💻 Quantum Computing Industry

Quantum Computing Industry

A researcher's guide to the hardware race — platforms, companies, metrics, and where neutral atoms fit in. From NISQ to fault tolerance: superconducting, trapped-ion, neutral-atom, silicon-spin, photonic, and topological qubits. Updated to May 2026.

6 Platforms DiVincenzo Criteria Radar Comparison Roadmap Timeline 25+ Companies AMO → QC Career
0
Qubit Platforms
0+
Leading Companies
0
Largest Coherent Atom Array
0.9%
Best 2Q Fidelity
0
Logical Qubits Demonstrated

2024–2026 Flash Points

The developments that moved the field — from lab breakthroughs to billion-dollar bets.

Dec 2024
🌟

Google Willow — Below-Threshold QEC

Google's Willow processors demonstrated below-threshold surface-code memories up to distance 7: logical error per cycle decreased as code distance increased, and the distance-7 memory exceeded the lifetime of its best physical qubit. Nature 2025.

Feb 2025

Microsoft Majorana 1 — Topological Qubit Chip

Microsoft announced the Majorana 1 chip, an InAs/Al heterostructure device aimed at topological qubits. The potential overhead reduction is large, but independent confirmation of topological qubit operation and braiding remains open; treat this as a high-risk frontier rather than a settled platform.

2025
💰

Quantinuum H2 — 56 Trapped-Ion Qubits

Quantinuum launched the 56-qubit H2-1 trapped-ion processor in 2024 and reported record quantum volume 33,554,432 in 2025. Its strength remains high-fidelity all-to-all gates, with scaling limited by gate speed and modular ion transport. Quantinuum 2024.

Nature 2024
⚛️

Harvard/QuEra — 48 to 96 Logical Qubits

Bluvstein et al. demonstrated a 48-logical-qubit reconfigurable neutral-atom processor in Nature 2024. A 2025 follow-up used up to 448 atoms to demonstrate key pieces of a universal fault-tolerant neutral-atom architecture with 96 logical qubits. Nature 2024; Nature 2025.

2024–25
🔧

IBM Roadmap — Utility to FTQC

IBM's public roadmap now emphasizes quality, modularity, and fault tolerance over raw qubit count: Heron-class 156-qubit processors now, Nighthawk/Kookaburra milestones in 2026, and Starling targeted for 2029 with 200 logical qubits and 100 million gates. IBM roadmap. company roadmap

2023–25
🏛️

PsiQuantum — Government Mega-Funding

Secured AUD $940 million from the Queensland government plus US DARPA and DoE contracts, betting on silicon photonics as the only fab-compatible path to fault-tolerant scale. Total private + public backing now exceeds $1.5 B.

The NISQ Era — And What Comes Next

Where the field stands in 2025–2026 and the precise gap to fault tolerance.

🎯 The End Goal

Fault-tolerant quantum computers with logical qubits running Shor's algorithm on RSA-2048 (~20 million physical qubits at 10⁻³ error rate — Gidney & Ekerå 2021) and simulating industrial molecules (FeMoco, P450) with ~100 logical qubits at <10⁻⁴ error rates.

📍 Status (2025–2026)

The field is transitioning from NISQ demonstrations to early fault-tolerance experiments. Google Willow crossed the below-threshold surface-code-memory milestone. Harvard/QuEra moved neutral atoms from 48 logical qubits in 2024 to a 96-logical-qubit universal fault-tolerant architecture demonstration in 2025. IBM and Quantinuum are pursuing utility-scale and logical-processor roadmaps. Physical qubit counts: tens to thousands, depending on whether one counts gate-operated qubits, loaded atoms, or annealer flux qubits.

🚧 The Gap

Surface codes need <10⁻³ two-qubit error rates to operate below threshold — current best physical gate fidelity is ~99.9% (trapped ions, Quantinuum 2023). Breaking RSA-2048 still requires ~20 million physical qubits, roughly 4–5 orders of magnitude beyond today's largest systems. Estimated timeline to cryptographically relevant QC: 2035–2040.

The Qubit Race

Physical qubit count per platform (log scale, 2025 data) — animated buildup toward fault tolerance threshold.

D-Wave counts are flux-qubit annealers (not gate-based). PsiQuantum bar = projected photonic-mode target. Dashed red = order 10⁷ physical qubits for cryptographically relevant RSA-2048 estimates, depending strongly on code, cycle time, and magic-state factory assumptions.

