How Quantum Computers Actually Work — From Qubits to Real Algorithms (2026)

Photo: Pachon in Motion / Pexels
A quantum computer encodes information in quantum bits — qubits — that exploit superposition and entanglement to evaluate many computational paths simultaneously. Unlike classical processors that execute one instruction at a time, quantum processors manipulate probability amplitudes so wrong answers cancel and correct answers reinforce. IBM's Heron processor (156 qubits, 2024) and Google's Willow chip (105 qubits, 2024) represent the current frontier: machines powerful enough to demonstrate quantum advantage on specific tasks, yet fragile enough that error correction remains the defining challenge of the field.
The Classical Computer's Limit — Why Quantum Computing Exists
Every classical computer — from a smartphone to a 21-megawatt supercomputer — operates on the same foundational premise: information is represented as bits, each of which is either 0 or 1 at any given moment. This binary certainty is not a design choice engineers made for convenience; it is the logical outcome of transistor physics. A transistor is a switch. It is open or closed. There is no third state.
This works extraordinarily well for most computation. Classical machines can execute trillions of operations per second. The fastest supercomputer in 2025 — Frontier at Oak Ridge National Laboratory — sustains over 1.1 exaflops, or 1.1 × 10¹⁸ floating-point operations per second (TOP500, 2025). That is a staggering number, yet it does not help when the problem's solution space grows exponentially with its size.
Consider factoring a 2,048-bit RSA integer — the kind that secures modern HTTPS connections. The best classical algorithm known, the General Number Field Sieve, would require roughly 10²⁵ operations on a classical machine, making the problem computationally intractable for any realistic hardware. Searching an unstructured database of N items classically requires N/2 steps on average. Simulating a molecule with 50 interacting electrons requires tracking 2⁵⁰ quantum states simultaneously — about 10¹⁵ entries in a state vector that grows exponentially with each added electron.
These are not engineering limitations. They are mathematical walls. The scale of the problem's state space outpaces any imaginable improvement in classical clock speeds or parallelism.
Richard Feynman identified the core insight in 1981 during a lecture at MIT: "Nature isn't classical, dammit, and if you want to make a simulation of nature, you'd better make it quantum mechanical." Feynman proposed that a quantum system could simulate other quantum systems efficiently — something no classical machine could do — and effectively sketched the conceptual blueprint for quantum computing decades before the hardware existed to realize it.
The classical computer's limit is not that it is slow. It is that certain problem structures — factoring, unstructured search, molecular simulation, combinatorial optimization — have exponentially large solution spaces that no amount of classical speed can tame. Quantum computing attacks these problems differently: not by being faster at classical operations, but by using quantum mechanical phenomena to restructure the computation itself.
What a Qubit Actually Is
Qubit — a quantum bit; the fundamental unit of quantum information, implemented as any two-level quantum system whose state can exist in a superposition of |0⟩ and |1⟩ until measured.
A classical bit is a physical quantity with two stable states — high voltage or low voltage, magnetic north or south, a pit or a land on a disc. A qubit is also a physical quantity with two basis states, conventionally written |0⟩ and |1⟩ using Dirac notation. The crucial difference is that quantum mechanics permits — and indeed requires — a qubit's state to be a linear combination of both basis states simultaneously before measurement.
Mathematically, any pure qubit state is written as:
|ψ⟩ = α|0⟩ + β|1⟩
where α and β are complex numbers satisfying |α|² + |β|² = 1. The quantities |α|² and |β|² represent the probabilities of obtaining 0 or 1 upon measurement. A qubit is therefore not simply "0 and 1 at the same time" in a colloquial sense — it is a probability amplitude vector whose components can interfere constructively and destructively, the way waves do.
