Quantum Computing in 2026 — IBM Heron R2, Google Willow, Quantinuum, IonQ, Pasqal, PsiQuantum Deep Dive

Prologue — NISQ is Over, What Comes Next?

The day Google announced quantum supremacy with Sycamore in 2019, we could not answer the question "what is it good for?" Fifty-three qubits, half-percent gate error, coherence in microseconds. The experiment was elegant; its practical value was zero. We then spent five years in the fog called NISQ — Noisy Intermediate-Scale Quantum.

In December 2024 Google published Willow in Nature, and the fog began to clear. It was the first experiment to drop below the surface-code threshold, meaning errors fell exponentially as code distance grew. In November of the same year IBM scaled Heron R2 to 156 qubits and pushed two-qubit error rates down to about 5e-4. In February 2025 Microsoft unveiled an 8-qubit Majorana 1 chip and AWS unveiled a 9-qubit Ocelot chip within the same week. Both companies were selling not "qubit count" but "error protection baked into the architecture."

This essay surveys the full quantum computing stack as of May 2026 in one place — modality comparison, company-by-company roadmaps, programming languages, error-correcting codes, and Korea/Japan national strategy. Long but quick to read.

1. The Big Picture — End of the NISQ Era

NISQ in one sentence was "trying to do something useful with 50 to 1000 qubits without error correction." Variational algorithms like VQE, QAOA, and quantum machine learning were the workhorses, and the results were, honestly, lukewarm. Classical algorithms caught up quickly, and noise kept circuit depth under 30 gates.

The 2026 landscape is different. Three pivots:

1. Crossing the error-correction threshold. Willow showed that for distance-3, 5, and 7 surface-code patches the logical error rate decreased monotonically. Each increment in code distance roughly halved logical error — the first experimental confirmation of Shor's 1996 fault-tolerance theorem.
2. Utility-scale quantum. IBM coined "Quantum Utility" for computations on 100-plus qubit circuits where classical simulation is effectively impossible. Heron R2 combined with error mitigation is starting to produce meaningful results in chemistry and materials science.
3. Modality diversification. The era when superconducting was the only game in town is over. Neutral atoms (Pasqal, Atom Computing, QuEra) have crossed 1000 qubits, ion traps (Quantinuum, IonQ) occupy the 99.99-percent gate fidelity regime, and photonics (PsiQuantum) is pitching room-temperature operation and fiber-network compatibility as data-center-ready.

NISQ is becoming "utility scale." Fault-tolerant comes next, with the consensus landing between 2029 at the earliest and 2032 at the latest.

2. Six Qubit Modalities — Superconducting, Ion Trap, Neutral Atom, Photonic, Topological, Cat

A qubit is the vessel for quantum information. The physical method used to make the vessel determines the modality. The six leading approaches in 2026:

Superconducting

An LC circuit of aluminum and niobium cooled to millikelvin temperatures becomes a quantum object. The transmon is the standard design. Gate times of 10 to 100 nanoseconds, fastest of all modalities, and friendly to standard semiconductor fabrication. Downsides: qubit connectivity is restricted to a nearest-neighbor lattice, and coherence sits around 100 microseconds. IBM, Google, Rigetti, and IQM are in this camp.

Trapped Ion

Ytterbium or barium ions are held in a vacuum chamber by electric fields and manipulated with lasers. Every qubit can entangle directly with every other (all-to-all connectivity), which cuts circuit depth by more than half. Gate fidelity reaches 99.99 percent, the highest, but gate times are microseconds to milliseconds, slower than superconducting. Quantinuum and IonQ lead this camp.

Neutral Atom

Rubidium or cesium atoms are caught in optical tweezers and excited to Rydberg states to entangle. Arrays of 1000-plus qubits can be reconfigured into arbitrary grids, making scaling the easiest. Gate fidelity around 99.5 percent. Pasqal, Atom Computing, and QuEra are the leaders.

Photonic

Photons themselves are the qubit. Operates at room temperature and qubits can travel along fiber, ideal for networking. Downside: photon generation is probabilistic, making circuit depth hard. PsiQuantum routes around this with fusion-based quantum computing (FBQC).

Topological

A hypothetical particle called the Majorana fermion holds qubit state topologically. Intrinsically resilient to noise, promising error-correction overhead 100 to 1000 times smaller. Microsoft's Majorana 1 in February 2025 was the first experimental demo.