The DiVincenzo Criteria

Five necessary (and simultaneously required) conditions for a viable qubit platform.

DiVincenzo (2000), A physical system is a viable qubit platform only if it satisfies all five criteria simultaneously, at scale, and with fault-tolerance margins. No platform has fully satisfied all five yet — this is what makes quantum computing hard.
1️⃣
Scalable Register

Well-defined |0⟩,|1⟩ two-level subspace, replicable in large numbers without performance degradation.

2️⃣
Initialisation

Reliable preparation of the fiducial |000…0⟩ state before each computation. Requires optical pumping + ground-state cooling.

3️⃣
Long Coherence

T₂ ≫ t_gate. Decoherence must be slow vs. gate time. For neutral atoms: T₂ > 1 s is achievable; tweezer photon scattering erodes it.

4️⃣
Universal Gates

High-fidelity single-qubit rotations + at least one entangling two-qubit gate (e.g. Rydberg blockade for neutral atoms).

5️⃣
Qubit Readout

Site-specific measurement in the computational basis with near-unity fidelity, without disturbing neighbours.

Analog vs. Digital Quantum Computing

Two fundamentally different approaches, and why neutral atoms can do both.

🔢 Digital (Gate-Based)

Programs decomposed into discrete unitary gates from a universal gate set, applied to qubits, then measured. Mathematically clean and compatible with quantum error correction.

Cost: Error correction needs 100s–1000s of physical qubits per logical qubit. Every gate must exceed the fault-tolerance threshold (~10⁻³ for surface codes).

Who: IBM, Google, IonQ, Quantinuum, QuEra (Rydberg gates)

🌊 Analog (Simulation / Annealing)

Engineer a Hamiltonian whose ground state encodes the answer, then adiabatically evolve from an easy initial state. Special-purpose, not universally programmable, but far larger system sizes are accessible in the NISQ era.

Examples: D-Wave's 5000+ superconducting flux qubit annealers; neutral-atom arrays studying quantum magnetism on 2D lattices of hundreds of atoms.

Neutral atoms are unusual: the same array can run Rydberg gates (digital) OR evolve a programmable spin Hamiltonian (analog), just by changing the control protocol. Pasqal & QuEra exploit this dual capability as a near-term strategy.

Platform Comparison

Six dimensions, normalised 0–10. No platform wins on all axes — that's the point.

Scores are editorial judgements based on published 2024–26 benchmarks and company roadmaps; roadmap claims are not equivalent to demonstrated hardware. Hover / tap datasets to isolate. peer-reviewed company roadmap rough estimate

Qubit Counts & Best Gate Fidelities (≈ 2025–26)

Atom Computing's 1,180-atom ¹⁷¹Yb array is a loading demonstration; gate operations use a subset. IBM Condor (1,121Q) is retired; IBM's active flagship is Heron r2 (156Q) with superior error rates. D-Wave's 7,000+ flux qubits are quantum annealers — not gate-based and not directly comparable. PsiQuantum bar = projected photon-mode target; no gate QPU exists yet.

Best published 2-qubit gate fidelities. Superconducting: Google Willow / IBM Heron r2 (~99.5%). Trapped ions: Quantinuum H2 99.9% (arXiv:2305.03828). Neutral atoms: Evered et al., Harvard, Nature 2023 (99.5%). Silicon spins: Mills et al. ²⁸Si, Sci. Adv. 2022 (99.8%). Red dashed line = surface code practical threshold ~99.9% (Fowler et al., PRA 2012).

Platform Deep Dives

How each technology works, key metrics, bottlenecks, leading companies, and key papers.

Superconducting Qubits

How it works

Transmon qubits — Josephson junction circuits cooled to ~15 mK in dilution refrigerators. The Josephson junction makes the LC oscillator anharmonic, isolating the |0⟩↔|1⟩ transition from the rest of the energy ladder. Microwave pulses (5–8 GHz) drive single- and two-qubit gates. The cross-resonance (IBM) or parametric (Google) interaction is the workhorse two-qubit mechanism.