The Bloch sphere (illustrated above) is the standard geometric picture: every pure qubit state corresponds to a unique point on the surface of a unit sphere. The north pole is |0⟩, the south pole is |1⟩, and every other point on the surface represents a distinct superposition. A classical bit occupies only the two poles. A qubit can occupy any point on the entire sphere — an infinite family of states, though measurement collapses it to a pole.
| Property | Classical Bit | Qubit | Qutrit |
|---|---|---|---|
| Basis states | 0, 1 | |0⟩, |1⟩ | |0⟩, |1⟩, |2⟩ |
| Information per unit | 1 bit | 1 qubit (up to log₂ 3 ≈ 1.58 bits extractable) | log₂ 3 ≈ 1.58 bits |
| Intermediate states | None | Superposition of 2 basis states | Superposition of 3 basis states |
| Measurement outcome | Definite (0 or 1) | Probabilistic (collapses to 0 or 1) | Probabilistic (collapses to 0, 1, or 2) |
| Real hardware examples | CMOS transistor, DRAM cell | Superconducting transmon, trapped ion, photonic | Trapped-ion qutrits (Oxford, 2022); superconducting qutrits (IBM research) |
| Error sensitivity | Low (noise margins wide) | High (decoherence timescale: microseconds to milliseconds) | Higher still (three-level control more complex) |
| Gates | NOT, AND, OR, XOR | Hadamard, CNOT, T-gate, Toffoli | Generalised SU(3) gates |
| Commercial availability | Ubiquitous | IBM, Google, IonQ, Quantinuum, Rigetti | Experimental only (2026) |
Table 1: Classical bits, qubits, and qutrits compared across eight dimensions. Qutrits offer richer state spaces but remain pre-commercial as of 2026.
Physical implementations of qubits include superconducting transmon circuits (IBM, Google), trapped atomic ions held in electromagnetic Paul traps (IonQ, Quantinuum), photonic qubits encoded in the polarisation of single photons, and silicon spin qubits (Intel, various academic labs). Each trades different error characteristics, operating temperatures, and gate speeds. Superconducting qubits operate at roughly 15 millikelvin — colder than deep space — to suppress thermal noise. Trapped ions operate at room temperature in the trap itself but require ultrahigh vacuum.
The choice of qubit modality shapes every downstream engineering decision: how the chip is packaged, how gates are implemented, what error rates are achievable, and how many qubits can be integrated on a single device.
Superposition Explained Without Misleading Metaphors
Superposition — a quantum system's property of existing in multiple eigenstates simultaneously, described by a linear combination of basis states; it is a feature of the state vector, not a physical splitting of the system into copies.
The "Schrodinger's cat" metaphor — a cat simultaneously alive and dead — is useful for philosophical provocation but misleading for computation. The cat metaphor implies the system is in two distinct classical realities at once. That is not what superposition means mathematically.
Superposition is a property of the wave function, or state vector. Before measurement, the qubit is in a definite state — it is just not a classical state. It is a quantum state, characterised by probability amplitudes. The amplitudes encode not just probabilities but phases, and phases allow interference.
The Hadamard gate H is the simplest way to create superposition. Applied to |0⟩:
H|0⟩ = (|0⟩ + |1⟩) / √2
This produces an equal superposition: measuring will yield 0 or 1 with 50% probability each. But the computational value is not randomness — it is that this state, when fed into subsequent quantum gates, carries both possibilities forward in parallel through the circuit, with their amplitudes evolving coherently.
The interference aspect is the heart of quantum speedup. A quantum algorithm is designed so that amplitude pathways corresponding to wrong answers cancel (destructive interference) while pathways corresponding to right answers add (constructive interference). Grover's algorithm, which we examine below, is the clearest example of this deliberate interference engineering.
Decoherence — the loss of a qubit's quantum properties due to uncontrolled interactions with the environment; it is the primary reason quantum computations fail.
Erwin Schrodinger formalised the wave function in 1926. Werner Heisenberg independently developed matrix mechanics the same year. Both formalisms predict the same outcomes; they are mathematically equivalent. The probabilistic nature of quantum measurement — the fact that no local hidden variable theory can reproduce quantum correlations — was confirmed by Alain Aspect's Bell test experiments in 1982 and definitively closed by loophole-free tests in 2015 (Hensen et al., Nature, 2015).
What superposition gives a quantum computer is a coherent workspace in which all possible inputs to a function can be processed simultaneously — provided the computation finishes before decoherence destroys that workspace. That race against decoherence is the central engineering challenge of quantum hardware, a theme we return to in the error section.
Entanglement: The Resource That Makes Quantum Computing Work
Entanglement — a quantum correlation between two or more qubits such that the state of each cannot be described independently of the others, regardless of the physical distance between them.
If superposition gives each qubit an extended state space, entanglement links those spaces together. An entangled two-qubit system does not have four independent amplitudes for the four classical combinations (00, 01, 10, 11) in any separable way. Instead, its state is a single vector in a four-dimensional Hilbert space that cannot be factored into two individual qubit states.
