Article 8 of 9 — Foundations of Quantum Computing
Strip away the press releases and the stock-market excitement, and a fair question remains: where is quantum computing actually, right now? The answer in 2026 is genuinely interesting — real, accelerating engineering progress — but it's surrounded by a fog of overstatement that makes it hard to read. This article lays out who's building what, the term that best describes the current era, the milestones that matter, and the persistent gap between an announcement and a useful machine.
The basics map the landscape and the era. Going Deeper covers how to read claims critically and the honest split between optimists and skeptics.
The basics: the NISQ era
The phrase that best captures today is NISQ — Noisy Intermediate-Scale Quantum. It describes exactly the machines we have: intermediate in scale (dozens to a few hundred physical qubits), and noisy (no full error correction yet, so errors limit how long a computation can run). NISQ machines are real and accessible — many are available over the cloud — and they're valuable for research and experimentation. What they are not, yet, is capable of broadly useful computations that beat classical computers on practical problems. We are between the proof-of-concept past and the fault-tolerant future.
The basics: who's building what
The field is a crowded race across the hardware approaches from Article 3:
- IBM (superconducting) has published a detailed roadmap toward a fault-tolerant machine — named Starling — targeted for around 2029, and has spoken of demonstrating verified quantum advantage by the end of 2026. It has rolled out successive processors emphasizing error correction.
- Google (superconducting) demonstrated below-threshold error correction with its Willow chip (late 2024) — a landmark showing errors falling as the system scales.
- Microsoft (topological) unveiled Majorana 1 in 2025, a bet on intrinsically more stable qubits, though the approach is the least mature.
- IonQ and Quantinuum (trapped ion) emphasize high fidelity; Quantinuum filed for what would be a landmark public listing.
- QuEra and Pasqal (neutral atom), PsiQuantum and Xanadu (photonic), plus Rigetti, D-Wave (quantum annealing), Intel, and Amazon round out a deep field.
- National programs in the US, China, the EU, and the UK are investing heavily, treating quantum as strategic infrastructure.
A notable theme of 2025 was a wave — some called it a "tsunami" — of error-correction results from many groups at once, prompting some researchers to argue the field has shifted from an open physics problem to an engineering one. That framing is meaningful but, as the deeper section notes, contested.
The basics: the gap between a press release and a machine
Here's the discipline that matters most when reading quantum news. Announcements cluster around impressive-sounding metrics — a new record qubit count, a "quantum advantage" benchmark, a "logical qubit" demonstration. Each can be real and still be a long way from a useful computer:
- A high qubit count says nothing without fidelity, connectivity, and how many are logical (Article 4).
- A quantum-advantage result is often on a contrived benchmark with no practical use (Article 5).
- A logical-qubit demonstration is frequently a narrow proof-of-concept — preserving a qubit, not running a full algorithm.
The progress is genuine. The leap from these milestones to a machine that does valuable work classical computers can't is still large.
Going deeper: reading claims, and the optimist–skeptic split
For readers who want to navigate the field themselves, two things help.
A credibility checklist. When you see a quantum claim, ask: Is it peer-reviewed or a press release? Does it report logical or merely physical qubits? Did it run a real algorithm or preserve a qubit in memory? Was the benchmark practical or contrived? Is the "advantage" measured against the best classical method, or an easy one? These questions separate substance from spin faster than any single headline number. The most informative metrics are usually the unglamorous ones — error rates, gate fidelities, and how deep a circuit can run — not the qubit count that leads the press release.
The honest optimist–skeptic split. Serious people disagree about pace. Optimists, pointing to the 2025 error-correction surge and rapidly improving codes, argue useful fault-tolerant machines could arrive within several years and that the hard part is now engineering. Skeptics counter that logical error rates remain far from what algorithms need, that overhead is enormous, and that scaling error correction may reveal problems invisible at small scale — making useful machines a decade or more away, if achievable on current paths. Both camps include respected physicists and engineers. The intellectually honest position is to report the disagreement rather than adopt one side's confidence.
Roadmaps are intentions, not guarantees. Company roadmaps (IBM's 2029 target, others') are valuable signals of direction and seriousness — but the field's history is littered with timelines that slipped. A roadmap is a plan made under uncertainty, not a delivery date. Treat dated promises, especially round-number ones, with appropriate caution.
The fair summary: 2026 finds quantum computing in a period of real, accelerating engineering progress, firmly in the noisy intermediate era, with fault tolerance demonstrated only in early, partial forms. The science is advancing genuinely; the distance to broad usefulness remains substantial; and the surrounding hype makes careful reading essential.
The takeaway
Quantum computing in 2026 sits in the NISQ era — noisy, intermediate-scale machines that are real and useful for research but not yet for broadly beating classical computers. A crowded field of companies and national programs is racing across rival hardware approaches, and 2024–2025 brought genuine milestones, especially in error correction. But qubit-count, advantage, and logical-qubit headlines each need careful reading, serious experts disagree about the timeline, and roadmaps are intentions rather than guarantees. Real progress, real distance remaining.
What people commonly get wrong
- Reading qubit counts as the measure of progress. Fidelity, connectivity, error rates, and logical-qubit counts matter more.
- Trusting press releases over peer review. The credibility of the source and the specifics of the claim are what count.
- Taking roadmap dates as guarantees. They're plans under deep uncertainty; quantum timelines have a long history of slipping.
- Assuming the optimists (or skeptics) are obviously right. The timeline is genuinely contested among serious experts.
- Confusing a milestone with usefulness. Below-threshold demos and advantage benchmarks are real but far from a valuable machine.
This article is educational and is not investment or professional advice. The field moves quickly; specific companies, milestones, and roadmaps reflect 2025–2026 reporting and should be refreshed at publish time.
Sources for context: company roadmaps and announcements from IBM, Google, Microsoft, IonQ, Quantinuum, and others; reporting in Nature, Network World, The Quantum Insider, and technology press on 2025–2026 error-correction milestones and the NISQ era. Refresh at publish time.
Next in the series: Article 9 — Outlook, Risks & Cutting Through the Noise: the honest uncertainty on timelines, the investment landscape, and a toolkit for reading overhyped announcements.