Key Metrics (2025–26)

Gate time: 10–200 ns 1Q fidelity: >99.9% 2Q fidelity: ~99.5% (Heron/Willow) T₁: 100–500 µs (best Heron) T₂: 50–300 µs Qubit count: 100–500 gate-quality

Principal Bottleneck

Classical wiring at millikelvin: each qubit requires several coaxial lines; heat load and cable density make scaling beyond a few thousand physical qubits extremely hard. Leading solutions: cryo-CMOS multiplexers (Intel Horse Ridge II) and microwave-to-optical transducers for inter-module coupling. IBM's modular Quantum System Two links multiple Heron chips via classical interconnects.

$$E_J = \frac{(\Phi_0/2\pi)^2}{2L_J} \quad \text{[Josephson energy]}$$ $$\omega_q \approx \frac{\sqrt{8E_J E_C}}{\hbar} \quad E_C = \frac{e^2}{2C_\Sigma}$$ Anharmonicity: $\alpha = \omega_{12} - \omega_{01} \approx -E_C/\hbar \approx -200$ to $-350$ MHz

Companies (2025–26)

IBM Quantum ↗

Heron r2 (156Q, 2024) is IBM's active flagship — Condor (1,121Q) retired. IBM Quantum System Two is a modular architecture linking Heron chips. Roadmap: Flamingo → Kookaburra for higher qubit quality. Cloud via IBM Quantum Platform.

Google Quantum AI ↗

Willow (105Q, Dec 2024): first chip to demonstrate below-threshold error correction — exponential error suppression as code distance increases. Sets the standard for near-term FTQC demonstrations.

Rigetti Computing ↗

Ankaa-3 (84Q, 2024). Publicly traded (RGTI). Cloud hybrid via AWS/Azure. Focus on algorithmic benchmarking over raw qubit count.

IQM Quantum Computers ↗

Finnish startup. IQM Resonance (20Q, Europe-deployed). Co-design QPU architecture. HPC integration focus with CSC Finland and Leibniz Supercomputing Centre.

Alice & Bob ↗

Cat qubits: biased-noise encoding suppresses bit-flips exponentially while accepting higher phase-flip rates, dramatically reducing QEC overhead. Series B raised (2024). Targeting logical qubit demonstration.

D-Wave (QBTS) ↗

Advantage2 (~7,000 flux-qubit annealers, 2024). Not gate-based — solves QUBO/Ising problems via adiabatic evolution. NASDAQ-listed. Commercial traction in logistics and materials optimisation.

Trapped Ion Qubits

How it works

Hyperfine "clock" states of ¹⁷¹Yb⁺ or ⁴⁰Ca⁺ ions confined in radiofrequency Paul traps. Ions laser-cooled to the motional ground state; shared vibrational modes (phonons) of the ion chain serve as a quantum bus. The Mølmer–Sørensen (MS) gate drives entanglement via off-resonant sideband excitation. Clock-state qubits are insensitive to first-order Zeeman shifts — T₂ can exceed minutes. All-to-all connectivity within a trap zone is native.

Key Metrics (2025–26)

T₂: 1 s – tens of minutes 2Q fidelity: 99.9% (Quantinuum) 1Q fidelity: <10⁻⁴ error rate Gate time: 20–500 µs Qubit count: 32–56 (gate-quality) Ops/T₂: >10⁴ All-to-all connectivity within zone

Principal Bottleneck & Scaling Path

Mode crowding: N ions share 3N vibrational modes; gates become slower and noisier as the chain grows. Solution — QCCD: segment the trap into storage, gate, and readout zones; shuttle ions between them. Quantinuum's H2 implements this with all-to-all connectivity on 56 qubits. Next step: photonic interconnects linking separate trap modules into a quantum network — the modular ion-trap path to thousands of logical qubits.

$$\hat{H}_{\rm MS} = \hbar\Omega(\hat{\sigma}^+_a\hat{\sigma}^+_b + \hat{\sigma}^-_a\hat{\sigma}^-_b)\cos(\delta t)$$ $\delta = \omega_{\rm laser} - \omega_{\rm qubit} - \omega_{\rm phonon}$ (blue sideband detuning)
Gate time: $\tau_g \sim \pi/(\eta^2\Omega)$ where $\eta = k x_{\rm zpf}$ is the Lamb–Dicke parameter
Infidelity ∝ $\bar{n}$ (mean phonon number): requires ground-state cooling $\bar{n} < 0.05$

Companies (2025–26)

IonQ ↗

Forte Enterprise (35 #AQ, 2024). NASDAQ-listed (IONQ). Proprietary #AQ (algorithmic qubit) metric benchmarks circuit performance rather than raw qubit count. Partners with AWS, Microsoft Azure, Google Cloud.