The canonical example is the Bell state:
|Φ+⟩ = (|00⟩ + |11⟩) / √2
This state has no classical analogue. Measure the first qubit and get 0 — the second qubit is immediately in state |0⟩. Get 1 — the second is immediately |1⟩. The correlation is perfect and instantaneous, regardless of separation. Einstein called this "spooky action at a distance" and believed it indicated quantum mechanics was incomplete (Einstein, Podolsky & Rosen, 1935). John Bell proved in 1964 that if hidden local variables existed, the correlations would obey certain inequalities. Experiment has consistently violated those inequalities, confirming entanglement is a real, non-classical resource.
For computation, entanglement enables several things classical processors cannot replicate:
- Exponential state space compression. An n-qubit entangled register represents 2ⁿ amplitudes simultaneously. 300 entangled qubits span a state space larger than the estimated number of atoms in the observable universe (roughly 10⁸⁰).
- Correlated gate operations. The CNOT gate flips a target qubit conditioned on a control qubit — a quantum conditional that creates entanglement and is the building block of all multi-qubit algorithms.
- Quantum teleportation and error correction. Entanglement is the resource that allows quantum information to be moved and reconstructed across a quantum network, and is essential to the syndrome-based error correction schemes needed for fault-tolerant quantum computers.
The distinction between quantum parallelism and classical parallelism is often misunderstood. Classical parallel computing runs independent threads on separate cores, each with its own memory. Quantum parallelism runs a single computation on a superposition of inputs, all sharing the same physical register. The outputs interfere. You cannot simply "read out all results" — measurement collapses the state. Quantum algorithm design is the art of choreographing interference so the right answer rises to high probability by the time of measurement.
This is why quantum computers are not universally faster than classical ones. They are faster for specific problem structures that map onto interference patterns. For problems without that structure — playing chess, rendering graphics, sorting arrays — quantum computers offer no fundamental advantage.
The Four Algorithms That Define What Quantum Computers Can Actually Do
Not all quantum algorithms are created equal. Four algorithms have driven most of the theoretical and commercial excitement around quantum computing, each addressing a different problem class with a different type of speedup.
| Algorithm | Problem solved | Speedup type | Hardware requirements | Status (2026) |
|---|---|---|---|---|
| Shor's algorithm (Peter Shor, 1994) | Integer factorisation; discrete logarithm | Exponential (polynomial vs. sub-exponential classical) | Thousands of logical qubits with error correction (fault-tolerant) | Not yet practically realised; requires ~4,000 logical qubits for 2048-bit RSA — estimated 10M+ physical qubits with current error rates |
| Grover's algorithm (Lov Grover, 1996) | Unstructured database search | Quadratic (√N vs. N classical steps) | NISQ-compatible; demonstrated on small registers | Demonstrated experimentally; quadratic advantage real but modest for most databases |
| VQE — Variational Quantum Eigensolver (Peruzzo et al., 2014) | Ground-state energy of molecules and materials | Heuristic quantum-classical hybrid; no proven polynomial speedup yet | 50–1000 qubits, moderate error rates | Actively used on NISQ hardware; most mature application; used by pharmaceutical and materials companies |
| QAOA — Quantum Approximate Optimisation Algorithm (Farhi et al., 2014) | Combinatorial optimisation (scheduling, routing, portfolio, Max-Cut) | Heuristic; quantum advantage unproven for large instances | 50–500 qubits | Actively deployed experimentally; advantage over best classical heuristics not yet demonstrated at scale |
Table 2: The four defining quantum algorithms. "Logical qubits" refers to error-corrected qubits; physical qubit counts depend on the error rate of the underlying hardware.
The distinction between algorithmic speedup types matters enormously for commercial timelines:
Exponential speedup (Shor) is the most dramatic but requires fault-tolerant hardware that does not exist yet. Breaking RSA-2048 would require roughly 4,000 logical qubits, each protected by hundreds to thousands of physical qubits under current error correction schemes (Gidney & Eker, 2021, arXiv:1905.09749). At today's physical error rates of roughly 0.1–1% per gate, the overhead is prohibitive.
