Quantum Computing Hardware

Superconducting Quantum Processors

Superconducting qubits are the most widely deployed quantum computing technology, used by IBM, Google, Rigetti, IQM, and others. The quantum processor is a chip roughly 1 to 4 centimeters on a side, fabricated on a silicon or sapphire substrate using thin-film deposition and photolithographic patterning similar to semiconductor manufacturing. Each qubit is a tiny superconducting circuit containing a Josephson junction, a nanometer-thin insulating barrier between two superconducting aluminum electrodes. The Josephson junction creates a nonlinear inductance that gives the circuit discrete energy levels, making it behave like an artificial atom with well-defined |0> and |1> states.

The entire chip must operate at approximately 15 millikelvin, about 0.015 degrees above absolute zero, because thermal energy at higher temperatures would overwhelm the quantum energy scales of the qubits (which operate at microwave frequencies of 4 to 6 GHz, corresponding to photon energies equivalent to about 200 millikelvin). A dilution refrigerator achieves these temperatures through a multi-stage cooling process. The outermost stage uses conventional pulse tube coolers to reach 4 kelvin. Inner stages use the mixing of helium-3 and helium-4 isotopes to cool progressively to 15 millikelvin. The refrigerator is a cylinder roughly 1 meter in diameter and 3 meters tall, with the quantum chip mounted at the bottom of an inverted tree structure called the chandelier.

Hundreds of coaxial cables carry microwave control signals from room-temperature electronics down to the quantum chip. Each qubit requires at least one control line for single-qubit gates and shares coupling lines with neighboring qubits for two-qubit gates. Readout resonators, one per qubit, allow the qubit state to be measured by probing the resonator's response at its resonant frequency. The cables pass through multiple thermal stages with attenuators at each stage to prevent thermal noise from reaching the chip. Signal amplification on the readout path uses quantum-limited amplifiers (Josephson parametric amplifiers) that add the minimum noise allowed by quantum mechanics.

The room-temperature electronics stack generates the precisely shaped microwave pulses that implement gate operations. Each single-qubit gate is a microwave pulse of specific frequency (matching the qubit's resonance), amplitude (determining the rotation angle), duration (typically 20 to 50 nanoseconds), and phase (determining the rotation axis). Arbitrary waveform generators produce these pulses, which are mixed with carrier signals, filtered, attenuated, and routed to the correct qubit through the cryogenic wiring. The timing precision required is sub-nanosecond, and amplitude control must be accurate to fractions of a percent to achieve gate fidelities above 99.9%.

Trapped Ion Systems

Trapped ion quantum computers confine individual atomic ions in electromagnetic traps and use laser beams to manipulate their quantum states. The trap is a set of precisely machined electrodes that create oscillating electric fields (radio-frequency traps) confining the ions in a linear chain. The ions are separated by approximately 5 micrometers, and the trap operates in an ultra-high vacuum chamber at pressures below 10^-11 torr (roughly a trillionth of atmospheric pressure) to prevent collisions with background gas molecules that would disrupt the quantum states.

Qubit states are encoded in two energy levels of the ion's electronic structure. For ytterbium-171 (used by IonQ), these are two hyperfine levels of the ground state, separated by 12.6 GHz. For barium-137 (used by some academic systems), optical transitions between the ground and a metastable excited state serve as the qubit levels. Single-qubit gates are performed by Raman laser pulses that drive transitions between the qubit levels. The gate time depends on the laser intensity and the specific transition used, typically 1 to 100 microseconds.

Two-qubit entangling gates use the shared motional modes of the ion chain as a communication bus. A laser pulse on one ion couples its internal state to the collective motion of the chain (the ions vibrate together like beads on a spring). A subsequent pulse on another ion picks up this motional information, creating an entangling interaction between the two ions' internal states. The Molmer-Sorensen gate and the light-shift gate are the most common entangling gate implementations. Because the motion couples all ions in the chain, any pair of ions can interact directly, giving trapped ion systems all-to-all connectivity without the SWAP gate overhead that grid-connected superconducting processors require.

Scaling trapped ion systems beyond a few dozen ions in a single chain requires multi-zone trap architectures. The Quantum Charge-Coupled Device (QCCD) architecture, developed at NIST and commercialized by Quantinuum, uses a trap with multiple zones connected by junctions. Small groups of ions are shuttled between zones using time-varying electrode voltages. Computation occurs within individual zones, and ions are physically moved to bring qubits together for two-qubit gates. Quantinuum's H2 processor has demonstrated 56 fully connected qubits using this approach, with two-qubit gate fidelities exceeding 99.8%.