Cat Qubit

A microwave cavity holds two coherent states of opposite phase (a Schrodinger cat state) as the qubit. Bit-flip errors are exponentially suppressed, leaving only phase-flip errors to correct. AWS announced Ocelot in February 2025.

3. IBM Quantum Heron R2 — 156 Qubits and System Two

IBM announced Heron R2 at 156 qubits in November 2025. R2's core message is "more qubits without losing two-qubit fidelity" — median two-qubit error around 5e-4, and circuits of depth 5000 gates start producing meaningful results.

The headline word in IBM's roadmap is Quantum System Two. Not a single chip but multiple Heron modules linked by m-couplers and l-couplers, scaling to 200-qubit and 1000-qubit classes. As of 2026 the first System Two is running at IBM Yorktown, and a second has been installed at Cleveland Clinic.

IBM's next chip Kookaburra targets 1386 qubits by late 2026, succeeding the Condor (1121 qubits) generation. But IBM is betting more on "error mitigation plus partial error correction" at utility scale than on raw qubit count. Qiskit Runtime's PEC (probabilistic error cancellation) and ZNE (zero-noise extrapolation) are the workhorses.

Sending a circuit to Heron R2 with Qiskit:

from qiskit import QuantumCircuit
from qiskit_ibm_runtime import QiskitRuntimeService, SamplerV2

service = QiskitRuntimeService(channel="ibm_quantum")
backend = service.backend("ibm_heron_r2")

qc = QuantumCircuit(3)
qc.h(0)
qc.cx(0, 1)
qc.cx(1, 2)
qc.measure_all()

sampler = SamplerV2(backend)
job = sampler.run([qc], shots=1024)
print(job.result()[0].data.meas.get_counts())

4. Google Willow (Dec 2024) — What Exponential Error Reduction Means

Willow is a 105-qubit superconducting chip. By qubit count alone it is smaller than Heron R2, but the message it carries is on a different axis.

Two key results:

Below threshold. Willow demonstrated that on distance-3, 5, and 7 surface-code patches the logical error rate roughly halved with each increment of code distance. This is the first experimental crossing into the regime where surface code actually works. Shor's 1996 paper-proof was finally realized in a lab 28 years later.

RCS benchmark. Random Circuit Sampling on 105 qubits at depth 40 finished in 5 minutes, a task that would take the Frontier supercomputer 10 to the 25 years. The 53-qubit Sycamore experiment from 2019 was beaten by two orders of magnitude in six years.

Willow's qubit coherence time is T1 around 100 us, five times longer than Sycamore. Gate fidelity is 99.95 percent for single-qubit and 99.7 percent for two-qubit. Google's next goal is a single logical qubit — bundle one distance-7 surface-code patch into one error-corrected qubit. From there, scaling to 100 and 1000 logical qubits is the real road to fault tolerance.

Sending a circuit to Willow with Cirq:

import cirq
import cirq_google

processor_id = "willow"
sampler = cirq_google.get_engine().get_processor(processor_id).get_sampler()

q0, q1 = cirq.GridQubit(4, 1), cirq.GridQubit(4, 2)
circuit = cirq.Circuit(
    cirq.H(q0),
    cirq.CNOT(q0, q1),
    cirq.measure(q0, q1, key="m"),
)

result = sampler.run(circuit, repetitions=1000)
print(result.histogram(key="m"))

5. Quantinuum H2/H3 — The Trapped-Ion Champions

Quantinuum is the merger of Honeywell's quantum division and Cambridge Quantum. It runs the H2 system at 56 qubits as of 2024 and brought H3 into beta in 2025. Largest ion-trap system in the field.

H2's differentiator is the QCCD (Quantum Charge-Coupled Device) architecture. Ions shuttle around the vacuum trap so that any two qubits can be brought into the same gate zone for entanglement. All-to-all connectivity falls out naturally. Two-qubit gate fidelity is 99.87 percent, industry leading.

On H2 in late 2024, Quantinuum encoded one fault-tolerant logical qubit using a carbon-12 color code at distance 3, achieving a logical error rate below the raw physical rate — a first for the field. In 2025 they extended to distance 5.

H3 starts at fewer qubits but trims gate time and switches to 2D traps. Quantinuum's roadmap continues with Sol (codename) in 2027 and Apollo in 2029.