Quadratic speedup (Grover) is achievable sooner but is less transformative. Searching 10¹² items requires 10⁶ quantum steps versus 10¹² classical ones — a million-fold speedup sounds remarkable until you realise the same task takes about one millisecond on classical hardware and that Grover's circuit depth makes decoherence the limiting factor.
Heuristic advantage (VQE, QAOA) is where near-term quantum computers actually operate today. These hybrid algorithms offload some work to classical computers, iterate, and may offer practical advantages for specific problem instances before full fault tolerance is achieved. This is the domain of the NISQ era.
Quantum supremacy — a demonstration that a quantum device performs a specific computational task faster than any classical computer could, regardless of practical utility of that task.
Google's 2019 Sycamore result (Arute et al., Nature, 2019) claimed a 53-qubit processor completed a sampling task in 200 seconds that would take 10,000 years classically. IBM subsequently argued their classical algorithm could do it in 2.5 days. The definitional battle around "supremacy" vs. "advantage" persists, but the underlying point stands: quantum processors can now, for specific tasks, outperform classical simulation.
How Real Quantum Computers Are Built
Building a quantum computer is not primarily a software problem. It is a materials science, cryogenic engineering, and microwave control problem.
Superconducting Qubits — The IBM and Google Approach
IBM and Google both use transmon qubits — superconducting circuits based on a Josephson junction, a thin insulating barrier sandwiched between two superconducting materials (typically aluminium). When the circuit is cooled below the superconducting transition temperature, it exhibits discrete energy levels governed by quantum mechanics rather than classical electrodynamics.
The chip itself is about the size of a fingernail and is suspended at the bottom of a dilution refrigerator — a device that cools the sample to approximately 15 millikelvin (–273.135 °C), colder than the cosmic microwave background radiation (2.7 K). At that temperature, thermal fluctuations are suppressed enough that the quantum coherence of the circuit survives long enough to compute.
IBM's Heron processor, released in late 2024, integrates 156 qubits with a tunable-coupler architecture that reduces unwanted crosstalk between adjacent qubits, achieving two-qubit gate error rates of approximately 0.1–0.3% (IBM Research, 2024). Google's Willow chip (105 qubits, announced December 2024) achieved a landmark result: as more qubits were added for error correction, the error rate fell below the threshold required for scalable fault-tolerant computation — the first experimental demonstration of below-threshold performance in a superconducting device (Acharya et al., Nature, 2024).
Trapped-Ion Qubits — The IonQ and Quantinuum Approach
Trapped-ion quantum computers use individual atoms — typically Ytterbium-171 (IonQ) or Barium-133 (Quantinuum) — suspended in vacuum by oscillating electromagnetic fields in a Paul trap. Laser pulses or microwave radiation implement the qubit gates.
Trapped-ion systems achieve significantly lower error rates than superconducting qubits — Quantinuum's H-Series reported two-qubit gate fidelities of 99.8% in 2024 (Quantinuum press release, 2024) — but gate speeds are slower (microseconds vs. nanoseconds for superconducting) and scaling to hundreds of qubits requires either longer ion chains (which become unstable) or photonic interconnects between multiple trap modules. IonQ's roadmap targets 1,000 "algorithmic qubits" (a quality-adjusted measure) by 2028.
Photonic Qubits
Photonic quantum computers encode qubits in the properties of single photons — polarisation, time bins, or path. They operate at room temperature and interface naturally with optical fibre networks, making them attractive for quantum communication. PsiQuantum and Xanadu are the primary commercial entrants. Photonic systems face challenges in deterministic photon–photon interactions (photons do not naturally interact without nonlinear media), which complicates two-qubit gate implementation.
Silicon Spin Qubits
Intel and academic groups at TU Delft and UNSW Sydney pursue silicon spin qubits — quantum dots confining a single electron, with the qubit encoded in the electron's spin. The appeal is compatibility with existing CMOS fabrication infrastructure, potentially enabling millions of qubits on a chip. Coherence times and gate fidelities remain below those of superconducting and trapped-ion systems, but the scalability argument is compelling for the 10-year horizon.
For a related look at how quantum phenomena underpin chemistry and materials at the atomic scale, see our article on how atoms and molecules work.
The Error Problem — Where Quantum Computing Actually Stands in 2026
NISQ — Noisy Intermediate-Scale Quantum; the current era of quantum computing, characterised by processors with 50–1,000 physical qubits that lack sufficient error correction for fault-tolerant operation.