Neutral Atom and Photonic Systems

Neutral atom quantum computers trap individual atoms in arrays of optical tweezers, tightly focused laser beams that hold atoms at their focal points through the gradient force. Atoms are loaded from a laser-cooled cloud and arranged into arbitrary 2D or 3D geometries by controlling the positions of the tweezer beams using spatial light modulators or acousto-optic deflectors. Arrays of 200+ atoms have been demonstrated, and the number is limited primarily by the available laser power rather than by any fundamental physical constraint.

Two-qubit gates in neutral atom systems exploit Rydberg interactions. When an atom is excited to a Rydberg state (a very high-energy electronic state with a large electron orbit), it develops a strong electric dipole moment that interacts with nearby Rydberg atoms. This interaction creates a controlled-phase gate between neighboring atoms. The Rydberg blockade mechanism, where the Rydberg excitation of one atom shifts the energy levels of its neighbors enough to prevent their simultaneous excitation, naturally implements a CNOT-like operation. Gate fidelities above 99.5% have been demonstrated, and the gate speed (roughly 1 microsecond) is intermediate between the fast superconducting gates and the slower trapped ion gates.

Photonic quantum computers use individual photons as qubits, with information encoded in polarization, time-bin, or path degrees of freedom. Xanadu's Borealis processor demonstrated quantum advantage using Gaussian boson sampling with 216 squeezed-state photon modes. PsiQuantum is pursuing a different approach: building a million-qubit fault-tolerant photonic quantum computer using silicon photonics manufacturing. Their architecture uses a measurement-based computation model where a large entangled state (a cluster state) is generated from photon sources and linear optical networks, and computation is performed entirely through single-photon measurements. The key advantage of photonic qubits is room-temperature operation and the potential for manufacturing using existing semiconductor fabrication infrastructure.

Calibration and the Daily Reality of Quantum Hardware

Quantum processors require constant calibration because qubit properties drift over time. A superconducting qubit's resonant frequency can shift by megahertz over hours due to two-level system defects in the substrate, changing material properties, and electromagnetic interference. These shifts change the optimal gate pulse parameters, degrading gate fidelity if not corrected. Quantum computing teams run automated calibration routines daily (or more frequently), measuring each qubit's frequency, T1 relaxation time, T2 coherence time, single-qubit gate fidelity, and two-qubit gate fidelity, then adjusting the control parameters accordingly.

Crosstalk between qubits is a persistent hardware challenge. When a gate pulse is applied to one qubit, it can partially affect neighboring qubits through unwanted electromagnetic coupling. This cross-talk is characterized through detailed experiments that measure how each qubit responds to pulses intended for every other qubit, producing a cross-talk matrix that calibration routines use to apply compensating corrections. As qubit counts increase, the number of cross-talk interactions grows quadratically, making calibration increasingly complex.

The yield problem affects manufacturing scalability. Not all qubits on a fabricated chip meet performance specifications. Some qubits have coherence times too short for useful computation, some qubit pairs have entangling gate fidelities below acceptable thresholds, and some qubits exhibit frequency collisions (their resonant frequencies are too close to neighboring qubits, causing unwanted interactions). Current processors typically have 5% to 15% of qubits or qubit pairs that are unusable, and circuit compilers must route computations around these defective components. Improving fabrication consistency to reduce defect rates is critical for scaling to larger processors.

Scaling Roadmaps

Every major quantum hardware company has published a roadmap projecting processor sizes over the next 5 to 10 years. IBM's roadmap progresses from Heron (133 to 156 qubits, 2024) through Flamingo (modular architecture, 2025) to Starling (fault-tolerant, 2029) and Blue Jay (100,000+ qubits, 2033). Google aims for a million-qubit error-corrected processor by the early 2030s. Microsoft is pursuing topological qubits that would require far fewer physical qubits per logical qubit, potentially reaching fault-tolerant computation with fewer total qubits. IonQ projects scaling trapped ion systems to thousands of qubits through modular architectures with photonic interconnects between multiple ion trap modules.

The path to million-qubit processors requires solving interconnected challenges in fabrication (making reliable qubits at scale), wiring (routing signals to millions of qubits without adding thermal noise), control electronics (generating and routing millions of independent gate pulses), and classical processing (decoding error correction syndromes fast enough to keep up with the quantum processor). Modular architectures, where multiple smaller processors are connected by quantum interconnects, are widely seen as more practical than building monolithic chips with millions of qubits. The quantum interconnects (optical links, superconducting links, or ion shuttling between modules) add latency and reduce fidelity compared to local operations, creating a performance hierarchy analogous to the cache/memory/disk hierarchy in classical computing.