Технологии / Железо
Железо: сравнение модальностей
The reference table. Seven ways to build a qubit plus the software layer above them. Physical principle, strengths, challenges, maturity, and who is pursuing each. Quantitative specs change quarterly, so this table carries honest unknowns rather than invented numbers.
| Формат | Как реализован кубит | Преимущества | Сложности | Зрелость | Компании |
|---|---|---|---|---|---|
| Superconducting circuits | Microwave-controlled electrical circuits at millikelvin temperatures | Fast gates · Semiconductor-style fabrication · Mature ecosystem | Cryogenics · Wiring · Crosstalk · Coherence | early-commercial (cloud access); fault tolerance in progress | IBM · Google Quantum AI · Rigetti Computing · IQM · Fujitsu / RIKEN · AWS (Amazon) · Alice & Bob |
| Trapped ions | Electromagnetically trapped charged atoms controlled with lasers | High-fidelity operations · Long coherence · Strong connectivity | Slower gates · Optical complexity · Scaling ion control | early-commercial (cloud access); scaling architectures in progress | IonQ · Quantinuum · Alpine Quantum Technologies · Universal Quantum |
| Neutral atoms | Laser-trapped neutral atoms arranged in programmable arrays | Large arrays · Flexible geometry · Promising scaling | Gate fidelity · Atom loss · Control and readout | prototype → early-commercial | QuEra · Pasqal · Atom Computing · Infleqtion · planqc |
| Photonic | Single photons and optical modes | Room-temperature transmission · Networking potential | Photon loss · Deterministic sources · Detectors · Large overhead | prototype; betting on a leap to fault tolerance | PsiQuantum · Xanadu · Quandela · ORCA Computing |
| Silicon spin qubits | Electron or nuclear spins in semiconductor devices | Tiny footprint · Compatibility with chip manufacturing | Uniformity · Control · Cryogenic integration | research → prototype | Intel · Diraq · Quantum Motion · Silicon Quantum Computing · Quobly |
| Topological | Quantum information protected by exotic quasiparticle states | Potentially intrinsic error resilience | Underlying physics and scalable implementation remain difficult | research | Microsoft |
| Quantum annealing | Analog evolution toward low-energy solutions (a distinct paradigm, not gate-based) | Commercial access · Large device sizes | Not equivalent to universal gate-model computing · Advantage is problem-specific | commercial for specific optimization workloads | D-Wave |
| Software & error suppression | The hardware-agnostic layer: control, error suppression, algorithm design, middleware | Wins whichever qubit recipe prevails | Value depends on the hardware ecosystem maturing | commercial today; grows with every machine shipped | Q-CTRL · Classiq · Multiverse Computing |
Gate speeds, fidelities, coherence times and qubit counts are deliberately absent: they change quarterly and vendor figures are not independently comparable. Where a spec matters, check the primary sources on each modality page. Treat "unknown" as the honest state, not a gap.
Physical vs logical architecture
A complete system also requires cryogenics or vacuum, control electronics, calibration software, compilers, networking, and classical compute. The race that matters is not toward bigger chips but toward reliable logical operations at useful scale. "Below threshold" behavior, where growing the code shrinks the logical error rate.
Взгляд инвестора: each row is a different bet with a different risk shape, and no modality has conclusively won. The same map with tickers and framing lives on the страница инвестиций; более наглядная версия в карточках — Разные квантовые компьютеры.
Углубиться (по 5 минут)
Хочешь закрепить материал? The Academy covers hardware at every tier: from the beginner hardware module to advanced systems architecture.
Начать весёлые уроки → Бесплатно · без оценок, без давления · весёлые викторины с неограниченным числом попыток