Submitting an OpenQASM 3 circuit to a Quantinuum backend:

from qiskit_quantinuum import QuantinuumBackend

backend = QuantinuumBackend(name="H2-1")

qasm = """
OPENQASM 3.0;
qubit[3] q;
bit[3] c;
h q[0];
cx q[0], q[1];
cx q[1], q[2];
c = measure q;
"""

job = backend.run(qasm, shots=1024)
print(job.result().get_counts())

6. IonQ Tempo — Forte's Successor at 64 AQ

IonQ released Forte Enterprise (36 algorithmic qubits) in 2024 and Tempo (64 algorithmic qubits) in late 2025. IonQ uses its own metric "algorithmic qubits (AQ)" — the effective qubit count usable to run meaningful algorithms in the presence of noise.

IonQ's ion is barium (Ba+), differing from Quantinuum's ytterbium (Yb). Barium can be manipulated with visible-light lasers, allowing cheap light sources like diode lasers, and is friendly to fiber optics. The trade-off is the need for isotope separation and more complex trap design.

Tempo is available on AWS Braket, Microsoft Azure Quantum, and Google Cloud — IonQ is the most open in terms of cloud access.

Submitting a circuit to IonQ Tempo via Braket:

from braket.aws import AwsDevice
from braket.circuits import Circuit

device = AwsDevice("arn:aws:braket:::device/qpu/ionq/Tempo")

circ = Circuit().h(0).cnot(0, 1).cnot(1, 2)
task = device.run(circ, shots=1000)
print(task.result().measurement_counts)

7. Pasqal Orion / Atom Computing — Neutral-Atom Surge

The neutral-atom camp has grown fastest between 2024 and 2026.

Pasqal Orion (France)

After 100- and 256-qubit systems, Pasqal's Orion series targets 1000 qubits. Optical tweezers arrange rubidium atoms into arbitrary lattices, so lattice, triangular, or arbitrary-graph topologies are configurable on the fly. Both quantum annealing and digital-circuit modes are supported.

Pasqal's SDK is Pulser, a pulse-level library. You design laser pulse shapes themselves rather than gates.

from pulser import Pulse, Sequence, Register
from pulser.devices import AnalogDevice
from pulser.waveforms import BlackmanWaveform

reg = Register.rectangle(2, 2, spacing=5)
seq = Sequence(reg, AnalogDevice)
seq.declare_channel("ising", "rydberg_global")

pulse = Pulse.ConstantDetuning(
    BlackmanWaveform(1000, 8.0), 0.0, 0.0
)
seq.add(pulse, "ising")

result = seq.draw()

Atom Computing (USA)

Atom Computing demonstrated a 1180-qubit array in 2023. They use cesium atoms and store the qubit in nuclear spin while using optical spin for gates. This separation gives them coherence times of 40 seconds, the longest of any modality.

Atom Computing is working with NIST on standards and integrating with Microsoft Azure Quantum.

QuEra (USA)

QuEra runs the 256-qubit Aquila system on AWS Braket. Strong on analog-digital hybrid quantum simulation.

8. PsiQuantum — Photonics, Billion-Qubit Plan

PsiQuantum builds photonic qubits in silicon photonics fabs. The core idea is to print quantum circuits on GlobalFoundries' 300-mm wafer lines. Rather than a single chip, the plan is a modular photonic system reaching one million to one billion qubits.

Photonic qubits run at room temperature and qubits can travel along fiber as-is, which is data-center friendly. The downside is that photon generation is probabilistic. PsiQuantum routes around this with fusion-based quantum computing (FBQC). Small entangled resource states are produced in advance, and "fusion measurements" stitch them into large entangled states.

In 2024 PsiQuantum announced its first fault-tolerant facility in Brisbane, Australia, followed in 2025 by a second site in Chicago, Illinois. Both target 2027 to 2029 for operation.

Whether PsiQuantum's promise is real remains an open question. While other companies show actual measurements on 100 to 1000 qubits, PsiQuantum draws the big picture of "millions of qubits in one shot." All-or-nothing.

9. Rigetti / IQM — The Superconducting Challengers

Beyond IBM and Google, two more companies hold the superconducting line.

Rigetti Computing (USA)

Rigetti released Ankaa-3 at 84 qubits in 2024 and demonstrated a 168-qubit multi-chip system by linking two 84-qubit modules in 2025. Available on both AWS Braket and Rigetti's own Quantum Cloud Services (QCS). Rigetti's strength is fast gate times — 70 ns for a two-qubit gate.