Quantinuum ↗

H2 processor (56 trapped ¹⁷¹Yb⁺ ions, 2024): best published 2Q gate fidelity (99.9%) of any commercial QPU. Joint venture: Honeywell + Cambridge Quantum. Major Microsoft strategic investment (2025). H3 roadmap targets 100+ qubits with modular QCCD.

Oxford Ionics ↗

Electronic control (microwave pulses, no lasers) — compatible with CMOS fab processes. Quieter gates, easier integration. £30M Series A (2023). Demonstrated single-qubit fidelity >99.9% in ⁴⁰Ca⁺ with microwave-only drives.

AQT (Alpine Quantum Technologies) ↗

⁴⁰Ca⁺ ion trap systems. European quantum flagship partner. Deployed systems at CERN (HPC integration pilot) and Vienna University of Technology.

eleQtron ↗

German startup. Microwave-driven ²⁵Mg⁺ trapped ions. Room-temperature-compatible control electronics. BMBF-funded.

Neutral Atom Qubits — Optical Tweezers + Rydberg Gates

How it works

Individual neutral atoms (⁸⁷Rb, ¹³³Cs, ¹⁷¹Yb, ⁸⁸Sr) trapped in tightly focused laser beams (optical tweezers; ~1 µm waist). Intensity gradient provides a conservative trapping potential. Hyperfine or nuclear-spin ground states form the qubit. Two-qubit gates use the Rydberg blockade: transient excitation to n ~ 60–100 Rydberg states, where the van der Waals interaction V = C₆/r⁶ (C₆ ∝ n¹¹) prevents simultaneous excitation of two atoms within the blockade radius r_b. Zone-based architectures (Gemini-class) separate storage, entangling, and readout zones — atoms are transported by AOD-controlled tweezers.

Key Metrics (2025–26)

Array size: 50–1,180 atoms Coherence T₂: 1–10 s (hyperfine) 2Q CZ fidelity: 99.5% best (Evered 2023) Gate time: 0.2–5 µs Identical qubits: atoms are perfect copies Reconfigurable 2D/3D geometry in real time Dual-mode: digital gates OR analog Hamiltonian Mid-circuit readout: demonstrated at 48 LQ scale

Engineering Frontiers (2025–26)

① Gate fidelity → 99.9%+: Evered et al. 2023 demonstrated 99.5% native CZ on ⁸⁷Rb. ¹⁷¹Yb and ⁸⁸Sr nuclear-spin qubits (I = ½, 0) offer longer T₂ and clock-transition magic wavelengths. Optimal-control pulse shaping + deep ground-state cooling (⟨n⟩ < 0.05) is the route to fault-tolerance-grade gates.

② Fault-tolerant logical qubits: Bluvstein et al. (Nature 2024) demonstrated 48 error-corrected logical qubits and 228 logical gates — the most complex fault-tolerant operation on any platform to date. [[7,1,3]] and surface-code QEC are both being explored.

③ Scale to 10⁴ atoms: Multiplexed AOD arrays + parallel imaging. QuEra's Gemini-class system targets thousands of atoms. Atom Computing showed 1,180-qubit ¹⁷¹Yb loading (2023); gate operation on large subsets is the next step.

④ Photonic interconnects: Entanglement between atom arrays via photon-mediated Bell pairs is in early research — essential for modular scale-out beyond a single vacuum chamber.

$$V_{\rm dd}(r) = \frac{C_6}{r^6},\quad C_6 \propto n^{11} \quad \text{(van der Waals, }n \gg 1\text{)}$$ $$r_b = \left(\frac{C_6}{\hbar\Omega_{\rm Ryd}}\right)^{1/6} \sim 5\text{–}15\;\mu\text{m for }n = 60\text{–}100$$ $$\varepsilon_{\rm motion} \approx \alpha\langle n\rangle, \quad \langle n\rangle = k_{\rm B}T/(\hbar\omega_{\rm trap})$$ State-of-art: $\langle n\rangle < 0.05$ via resolved-sideband cooling on ¹⁷¹Yb clock transition

Companies (2025–26)

QuEra Computing ↗

Harvard/MIT spin-out (Lukin/Greiner groups). Aquila (256 ⁸⁷Rb atoms) on AWS Braket — largest publicly accessible neutral-atom QPU. Gemini-class next-gen targets thousands of atoms in zone-based architecture. Digital gate mode + analog Hamiltonian simulation in one platform. $230M Series B (2023).