Quantum information is fragile. Every qubit interacts with its environment — stray magnetic fields, temperature fluctuations, microwave crosstalk from neighbouring qubits, cosmic rays. These interactions are collectively called decoherence, and they destroy superposition and entanglement on timescales measured in microseconds to milliseconds depending on the hardware.
The two principal error mechanisms are:
- T1 (relaxation time): How long before the qubit spontaneously decays from the excited state |1⟩ to the ground state |0⟩. Typically 50–500 microseconds for superconducting transmons, milliseconds for trapped ions.
- T2 (dephasing time): How long before phase coherence between |0⟩ and |1⟩ components is lost, scrambling the superposition. Often shorter than T1.
Physical gate error rates on leading hardware in 2026:
- IBM Heron: single-qubit gates ~0.02–0.05% error; two-qubit gates ~0.1–0.3% error
- Quantinuum H2: two-qubit gates ~0.15–0.2% error; single-qubit gates <0.01% error
- Google Willow: demonstrated below fault-tolerance threshold for surface code QEC
These error rates sound small but compound catastrophically. A circuit with 1,000 gate operations at 0.1% two-qubit error rate has roughly a 63% probability of at least one error (1 − 0.999¹⁰⁰⁰ ≈ 0.63). Running Shor's algorithm on a 2,048-bit key requires tens of millions of gate operations. Without error correction, the output is noise.
Quantum Error Correction
Quantum error correction (QEC) encodes one logical qubit across many physical qubits, distributing information redundantly so that errors on individual physical qubits can be detected and corrected without measuring — and thereby collapsing — the logical qubit's state. The surface code is the leading QEC scheme for superconducting hardware. A distance-d surface code uses roughly d² physical qubits per logical qubit and corrects up to ⌊(d–1)/2⌋ errors per round.
The resource overhead is severe. Correcting Shor's algorithm to the precision needed to factor RSA-2048 reliably is estimated to require approximately 20 million physical qubits at current error rates (Babbush et al., PRX Quantum, 2021). Leading machines have hundreds of physical qubits, not millions.
Google's Willow result in December 2024 was significant precisely because it demonstrated that adding more qubits to the surface code actually reduced the logical error rate — a prerequisite for scalable fault tolerance. Below the threshold, more physical qubits means more reliable logical qubits. Above the threshold, adding qubits adds more errors than the code can correct. Willow appears to have crossed below the threshold, at least for the surface code metrics tested.
The industry consensus for the timeline to fault-tolerant, cryptographically relevant quantum computers is roughly the mid-2030s to 2040s, depending on the rate of progress in fabrication yield and error rates. The NISQ era — where quantum advantage may appear for specific tasks without full error correction — is likely to extend through the late 2020s.
For further context on quantum mechanics foundations, see quantum mechanics explained: the physics of the very small and our deep dive on quantum entanglement explained simply.
What Quantum Computers Can and Cannot Do
The most persistent misconception about quantum computing is that it will eventually replace classical computing across the board. It will not. Quantum computers are a specialised tool for a narrow but important class of problems.
What quantum computers are good at:
- Problems with exponentially large structured solution spaces: Factorisation (Shor), lattice problems, discrete logarithm. These have proven exponential speedups.
- Unstructured search: Grover's algorithm provides a quadratic speedup — real but modest.
- Quantum simulation: Simulating molecules, materials, and quantum field theories. Lithium-ion battery cathodes, nitrogen fixation enzymes, high-temperature superconductors — all involve quantum effects that are tractable on quantum hardware but exponentially expensive classically. IBM and Microsoft have active pharmaceutical and materials partnerships.
- Certain optimisation problems: Scheduling, portfolio optimisation, traffic routing — QAOA may provide practical advantage for certain instance structures, though this remains actively debated.
- Quantum cryptography and quantum networking: Quantum key distribution (QKD) uses quantum properties to detect eavesdropping with information-theoretic security. This is deployable now with current hardware.
What quantum computers are not good at, and never will be:
- General-purpose computation: Browsing the web, running an operating system, playing video games, training large language models. No quantum speedup exists for these tasks.
- Tasks that are already fast classically: Sorting, basic arithmetic, most database queries. Grover's speedup is quadratic — if a database query takes 1 nanosecond classically, Grover saves you the square root of that factor.