Rigetti's SDK is Forest, and you write circuits in pyQuil:

from pyquil import Program, get_qc
from pyquil.gates import H, CNOT, MEASURE

p = Program()
ro = p.declare("ro", "BIT", 2)
p += H(0)
p += CNOT(0, 1)
p += MEASURE(0, ro[0])
p += MEASURE(1, ro[1])

qc = get_qc("Ankaa-3")
result = qc.run(p.wrap_in_numshots_loop(1000))
print(result.readout_data.get("ro"))

IQM (Finland)

IQM is the standard-bearer of European quantum. Headquartered in Espoo, Finland. They released the 54-qubit IQM Star in 2024, installed at the Jungfrau supercomputer center in Germany. In 2025 they announced the 150-qubit IQM Radiance.

Through EuroHPC funding, IQM has won successive contracts to supply quantum systems to supercomputer centers in Poland, Czechia, and Spain. Their positioning as "HPC-accompanying quantum computer" is clear.

IQM's OpenQASM 3 support is a strength. Compatible with both Qiskit and Cirq.

10. Microsoft Majorana 1 (Feb 2025) — Topological Qubit's First Results

In February 2025 Microsoft published Majorana 1 in Nature. An 8-qubit chip measuring Majorana zero modes on a new material called the "topoconductor" — a combination of indium arsenide (InAs) nanowires and aluminum (Al) superconductor.

The core claim: two Majorana fermions form a pair and together hold one qubit, with the two fermions spatially separated so that local noise cannot affect the qubit state. In theory, error-correction overhead is 100 to 1000 times smaller than surface code.

Skeptics abound. In 2018 Microsoft announced Majorana observations and later retracted the paper due to data-analysis errors. Whether Majorana 1 really demonstrates a topologically protected qubit or merely Andreev bound states is contested in the literature.

Microsoft's promise is "one million topological qubits in five years." If true, a game changer; if not, another large setback. The next data in 2026 to 2027 will decide.

Writing a topological-qubit algorithm in Q#:

namespace Sample {
    open Microsoft.Quantum.Canon;
    open Microsoft.Quantum.Intrinsic;
    open Microsoft.Quantum.Measurement;

    operation BellPair() : (Result, Result) {
        use q = Qubit[2];
        H(q[0]);
        CNOT(q[0], q[1]);
        let r0 = M(q[0]);
        let r1 = M(q[1]);
        ResetAll(q);
        return (r0, r1);
    }
}

11. AWS Ocelot (Feb 2025) — Cat Qubits Arrive

In the same February 2025, AWS announced Ocelot. A 9-qubit system, the result of collaboration with Caltech's quantum lab.

A cat qubit uses two coherent states 180 degrees out of phase — a Schrodinger's cat state — inside a microwave cavity. The two states are intrinsically far apart, so "bit-flip" errors (flips between 0 and 1) are exponentially suppressed. Only phase-flip errors remain, and because only one error type needs correction, error-correction overhead drops dramatically.

Ocelot is the first system to integrate 5 data qubits, 4 measurement buffers, and phase-flip correction into a single chip. AWS reported a 100-fold suppression of bit-flip error.

AWS is opening Ocelot access on Braket in phases. Microsoft, Google, AWS, and IBM are each walking different paths to fault tolerance — topological (MS), surface code (Google), Heron plus error mitigation (IBM), and cat qubits (AWS).

12. Qiskit / Cirq / Q# / OpenQASM 3 — The Programming Standards

Four quantum programming languages compared.

Qiskit (IBM)

Largest ecosystem. Python based, with backends for IBM Quantum plus IonQ, Quantinuum, IQM, and more. Qiskit Runtime is the cloud execution environment. Qiskit 1.0 stabilized in 2024.

from qiskit import QuantumCircuit, transpile
from qiskit_aer import AerSimulator

qc = QuantumCircuit(2, 2)
qc.h(0)
qc.cx(0, 1)
qc.measure([0, 1], [0, 1])

sim = AerSimulator()
tqc = transpile(qc, sim)
result = sim.run(tqc, shots=1024).result()
print(result.get_counts())

Cirq (Google)

Started as a Google-hardware-only library and gradually added other backends. NumPy friendly, designed for AI researchers. cirq-google, cirq-ionq, cirq-pasqal are the adapters.

