computing platform profile
Superconducting quantum processors
Superconducting processors use patterned electrical circuits as controllable artificial atoms. Early systems are available through cloud research services, although useful fault-tolerant machines do not yet exist.
Technical deep dive
How it works
Josephson junctions add a nonlinear element to a cold resonant circuit, isolating two energy levels that act as a qubit. Microwave pulses perform gates, tunable couplers mediate interactions, and resonators support readout. Scaling requires better materials and packaging, lower correlated error, extensive cryogenic wiring, and error correction that converts many physical qubits into reliable logical qubits.
An artificial atom made from a circuit
A Josephson junction passes Cooper-pair current through a thin insulating barrier and supplies a nonlinear inductance. Combined with a capacitor, it creates an anharmonic oscillator: the two lowest energy levels can act as |0⟩ and |1⟩ without a control pulse equally exciting every higher level. The common transmon design trades lower sensitivity to charge noise for relatively weak anharmonicity.
A dilution refrigerator cools the chip to millikelvin temperature so thermal excitation is far below the microwave transition energy. Patterned aluminum, niobium, or tantalum films sit on sapphire or silicon; dielectric defects, interfaces, radiation, magnetic vortices, and nonequilibrium quasiparticles can all absorb energy or randomize phase.
Control, coupling, and readout
Shaped microwave pulses rotate individual qubits. Fixed or tunable capacitive couplers mediate two-qubit interactions; flux bias can tune qubit or coupler frequency but introduces another noise path. Leakage into higher transmon levels, residual coupling, pulse distortion, crosstalk, and frequency collisions grow more difficult as devices become dense.
For readout, each qubit shifts the resonance of a coupled microwave resonator depending on its state. Amplifiers measure the outgoing field, and room-temperature electronics classify the result. Readout is not passive: measurement photons, reset time, amplifier noise, and multiplexing all enter an error-correction cycle.
Logical qubits and the threshold test
A surface code arranges data and measurement qubits on a lattice. Repeated parity checks reveal syndromes without directly measuring the encoded state; a classical decoder infers likely error chains. Code distance d sets the minimum chain length able to cause an undetected logical error and ordinarily requires a growing patch of physical qubits.
Below threshold, increasing code distance should reduce logical error despite adding more imperfect components. That is a systems result, but a logical memory is still not a universal processor: fault-tolerant logical gates, state injection, routing, decoding latency, leakage removal, and long algorithmic depth must work at the same time.
Reason to use it
What distinguishes the approach
Fast gates, lithographic fabrication, mature microwave control, and an extensive experimental ecosystem make this one of the most developed hardware platforms.
Unresolved engineering
Main scaling problem
Reducing errors and control overhead faster than systems scale
Monthly evidence layer
Research watch
Named programs, dated developments, and primary or official sources. Company claims are identified separately from independent or peer-reviewed evidence.
Last reviewed 6 September 2026
Research groups & programs
- Google Quantum AI
Superconducting processors, surface-code memories, decoding, fabrication, and logical operations.
- IBM Quantum
Modular superconducting processors, dynamic circuits, error mitigation, and fault-tolerant architecture.
Current developments
December 2024
A larger surface code reduced logical errorThe distance-7 memory supplied below-threshold evidence: adding the physical qubits needed for a larger code lowered the measured logical error.
Next frontier
Move from protected memory to protected computationLogical gates, magic-state resources, fast decoding, leakage control, and modular scaling must preserve the suppression seen in memory experiments.
Papers & official resources
- paperQuantum error correction below the surface-code threshold
Nature · 2024
- official resourceMaking quantum error correction work
Google Research · 2024