- Big data analytics: Classical GPUs process petabytes of data efficiently. Quantum computers have no classical memory bandwidth advantages.
- Any problem where the answer cannot be amplified by interference: The algorithm design problem remains open for many practically important optimisation problems.
The honest near-term picture: quantum computers in 2026 are research instruments and early-stage industrial prototypes. Pharmaceutical companies are running VQE experiments to model molecular interactions. Financial institutions are exploring QAOA for portfolio optimisation. National security agencies are preparing post-quantum cryptography standards — NIST finalised its first post-quantum cryptographic standards in 2024 (NIST, FIPS 203, 204, 205) — precisely because Shor's algorithm will eventually threaten RSA and ECC.
The comparison to classical computing in the 1950s is apt but imprecise. Early vacuum-tube computers were expensive, unreliable, and limited to a handful of tasks. They scaled. Quantum computers may scale similarly — or they may plateau at NISQ capabilities and serve a specialised niche alongside classical computing rather than supplanting it.
For a broader view of how computing and physics have intersected historically, see our article on Alan Turing and the foundations of computing and artificial intelligence. The story of quantum computing's conceptual roots connects directly to the history of the atom from Dalton to quantum mechanics.
Frequently Asked Questions About Quantum Computing
Q: What is a quantum computer in simple terms? A: A quantum computer is a machine that uses the quantum mechanical properties of particles — superposition and entanglement — to perform computations. Instead of processing one state at a time like a classical computer, it manipulates probability amplitudes across many states simultaneously, engineering interference so correct answers become more probable.
Q: How many qubits does a useful quantum computer need? A: For heuristic near-term tasks (molecular simulation, optimisation), 100–1,000 high-quality physical qubits may provide practical value. For cryptographically relevant tasks like breaking RSA-2048 via Shor's algorithm, estimates range from 4,000 error-corrected logical qubits to 20 million physical qubits under current surface-code error rates (Babbush et al., PRX Quantum, 2021).
Q: Is a quantum computer faster than a classical computer? A: Not universally. Quantum computers offer provable speedups only for specific problem types — factoring integers, searching unstructured data, simulating quantum systems. For most everyday computing tasks, a classical processor is equally fast or faster, and far more practical.
Q: What is quantum supremacy and has it been achieved? A: Quantum supremacy means a quantum device outperforms any classical computer on a specific task. Google's Sycamore processor claimed this in 2019 for a random circuit sampling problem. The claim is contested for practical workloads, but the underlying hardware capability has advanced significantly since then, with Google's Willow chip in 2024 achieving below-threshold error correction for the first time.
Q: Why do quantum computers need to be so cold? A: Superconducting qubits must operate near absolute zero — around 15 millikelvin — to eliminate thermal fluctuations that would collapse quantum superposition in nanoseconds. At room temperature, thermal energy (~25 meV) vastly exceeds the energy spacing between qubit levels (~0.1 meV for transmons), making quantum operation impossible.
Q: Will quantum computers break encryption? A: Eventually, fault-tolerant quantum computers running Shor's algorithm could break RSA and ECC encryption. This is not an immediate threat — current machines lack the error-corrected qubit count needed. NIST finalised post-quantum cryptography standards in 2024 (FIPS 203 using CRYSTALS-Kyber, FIPS 204 using CRYSTALS-Dilithium) specifically to address this future threat.
Q: What is the difference between NISQ and fault-tolerant quantum computing? A: NISQ (Noisy Intermediate-Scale Quantum) describes current machines with 50–1,000 physical qubits that lack sufficient error correction for deep, reliable circuits. Fault-tolerant quantum computing uses quantum error correction to protect logical qubits from physical errors, enabling arbitrarily long computations. The crossover requires physical error rates below ~1% per gate and millions of physical qubits for practically useful algorithms.
Q: What companies are leading quantum computing in 2026? A: IBM (superconducting, Heron 156-qubit processor, public cloud access via IBM Quantum), Google (superconducting, Willow 105-qubit chip, below-threshold QEC milestone), IonQ (trapped ion, targeting 1,000 algorithmic qubits by 2028), Quantinuum (trapped ion, highest gate fidelities commercially available), Rigetti (superconducting, cloud access), PsiQuantum (photonic, targeting silicon fab production), and Microsoft (topological qubits based on Majorana zero modes, first announced hardware milestone in 2025).