Q# (Microsoft)

A domain-specific language. Called from Python or C# as an external language. Integrated with Azure Quantum backends. Functional style, strongly typed, and the quantum resource estimation toolchain is a strength.

OpenQASM 3

A vendor-neutral quantum assembly language. Successor to OpenQASM 2, standardized in 2021. As of 2026, IBM, Quantinuum, IonQ, Pasqal, and others accept OpenQASM 3 as input. Qiskit can emit OpenQASM 3, and importing from OpenQASM 3 into Qiskit works as well.

OPENQASM 3.0;
include "stdgates.inc";

qubit[3] q;
bit[3] c;

h q[0];
cx q[0], q[1];
cx q[1], q[2];

c = measure q;

PennyLane (Xanadu)

Specializes in quantum machine learning. Integrates with PyTorch, TensorFlow, and JAX, treating quantum circuits as differentiable neural-network layers.

import pennylane as qml
import numpy as np

dev = qml.device("default.qubit", wires=2)

@qml.qnode(dev)
def circuit(params):
    qml.RX(params[0], wires=0)
    qml.RY(params[1], wires=1)
    qml.CNOT(wires=[0, 1])
    return qml.expval(qml.PauliZ(0))

print(circuit([0.5, 0.3]))

13. Quantum Error Correction — Surface, Color, Floquet

The bridge from NISQ to fault tolerance is quantum error correction (QEC).

Surface Code

Place qubits on a planar lattice and measure only neighboring qubits. Threshold error rate is about 1 percent, the most generous — once two-qubit physical error is under 1 percent, increasing code distance shrinks logical error. A distance-d patch packages d-squared qubits into one logical qubit. Distance-5 uses 25 qubits, distance-7 uses 49. Google's Willow demonstrated below-threshold at distance 7.

Color Code

A code on a triangular lattice. More stabilizers to measure than surface code, but transversal gates like CNOT come for free. Quantinuum's carbon-12 color-code demo is the representative result. Good fit for ion-trap all-to-all connectivity.

Floquet Code

A dynamic code that varies the measurement pattern in time. Proposed by Microsoft researchers in 2022. More qubit-efficient than surface code, and measurement overhead can be spread over time, which is hardware friendly. Some experimental implementations appeared in 2025.

Concatenated / qLDPC

Beyond classical concatenated codes (Steane, Bacon-Shor), quantum LDPC (low-density parity-check) codes have drawn attention since 2024. IBM's bivariate bicycle code is representative — a distance-12 code in 144 physical qubits, 5 to 10 times more qubit-efficient than surface code.

14. Korea (KISTI, ETRI) and Japan (RIKEN, NEC, Fujitsu, NTT)

Asia's quantum strategy differs in flavor from the US and Europe. State-led with industry-academia-government consortia.

Korea

KISTI (Korea Institute of Science and Technology Information) announced installation of an IBM Quantum System Two at Yonsei University Songdo Campus in November 2024 and brought it online in 2025. The first IBM System Two in Asia. Part of the Korean government's Quantum Computing, Communication, and Sensing Master Plan 2030.

ETRI (Electronics and Telecommunications Research Institute) is developing homegrown superconducting qubits. A 20-qubit prototype in 2024, with a 50- to 100-qubit demonstration targeted for 2026. Gate fidelity around 99.5 percent — not at IBM/Google levels but significant as domestic capability.

Q-Center is the KAIST-SNU-POSTECH consortium. Strong in quantum algorithm and software research.

Samsung, SKT, and LG focus more on quantum communication (QKD). SK Telecom holds a stake in ID Quantique.

Japan

RIKEN, jointly with Fujitsu, unveiled a 64-qubit superconducting quantum computer in March 2023 and brought a 256-qubit system online in 2025. Japan's quantum flagship.

NEC develops quantum annealing systems. Both D-Wave-like annealers and superconducting gate-based machines.

Fujitsu, beyond RIKEN, also runs Quantum-Inspired Computing — classical hardware that mimics quantum algorithms, branded as Digital Annealer.

NTT focuses on photonic quantum computing. Leveraging its strength in handling fiber optics and quantum information together, it takes the photonic-quantum route.

Toshiba is a global leader in commercial quantum key distribution (QKD). Their business is more quantum security than quantum computing itself.

