Q Bits, 4 Bits, 6 Bits, a Dollar

[EndtheMadnessNow]

Quantum computing is one of those technologies that makes sensible people sound deranged. It promises machines that can explore huge mathematical spaces, simulate molecules atom by atom, optimise impossible systems, and someday casually mug parts of modern cryptography in a dark alley. It is also, at present, a field where “breakthrough” often means “we made the thing fail slightly less catastrophically than last Tuesday”. Progress is real. The hype is also real. Both can be true; science is untidy that way.

The state of the art in 2026 is this: quantum computers exist, they work, they are improving fast, but they are not yet broadly useful general-purpose machines. The field has moved beyond laboratory curiosities and into cloud-accessible processors, corporate roadmaps, government funding, and early fault-tolerance demonstrations. IBM, Google, Quantinuum, IonQ, QuEra, Microsoft, Rigetti, D-Wave and others are all pursuing different hardware strategies. The central prize is not merely more qubits, but reliable logical qubits: error-corrected quantum units that can survive long enough to do useful work.

That distinction matters. A physical qubit is the raw hardware element: a superconducting circuit, trapped ion, neutral atom, photon, spin defect, or some other delicate quantum snowflake. A logical qubit is built from many physical qubits arranged so that errors can be detected and corrected. Classical computers can copy bits and check them easily. Quantum computers cannot simply copy unknown quantum states because physics, in its usual charming manner, says no. So quantum error correction has to infer errors indirectly without destroying the information. This is why quantum engineers age in dog years.

Google’s Willow chip marked one of the clearest recent milestones. In work published in Nature, Google demonstrated below-threshold surface-code quantum error correction, meaning that increasing the code size reduced the logical error rate rather than making the whole contraption worse. Its larger distance-7 code used 101 physical qubits and achieved a logical error rate of roughly 0.143% per error-correction cycle. That is not “Star Trek computer, make me a vaccine”, but it is a serious step towards scalable fault tolerance.

IBM is taking the industrial-roadmap approach: lay out targets, build systems, chase engineering scale, and try not to trip over the superconducting plumbing. IBM says its Starling system, planned for 2029, is intended to deliver 200 logical qubits and run circuits with 100 million quantum gates; its later Blue Jay target aims at 2,000 logical qubits and one billion gates beyond 2033. Reuters also reported in May 2026 that IBM plans to invest more than $10 billion over five years towards large-scale quantum computing. That is not pocket change; that is “we have convinced adults in suits” money.

Trapped-ion systems, meanwhile, remain attractive because their qubits tend to be highly accurate and well connected. Quantinuum’s Helios system, announced in November 2025, claims 98 fully connected physical qubits, 99.9975% single-qubit gate fidelity and 99.921% two-qubit gate fidelity across all qubit pairs. That level of precision matters because error correction becomes less punishing when the underlying qubits are less drunk. IonQ is also pushing an aggressive trapped-ion roadmap, claiming a path to 2 million physical qubits and 80,000 logical qubits by 2030, though such roadmaps should be read with the same caution one applies to film producers promising “locked financing”.

Neutral atoms are another serious contender. They offer arrays of identical atoms, flexible geometry, parallel operations and relatively long coherence times. QuEra and academic collaborators have demonstrated logical-qubit processors using reconfigurable atom arrays, and a 2025 Nature paper described a fault-tolerant neutral-atom architecture for universal quantum computation. Neutral atoms may scale beautifully, but they still face the brutal question every architecture faces: can they run long, useful, error-corrected algorithms at industrial reliability?

Microsoft’s topological-qubit strategy is the glamorous long shot. The pitch is seductive: use Majorana-based qubits that are intrinsically more protected against noise, reducing the enormous overhead of error correction. In February 2025 Microsoft announced its Majorana 1 chip and argued that useful quantum computers could be “years, not decades” away. The problem is that the underlying evidence has faced serious scrutiny, including criticism over earlier Majorana-related work and questions in the physics community about whether Microsoft has really shown what it claims. If Microsoft is right, it has a shortcut. If not, it has a very expensive science-fiction prop with peer-reviewed footnotes.

So what can quantum computers actually do now? They can demonstrate quantum advantage on carefully chosen benchmark problems. They can run small chemistry, materials, optimisation and simulation experiments. They can serve as test beds for algorithms, compilers, error-correction codes and hybrid quantum-classical workflows. They are useful for research. They are not yet routinely outperforming classical supercomputers on commercially important problems. The grown-up phrase here is NISQ — noisy intermediate-scale quantum. Translation: “interesting, expensive and not yet house-trained”.

The biggest barrier remains error. Qubits are absurdly sensitive to noise: heat, vibration, stray electromagnetic fields, cosmic rays, imperfect gates, measurement errors and whatever fresh indignity the universe invents before lunch. Superconducting qubits require extreme cryogenic systems. Trapped ions need lasers, vacuum systems and precise control. Neutral atoms need optical tweezers and exquisite calibration. Photonic systems need efficient sources, detectors and loss management. Every platform has a demon; they merely wear different hats.

The second barrier is scale. A useful fault-tolerant quantum computer will require not just hundreds or thousands of physical qubits, but likely many thousands to millions depending on architecture, gate fidelity and error-correction code. Logical qubits are expensive. Magic-state distillation — needed for many universal fault-tolerant operations — adds further overhead. The machine does not merely need qubits; it needs control electronics, cryogenics or optical systems, calibration, compilers, decoders, interconnects and software stacks that do not collapse like a community theatre set in a thunderstorm.

The third barrier is finding genuinely valuable applications. Quantum computing is expected to shine in quantum chemistry, materials science, some optimisation problems, cryptanalysis, and certain linear algebra and simulation tasks. But “expected” is doing a lot of unpaid labour there. Classical computing is not standing still. GPUs, specialised accelerators, AI-assisted simulation, improved algorithms and high-performance computing keep moving the goalposts. Quantum will need to beat not a frozen 1998 desktop, but the full fury of modern classical engineering.

Cryptography is the exception where the threat is clear even before the machines arrive. A sufficiently large, fault-tolerant quantum computer running Shor’s algorithm would break widely used public-key systems such as RSA and elliptic-curve cryptography. Sensibly, governments and industry are already moving. In August 2024, NIST released its first three finalised post-quantum cryptography standards: ML-KEM for key encapsulation, ML-DSA for digital signatures and SLH-DSA as a hash-based signature alternative. That does not mean the cryptographic asteroid hits tomorrow. It means migration takes years, and “harvest now, decrypt later” attacks make long-lived secrets vulnerable today.

The upbeat view is that quantum computing has crossed from physics stunt to engineering programme. The field now has credible error-correction demonstrations, competing hardware platforms, serious capital, better software, and clearer benchmarks. The lightly sarcastic view is that every press release still needs to be hosed down before it enters the house.

The most honest forecast is this: the next five years will decide whether quantum computing becomes a practical accelerator for selected high-value problems or remains a magnificent laboratory beast with excellent branding. The odds of progress are good. The odds of disappointment are also good. That is what happens when humans try to industrialise a machine built from probability, phase, entanglement and optimism.

Quantum computing is not magic. It is physics under extreme management. And like most ambitious productions, the miracle will happen only after the crew solves the wiring, the cooling, the timing, the budget and the part where the star keeps forgetting its lines.

=====

I was torn on today’s cine-pick. There are two topic-adjacent films I think fit today’s rant. The first is Lapsis (2020), a sci-fi mystery/thriller about gig work in the New Economy. The second choice is a bit more mainstream: Everything Everywhere All at Once (2022), a sci-fi comedy karate Heinlein-esque romp where Chinese immigrants save the Universe. Your call.