Sources
Arute, F. et al. (2019). "Quantum supremacy using a programmable superconducting processor." Nature, 574, 505–510. https://doi.org/10.1038/s41586-019-1666-5
Acharya, R. et al. (Google Quantum AI, 2024). "Quantum error correction below the surface code threshold." Nature, 638, 920–926. https://doi.org/10.1038/s41586-024-08449-y
Babbush, R. et al. (2021). "Focus beyond quadratic speedups for error-corrected quantum advantage." PRX Quantum, 2, 010103. https://doi.org/10.1103/PRXQuantum.2.010103
Gidney, C. & Eker, M. (2021). "How to factor 2048 bit RSA integers in 8 hours using 20 million noisy qubits." Quantum, 5, 433. https://doi.org/10.22331/q-2021-04-15-433
Hensen, B. et al. (2015). "Loophole-free Bell inequality violation using electron spins separated by 1.3 kilometres." Nature, 526, 682–686. https://doi.org/10.1038/nature15759
IBM Research (2024). IBM Heron Processor Technical Specifications. https://research.ibm.com/blog/heron-processor
NIST (2024). FIPS 203, 204, 205: Post-Quantum Cryptography Standards. https://csrc.nist.gov/pubs/fips/203/final
Peruzzo, A. et al. (2014). "A variational eigenvalue solver on a photonic quantum chip." Nature Communications, 5, 4213. https://doi.org/10.1038/ncomms5213
TOP500 (2025). TOP500 Supercomputer Sites — June 2025 List. https://www.top500.org/lists/top500/2025/06/
Quantinuum (2024). H-Series Quantum Processor Performance Report. https://www.quantinuum.com/hardware
Feynman, R. P. (1982). "Simulating physics with computers." International Journal of Theoretical Physics, 21, 467–488. https://doi.org/10.1007/BF02650179
Farhi, E., Goldstone, J. & Gutmann, S. (2014). "A Quantum Approximate Optimization Algorithm." arXiv:1411.4028.
Follow new articles
Add SciHub101 to your feed reader to see new science and technology stories.
About the Author
SciHub101 is an independent science and technology publication founded and edited by Buğra Sözeri, a software engineer and app developer based in Türkiye. AI tools assist with drafting and research. Source documentation in the existing archive is uneven, and articles are being reviewed and corrected. The Editorial Team byline identifies editorial responsibility; it does not certify specialist or independent human review. Where we compare products, anything built by our own team is labelled as such — see the editorial guidelines and About page for how corrections and disclosures work.
Related Articles

Nuclear Fusion in 2026: Is It Finally Working?
A comprehensive look at the state of nuclear fusion in 2026 — the NIF ignition breakthrough, ITER tokamak, stellarators, inertial confinement, and the private companies racing to deliver fusion power to the grid.

Cosmic Rays and Particle Physics — Sources, Detection, and the Oh-My-God Particle
What cosmic rays are, how Victor Hess discovered them in 1912, their sources from the Sun to AGN, detection methods, biological effects, and the ultra-high-energy mystery.

Quantum Error Correction Explained: Physical Qubits, Logical Qubits and Thresholds
Quantum error correction protects one logical qubit across many noisy physical qubits. Learn how syndromes, surface codes, and thresholds work in practice.

How Do Quantum Computers Actually Work? Beyond the Hype
Quantum computers don't replace your laptop — they solve a small set of problems exponentially faster. Here's what they really are, why they're hard to build, and what they'll change.

The Most Important Scientific Discoveries of 2025: Breakthroughs and Their Impact
2025 marked significant scientific progress across multiple disciplines: a structural biology breakthrough in protein folding, advances in quantum computing error correction, discovery of a novel class of high-temperature superconductors, clinical trials demonstrating efficacy of novel Alzheimer's disease therapies, successful gene therapy applications for hereditary blindness, and confirmation of predicted gravitational wave signatures. This article examines the most consequential discoveries of 2025 — those that represent genuine advances in human knowledge, demonstrate technological capability, or have immediate clinical applications — explaining the science, why each discovery matters, and the implications for future research.

Quantum Mechanics Foundations Explained — Wave, Uncertainty, and Entanglement
Quantum mechanics rests on five foundational principles that explain all atomic and subatomic behavior. This primer covers each principle, its experimental basis, and what it means.