Pasqal ↗

French startup (Browaeys/Lahaye, IOGS). Fresnel QPU (100 ⁸⁷Rb atoms). Hybrid analog-digital. HPC cloud integration (EDF, BASF, Crédit Agricole pilots). Merged with Qu&Co for quantum software stack. European Quantum Flagship partner.

Atom Computing ↗

Berkeley-based. 1,180-qubit ¹⁷¹Yb array loading demonstrated (Oct 2023). ¹⁷¹Yb nuclear spin qubit (I = ½) gives exceptionally long T₂; Sr-88 clock-transition enables magic-wavelength trapping. Next: gate operations at array scale.

Infleqtion ↗

Formerly ColdQuanta. Sqynet cloud QPU (Rb/Cs-based). Quantum networking programme. Acquired SuperTech; contracts with DARPA and US Air Force for quantum sensing + computing. ~$100M raised.

Silicon Spin Qubits

How it works

Electron (or hole) spins confined in Si or Si/SiGe quantum dots defined by electrostatic gates. Spin-up |↑⟩ and spin-down |↓⟩ form the computational basis. Single-qubit gates via electron spin resonance (ESR) driven by microwave pulses. Two-qubit gates: exchange interaction J tuned by gate voltage V_ex — the CPHASE gate follows. In isotopically purified ²⁸Si (0.01% residual ²⁹Si) the dominant nuclear spin bath dephasing is removed, enabling T₂* > 10 ms and T₂ > 1 ms with dynamical decoupling.

Key Metrics (2025–26)

Gate time: 20–100 ns 2Q fidelity: 99.8% (²⁸Si best) T₂ (echo): >1 ms in ²⁸Si Ops/T₂: ~10⁴–10⁵ Qubit count: 6–16 (gate-quality, 2025) Variability: no two dots are identical

The Long Game

Integration density is silicon's superpower: existing 300 mm CMOS fabs could in principle place millions of quantum dots on a single chip — the same infrastructure that makes CPUs. If device-to-device variability can be solved (through automated ML-based tuning or foundry-level uniformity), silicon spins represent the most credible path to millions of physical qubits. Timeline: ~5–10 years from demonstration-scale to hundreds of qubits.

Companies (2025–26)

Intel ↗

Tunnel Falls (12Q, 2023). Built on Intel's 300 mm fab process. Horse Ridge II cryo-CMOS controller. Automotive-grade variability reduction is a core research focus.

Quantum Motion ↗

UK startup. CMOS-compatible Si/SiO₂ platform. Series B raised (2023). Target: foundry-compatible qubit arrays with in-situ tuning. Partnership with GlobalFoundries.

Silicon Quantum Computing ↗

Australian national initiative (UNSW Sydney). Precision atom-by-atom Si:P doping — individual phosphorus donors as qubits. Sub-nm positioning control via STM lithography. 10-qubit demonstrated (2023).

Equal1 ↗

Irish startup. Full-stack QPU on a single CMOS chip — qubits + classical control co-integrated at 4K, eliminating the dilution fridge for the control circuitry.

Photonic Qubits

How it works

Qubits encoded in polarisation, path, or time-bin modes of single photons in integrated silicon photonics waveguides. Decoherence during transmission is negligible — photons are flying qubits. Single-qubit gates: beam splitters (50:50) and electro-optic phase shifters. The fundamental problem: photons don't interact — probabilistic entangling gates (KLM protocol) or matter-mediated nonlinearities are required. Both scale poorly without further architecture innovation.

Fusion-Based QC — PsiQuantum's Bet

Avoid probabilistic two-qubit gates entirely. Pre-generate small entangled resource states (4–6 photons). Fuse them via Bell measurements (fusion gates). Individual fusion failures are heralded and corrected architecturally through the fault-tolerant resource state network. This approach, if single-photon sources and detectors reach ~99%+ efficiency, allows scaling without each gate needing to succeed. PsiQuantum's target: 1M+ photonic modes on GlobalFoundries 300 mm silicon photonics.

Near-Term Value

Quantum key distribution (QKD) networks are commercially deployed today (Toshiba, ID Quantique). Photonic interconnects are essential for linking any qubit module over long distances. Gaussian boson sampling (GBS) on squeezed-light networks offers provable quantum speedup on specific sampling problems.

Companies (2025–26)

PsiQuantum ↗

Largest photonic QC bet: >$700M private + AUD $940M Queensland government + US DARPA/DoE contracts (>$1.5B total). Fusion-based architecture on GlobalFoundries 300 mm silicon photonics. No gate QPU yet — building the fab-compatible supply chain first.