The Japanese government allocates tens of billions of yen to Moonshot Program Goal 8 — fault-tolerant quantum computer by 2050.

15. Quantum Applications — Chemistry, Materials, Cryptography, Finance, ML

The five nearest-term application areas as of 2026.

Quantum Chemistry

Molecular-energy calculation. VQE (Variational Quantum Eigensolver) and quantum phase estimation compute ground-state energies of small molecules (H2O, LiH, BeH2). In 2024 IBM published a 78-qubit simulation of a perovskite crystal surface.

Materials Science

Simulation of magnetic, superconducting, and topological materials. Hubbard and Ising models run directly on quantum simulators.

Post-Quantum Cryptography (PQC)

Shor's algorithm, which breaks RSA and ECC, is estimated to require about 2000 to 4000 logical qubits and 10 hours to break RSA-2048. No system in 2026 can yet do this, but NIST finalized PQC standards in August 2024 (ML-KEM, ML-DSA, SLH-DSA). Governments and financial institutions are already migrating to PQC.

Quantum Machine Learning (QML)

In theory quantum computers offer exponential speedup for some ML problems (the HHL algorithm). In practice, data input-output bottlenecks limit the speedup. PennyLane and TensorFlow Quantum are the tools.

Financial Modeling

Option pricing and portfolio optimization. Goldman Sachs and JP Morgan are running pilots with IBM and Quantinuum, using QAOA.

16. Roadmap Beyond 2026 — When Does Useful Quantum Advantage Arrive?

Company-by-company promises:

"Useful quantum advantage" — the moment quantum solves a real problem classical cannot — lands at 2027 to 2029 in the consensus view. Chemistry and materials simulation come first, then optimization and ML, with cryptanalysis last.

17. References

• Google Willow announcement (Nature, Dec 2024) — https://blog.google/technology/research/google-willow-quantum-chip/
• Google Quantum AI page — https://quantumai.google/
• IBM Quantum Heron R2 — https://www.ibm.com/quantum/blog/quantum-roadmap-2025
• IBM Qiskit — https://qiskit.org/
• IBM Quantum Platform — https://quantum.ibm.com/
• Quantinuum H2 — https://www.quantinuum.com/products/h2
• Quantinuum Color Code (Nature, 2024) — https://www.nature.com/articles/s41586-024-08316-w
• IonQ Tempo announcement — https://ionq.com/news
• Pasqal Orion — https://www.pasqal.com/
• Pasqal Pulser SDK — https://pulser.readthedocs.io/
• Atom Computing 1180q (2023) — https://atom-computing.com/
• QuEra Aquila — https://www.quera.com/
• PsiQuantum Brisbane — https://psiquantum.com/
• Microsoft Majorana 1 (Nature, Feb 2025) — https://www.nature.com/articles/s41586-025-08703-x
• Microsoft Q# / Azure Quantum — https://azure.microsoft.com/en-us/products/quantum
• AWS Ocelot announcement (Feb 2025) — https://aws.amazon.com/blogs/quantum-computing/
• AWS Braket — https://aws.amazon.com/braket/
• Rigetti Ankaa-3 — https://www.rigetti.com/
• IQM Star / Radiance — https://www.meetiqm.com/
• OpenQASM 3.0 spec — https://openqasm.com/
• Cirq (Google) — https://quantumai.google/cirq
• PennyLane (Xanadu) — https://pennylane.ai/
• arXiv quant-ph — https://arxiv.org/list/quant-ph/recent
• Surface code review (Fowler et al, 2012) — https://arxiv.org/abs/1208.0928
• IBM qLDPC bivariate bicycle code — https://arxiv.org/abs/2308.07915
• Floquet code (Hastings, Haah, 2022) — https://arxiv.org/abs/2107.02194
• NIST PQC standards — https://csrc.nist.gov/projects/post-quantum-cryptography
• KISTI Quantum Computing Center — https://www.kisti.re.kr/
• ETRI Quantum Computing — https://www.etri.re.kr/
• RIKEN RQC 64-qubit system — https://www.riken.jp/en/research/labs/rqc/
• Japan Moonshot Goal 8 — https://www8.cao.go.jp/cstp/moonshot/

Quantum computing is no longer "ten years away." In May 2026 we stand somewhere between the end of NISQ and the beginning of fault tolerance. The next three years will decide.