Building a quantum computer is not just a software problem. The physical platform that stores and manipulates qubits determines almost everything: operating temperature, gate speed, error rates, and how far the system can scale. Three approaches dominate the current landscape: superconducting qubits, trapped ions, and photons.
## Superconducting Qubits: The Industrial Front-Runner
IBM and Google have both committed to superconducting qubits, which are microwave-frequency circuits cooled to about 15 millikelvin — colder than outer space. At that temperature, specially designed Josephson junctions exhibit quantum behavior between two energy levels, acting as artificial atoms.
The advantages are compelling. Gate operations take nanoseconds, manufacturing leverages semiconductor fabrication techniques, and the approach has already produced landmark results: Google’s Sycamore chip completed a quantum computation in 200 seconds that Google estimated would take a classical supercomputer 10,000 years (though that estimate has been contested). IBM’s roadmap targets fault-tolerant systems by the late 2020s.
The challenges are equally real. The required dilution refrigerators are expensive and bulky. Qubit coherence times are measured in microseconds, placing strict limits on circuit depth. And chip connectivity is constrained by physical wiring.
## Trapped Ions: Highest Fidelity, Slower Speed
IonQ and Quantinuum suspend individual ions — typically ytterbium or barium isotopes — in electromagnetic traps and use laser pulses to perform gate operations. Because the qubits are identical natural atoms, their properties are extraordinarily uniform.
Trapped-ion systems routinely achieve two-qubit gate fidelities above 99.5%, coherence times measured in minutes, and full all-to-all connectivity among qubits. Quantinuum’s H2 processor demonstrated 56 logical qubits in 2024 with the lowest logical error rates reported at the time. See [Quantinuum’s documentation](https://www.quantinuum.com/hardware) for current benchmarks.
The trade-off is speed: ion gates take milliseconds, roughly a million times slower than superconducting gates. Scaling beyond a few hundred ions while maintaining laser control is a hard engineering problem.
## Photonic Quantum Computing: Room-Temperature Promise
PsiQuantum and Xanadu use individual photons as qubits, routed through optical circuits. Because photons travel at the speed of light and barely interact with their environment, they are naturally long-lived and easy to transmit — but notoriously difficult to make interact with each other.
Xanadu’s [Strawberry Fields](https://strawberryfields.ai/) platform and its Borealis system demonstrated photonic quantum advantage in 2022. PsiQuantum is building toward a million-qubit system fabricated using standard semiconductor foundries, with photon detection integrated on chip.
Room-temperature operation is attractive for data centers, but deterministic photon-photon gates remain an open research problem. Current approaches rely on measurement-based schemes that require enormous numbers of photons to correct for loss.
## Other Contenders
**Neutral atoms**: QuEra and Pasqal trap arrays of neutral atoms with optical tweezers and use Rydberg state interactions for two-qubit gates. QuEra achieved 48 logical qubits in 2023. The platform combines long coherence times with programmable connectivity. See [QuEra’s research page](https://www.quera.com/research).
**Topological qubits**: Microsoft’s long-term bet is on qubits protected by topology — specifically Majorana zero modes — that would be inherently resistant to decoherence. In 2025, Microsoft announced the Majorana 1 chip, though independent validation of topological qubit behavior is ongoing.
## No Clear Winner Yet
Each platform is closest to quantum advantage in a different regime. Superconducting systems excel at fast, deep circuits; trapped ions at high-fidelity, fully connected operations; photons at communication and networking tasks. The likeliest outcome is that multiple platforms mature in parallel, with applications driving which architecture dominates in each use case.
For context on what quantum computers can actually do, see [Quantum Algorithms Explained](https://sunqi.org/quantum-algorithms-en/) and the overview at [arxiv:quant-ph](https://arxiv.org/list/quant-ph/recent).
—