Xanadu ↗

Borealis (216 squeezed modes, 2022): demonstrated beyond-classical speed on GBS. PennyLane: open-source quantum ML framework (60k+ GitHub stars). Amazon Braket partner. Photonic fault-tolerance roadmap via GKP encoding.

QuiX Quantum ↗

Si₃N₄ waveguide processors for boson sampling. 20-mode photonic chip (2021). European Quantum Flagship member. Focus on linear-optical circuits for near-term experiments.

Quandela ↗

InGaAs quantum dot single-photon sources with >99% purity. Muse cloud QPU. Focus on high-efficiency, high-indistinguishability photon generation — the rate-limiting hardware for photonic QC.

Solid-State Defects & Topological Qubits

NV Centres & Diamond Defects

Nitrogen-vacancy (NV) centres — a nitrogen atom adjacent to a lattice vacancy — are electron spin qubits with millisecond-to-hour coherence at room temperature, unique among all qubit platforms. Optically addressable: spin state initialised and read out via spin-selective fluorescence (ZPL at 637 nm). Related platforms: SiV and GeV centres in diamond (Lukin group, Harvard), defects in SiC (neutral divacancy, V₁). Principal near-term roles: quantum repeater network nodes, nanoscale magnetometry and NMR, and as memory qubits in hybrid quantum networks.

Topological Qubits — Majorana Zero Modes

Majorana fermions at the ends of 1D topological superconductors (semiconductor nanowire + s-wave SC in a magnetic field) encode a qubit non-locally — meaning local perturbations cannot flip it. Theoretically this gives hardware-level error protection far below 10⁻⁶, dramatically reducing QEC overhead.

Feb 2025 — Microsoft Majorana 1: Microsoft's Station Q team announced the Majorana 1 chip: an InAs/Al heterostructure device designed to host Majorana zero modes, with a topological gap measured via interferometric parity readout. This is the most concrete hardware demonstration yet, though independent verification of non-Abelian braiding remains an open experimental challenge.

⚠️ Topological qubits: promising physics, unproven computation. The Feb 2025 Majorana 1 announcement demonstrates a topological gap and parity readout; braiding (the operation needed for fault-tolerant gates) has not yet been demonstrated. Peer review and independent replication are ongoing. Treat commercial timelines with skepticism — this remains a physics frontier.

Companies & Groups (2025–26)

Microsoft Azure Quantum ↗

Majorana 1 chip (InAs/Al, Feb 2025). Station Q research lab. Azure Quantum also partners with IonQ, Quantinuum, Rigetti, and Pasqal as a multi-hardware cloud provider. Strategic investment in Quantinuum (2025).

Quantum Brilliance ↗

Room-temperature NV-centre quantum accelerators. Diamond QPU at 20–25°C — no dilution fridge. Partnered with Oak Ridge National Lab (ORNL). Focus: edge-deployable quantum co-processors for HPC nodes.

Q-NEXT / Argonne ↗

US DOE quantum center. NV centres for quantum repeater nodes. Operating a Chicago–Argonne quantum network testbed (first metropolitan QKD links). 2025 goal: entanglement distribution across 50 km.

QuTech (TU Delft) ↗

Pioneered NV-centre quantum networking (Hensen 2015 loophole-free Bell test). Demonstrated entanglement distribution across a three-node network (Oxford–Delft–Eindhoven). Topological Josephson junction research also active.

Road to Fault-Tolerant Quantum Computing

Milestone timeline across the three leading platforms (2019–2032). Solid dot = achieved; open dot = projected.

🛡️ Surface Code — Key Numbers

Encodes 1 logical qubit in ~2d² physical qubits (code distance d). Corrects any error on <⌊d/2⌋ qubits per syndrome round. At physical error rate 10⁻³, distance-21 code (~882 physical/logical qubit) gives logical error rate ~10⁻¹⁰/cycle. Google Willow (Dec 2024) demonstrated exponential error suppression below threshold: adding more qubits reduced the logical error rate. Breaking RSA-2048 needs ~20 million physical qubits — Gidney & Ekerå, Quantum 5, 433 (2021).

