is needed, whereas Grover requires about √N, a proven and optimal quadratic advantage. The mechanism is amplitude amplification. One starts from the uniform superposition, an oracle inverts the phase of the sought item and a reflection about the mean transfers amplitude towards it; by repeating the pair of operations about (π/4)√N times, the probability of measuring it becomes almost certain. On a billion entries this means going from five hundred million to just over thirty thousand queries. It is not an exponential miracle, but it is a respectable multiplier on brute-force problems, from searching unindexed databases to attacks on symmetric keys, where it indeed requires key lengths to be doubled. A practical constraint should be remembered. To beat massive classical parallelism, fast quantum gates with low error rates are needed; here too it is the total accounting that decides whether the theoretical advantage survives on the hardware.
What makes Shor's algorithm special?
It is the algorithm that turned quantum computing from an academic curiosity into a matter of national security. It factors integers in polynomial time, where the best classical methods remain superpolynomial, and in doing so undermines RSA and much of the public-key cryptography in use. The advantage is exponential, and its structure, which reduces the problem to finding a period and uses the quantum Fourier transform to find it, is the prototype of an entire class of techniques. It must be said just as clearly that running it on real keys requires millions of physical qubits with error correction, a scale still far off; the threat is nonetheless already real for those who archive encrypted data today in order to decrypt it tomorrow. For the complete analysis, from the mathematical foundations to the resource estimates for RSA-2048, we refer to the dedicated page on Shor's algorithm and to the countermeasures of post-quantum cryptography.
What is the RSA public-key cryptosystem
RSA, named after its inventors Rivest, Shamir and Adleman, who published it in 1977, is the most widely used public-key cryptosystem in the world. It solves an old problem, namely exchanging encrypted messages without having first exchanged a secret key, and it does so with a pair of mathematically linked keys. The public key is distributed to anyone and is used to encrypt, while the private key stays with the recipient and is used to decrypt; the same mechanism, used in reverse, produces digital signatures. Its security rests on a computational asymmetry. Generating the keys is easy, because it is enough to choose two very large prime numbers p and q and multiply them to obtain n; breaking the system requires the inverse operation, namely factoring n to recover p and q, a task that for a 2048-bit integer takes, with the best classical methods, longer than the age of the universe. Technically, the message becomes a number m, encryption computes c = me mod n with the public exponent e, and decryption computes m = cd mod n with the private exponent d, which can be derived only by knowing p and q. In practice RSA almost always encrypts a disposable symmetric key, for example AES, which then protects the actual traffic; this is the hybrid model on which HTTPS, VPNs, digital signatures and banking transactions rest. Shor's algorithm factors in polynomial time and therefore demolishes exactly the asymmetry on which RSA is founded. Against any known classical attack, by contrast, RSA remains secure if correctly implemented and with adequate keys; real breaches almost always stem from faulty implementations, not from the mathematics.
What are the variational algorithms VQE and QAOA?
The flagship family of the NISQ era. The idea is simple. A short parametric circuit, executable even on noisy hardware, prepares a trial state; the quantum processor measures its energy or cost; a classical optimiser updates the parameters and the process iterates until convergence. VQE (Variational Quantum Eigensolver) searches for ground states of molecules and materials, and is the workhorse of applied quantum simulation; QAOA (Quantum Approximate Optimization Algorithm) tackles combinatorial optimisation. The flip side must be stated honestly. The theoretical guarantees of advantage are weak and the quality depends on the circuit design and on the optimiser, which must also avoid so-called barren plateaus, regions in which the gradient flattens and learning stops. For this reason empirical prototyping, typically in emulation, is indispensable before any promise to management.
What is quantum phase estimation?
This section is for readers with a scientific background. Phase estimation (QPE) is the subroutine that unifies the exact algorithms. Given a unitary operator U and one of its eigenvectors |u⟩ with U|u⟩ = e2πiφ|u⟩, QPE estimates the phase φ to n bits of precision using an ancilla register of n qubits. The controlled applications of U raised to powers of 2 write the phase into the ancilla qubits, and the inverse quantum Fourier transform converts it into readable binary. The QFT on n qubits requires O(n²) gates against the O(n2n) operations of the classical FFT on the whole vector. Shor is, in essence, a QPE applied to the modular multiplication operator, while exact quantum chemistry uses it on Hamiltonian evolution. Anyone who masters QPE therefore holds the key to reading much of the exact catalogue.
Which speed-ups are proven and which are only expected?
Here is the honest taxonomy, updated to the state of research. Mathematically proven are the exponential speed-up for factoring and the discrete logarithm (Shor) and for the simulation of quantum dynamics, and the quadratic one for unstructured search (Grover). Expected but not proven in general are the variational advantages in optimisation and quantum machine learning, where ‘dequantisation’ results have scaled back more than one announcement. Any claim of speed-up that does not specify with respect to which classical algorithm and with which total resources should be read with caution. The professional rule is always to compare against the best known classical baseline, never against brute force. A second criterion of hygiene is to distinguish asymptotic results, valid for large instances on error-corrected machines, from performance measurable today on noisy hardware. Many published disappointments stem from confusing the two levels, and many commercial presentations rely on precisely this confusion.
How are quantum algorithms designed and tested?
With a cycle reminiscent of ordinary software development, but with its own constraints. One starts from small instances of the problem, writes the circuit in a framework such as Qiskit or PennyLane, and runs it where every quantity is observable, that is, in an emulation environment, in which amplitudes, phases and intermediate states are inspectable and every execution is reproducible. Here one verifies correctness, measures noise resistance by activating calibrated error models, and estimates how depth and shots scale as the instance grows. Only afterwards, and only if needed, does one move to real hardware with a reference baseline already in hand, so that every euro spent on machine access produces a comparison and not a hope. This is the native workflow of the EQM platform, described on the page on how a quantum computer works.
Where to start studying them?
From the path that teaching experience has selected. First the concepts of qubit, superposition and interference; then Deutsch-Jozsa run and inspected step by step; then Grover on instances of a few qubits, where amplitude amplification can literally be seen growing; finally a simple variational algorithm on a reduced real problem. Each stage takes place in emulation, where a conceptual error emerges immediately and at no cost. For corporate teams, HumanQ structures this path as guided training of a few weeks, with the aim of reaching a proof of concept, because quantum algorithms are learned by running them, not by reading them. The typical time to bring a software engineer from zero to a working variational algorithm is four to six weeks of guided practice, a small investment compared with the value of being able to read the sector critically and to tell a result from a press release.
Design your first quantum algorithm
With the EQM platform you develop, inspect and compare quantum algorithms on real use cases, with guided training and objective metrics. Discover quantum emulation, assess your company's quantum readiness or contact the HumanQ team.