🔵 Superconducting
2019 Google Sycamore 53Q quantum supremacy
2023 Google: surface code logical qubit demonstrated
2024 Willow 105Q: below-threshold QEC — a first
2024 IBM shifts to Heron r2; retires Condor; utility-scale focus
2027 ~10 high-quality logical qubits (Google/IBM roadmaps)
2030+ 100–1000 logical qubits; cryo-CMOS interconnects
🟢 Trapped Ions
2021 IonQ 32-qubit system on AWS/Azure
2023 Quantinuum H2: 99.9% 2Q gate fidelity
2024 Quantinuum H2-1: 56Q all-to-all connected
2025 H2 reports record quantum volume 33,554,432
2027 Modular QCCD network; H3 with 100+ qubits
2032 Fault-tolerant logical qubit processor
🟣 Neutral Atoms
2021 QuEra 256-atom analog processor (Ebadi et al.)
2023 99.5% Rydberg CZ gate (Evered et al., Harvard)
2023 1,180-qubit ¹⁷¹Yb loading (Atom Computing)
2024 48 logical qubits + logical circuits (Bluvstein et al.)
2025 6,100 coherent Rb qubits; 96-logical-qubit FT architecture
2026–28 Repeated active QEC cycles and larger logical processors

Open circles = projected. Pulsing dots = in progress (2025–26). Timelines are roadmap estimates, not guarantees.

Platform Snapshot Table (2025–26)

Two-qubit operation values throughout. T₂ = dephasing time; Ops/T₂ = two-qubit gates per coherence window. Highlighted row = neutral atoms (most relevant to AMO physicists).

Platform Active Qubits (2025) 2Q Gate Time 2Q Fidelity (best) T₂ Ops / T₂ Key Bottleneck (2025–26)
🔵 Superconducting 105–156 (gate-quality) 50–200 ns 99.5% (Heron/Willow) 100 µs – 0.5 ms ~10³ Classical wiring at mK; cryo interconnects
🟢 Trapped Ions 32–56 20–500 µs 99.9% (Quantinuum H2) 1 s – minutes >10⁴ Gate speed; modular QCCD scaling
🟣 Neutral Atoms 50–1,180 (loading) 0.2–5 µs 99.5% (Evered 2023) 1–10 s 10³–10⁴ 2Q fidelity → 99.9%; mid-circuit measurement scale
🟡 Silicon Spins 6–16 20–100 ns 99.8% (²⁸Si, Mills 2022) >1 ms (²⁸Si echo) ~10⁴–10⁵ Device variability; automated tuning
🩷 Photonic >216 modes (GBS) Probabilistic 93–97% N/A Limited No deterministic photon–photon gate; source efficiency
🩵 NV / Topological 1–few µs–ms ~99% (NV); undemonstrated (topo) ms – hours (NV) Varies Scale (NV); braiding undemonstrated (topo)

Investment Landscape (2023–2026)

Capital is concentrating around a handful of hardware bets — and governments are piling in.

$1.5B+
PsiQuantum
Private rounds + AUD $940M Queensland government + US DARPA/DoE contracts. Silicon photonic fusion-based QC.
$1B+
Quantinuum
Microsoft strategic investment (2025) — validates trapped-ion quality as the path to early fault tolerance.
$230M
QuEra Computing
Series B (2023). Harvard/MIT spin-out. Largest neutral-atom funding round ever. AWS cloud partner.
$3B+
IBM Quantum (cumulative)
Internal R&D since 2016. Utility-scale QPUs, Qiskit ecosystem, cloud access to 100+ organisations worldwide.
Pattern: Governments (US, UK, EU, Australia, Japan) have collectively committed >$20B to national QC programmes. The bet is: whoever reaches fault tolerance first will have a decisive advantage in cryptography, drug discovery, and materials science.

Neutral Atom Roadmap: Engineering Milestones

Six milestones on the path from today's 96-logical-qubit demonstrations to larger fault-tolerant neutral-atom processors.

✓ 48 → 96 Logical Qubits (2024–2025)

Bluvstein et al. (Nature 2024) demonstrated a 48-logical-qubit reconfigurable Rb processor. The 2025 architecture paper used up to 448 atoms and 96 logical qubits to demonstrate the core elements of universal fault-tolerant neutral-atom computation: zone-based transport, mid-circuit measurement, feed-forward, and logical operations.

Done ✓
① Gate Fidelity → 99.9% Without Post-Selection

Evered et al. reached 99.5% native CZ in Rb. Muniz et al. reported 99.72(3)% for ¹⁷¹Yb with post-selection and 99.40(3)% without. The next milestone is reproducible ≥99.9% two-qubit fidelity without post-selection, at scale and with mid-circuit operations enabled.

Active 2025
② Scale to 10,000+ Atoms

Multiplexed AOD/SLM arrays + parallel imaging. Atom Computing demonstrated 1,180 loaded ¹⁷¹Yb atoms in 2023; Manetsch et al. reported 6,100 highly coherent ⁸⁷Rb atoms across 12,000 tweezer sites in 2025. The remaining challenge is not loading alone, but maintaining gate fidelity, coherence, transport, and low-loss readout across the full array.

Near-term
③ Full Mid-Circuit Measurement + Feed-Forward at Scale

Demonstrated at the 48-logical-qubit level (Bluvstein 2024). Extending to larger arrays with fast FPGA-based real-time decision making is required for active QEC. The bottleneck is fluorescence imaging speed vs. coherence time — cameras and optics must read out ancilla qubits without disturbing data qubits.

Near-term
④ Repeated Active QEC Below Threshold

Neutral atoms now have logical processors and a universal fault-tolerant architecture demonstration. The next harder milestone is repeated active syndrome extraction where increasing code distance reliably lowers logical error over many cycles, while preserving atom number and avoiding measurement cross-talk.

2027–28
⑤ Photonic Interconnects — Modular Scale-Out

A single vacuum chamber cannot hold millions of atoms. Modular architecture links separate tweezer modules via photon-mediated Bell pairs (atom → photon → fibre → atom entanglement). NV-centre quantum repeater technology provides a blueprint. Relevant for neutral-atom quantum networks and distributed quantum computing.

Research

AMO Physics → Quantum Computing Industry

How your experimental skills translate — and what to add to be competitive in 2025–26 hiring.

Your AMO Skills → QC Industry Value

Optical tweezers Neutral atom QC hardware (QuEra, Pasqal, Atom Computing)
Rydberg atoms / blockade Rydberg gate development; two-qubit gate optimisation
Laser locking (PDH, SAS) Qubit drive laser stability; photon-source engineering
Ultra-high vacuum systems All hardware platforms; QPU system engineering
Fluorescence imaging / SNR Qubit readout fidelity; mid-circuit measurement
MOT / evaporative cooling State preparation, cooling to ground state (⟨n⟩ < 0.05)
RF / microwave spectroscopy Hyperfine qubit drives; superconducting gate pulses
FPGA / Python data analysis Quantum control software; real-time QEC feed-forward
Precision measurement / statistics Gate benchmarking (RB, XEB, process tomography)

Skills to Add for Industry

Quantum Software Frameworks
Qiskit (IBM) Bloqade (QuEra, Julia) PennyLane (Xanadu) Cirq (Google)
Theory to Know
Surface codes & QEC Randomised benchmarking VQE / QAOA basics Quantum error models
Job Roles for AMO PhDs
Quantum Hardware Scientist Systems Engineer Quantum Control Engineer Applications Physicist Error Correction Researcher

Where AMO PhDs Are Most Competitive — by Platform

QuEra / Pasqal / Infleqtion / Atom Computing — Direct fit. Your tweezer + Rydberg + laser experience is the core of their hardware stack. These are the most natural landing spots for an AMO postdoc transitioning to industry.
Quantinuum / IonQ / Oxford Ionics — Strong overlap: laser cooling, vacuum, spectroscopy, microwave control. Trapped-ion teams heavily recruit from AMO backgrounds.
IBM Research / Google Quantum AI — Research scientist roles for AMO physicists with strong QEC or quantum optics background. Typically require demonstrated software fluency (Qiskit/Cirq) + familiarity with superconducting physics.
National Labs (JILA, NIST, Sandia, Argonne, Oak Ridge) — AMO + QC intersection is well-funded. NIST Ion Storage Group, Argonne Q-NEXT, Sandia Quantum Network all hire AMO PhDs directly. Strong bridge if you want to stay research-adjacent.
PsiQuantum / Xanadu / Quandela — Photonics-adjacent. Relevant if you have quantum optics, cavity QED, or single-photon source experience. Harder cold fit from pure tweezer background unless you build quantum optics skills.
Quantum Software / Cloud (1QBit, Algorithmiq, QC Ware) — Open to strong physicists who learn Python, QEC, and algorithm basics. Less hardware, more classical-quantum interface. Good pivot if you want industry pace over lab life.

References & Further Reading