SUPERPOSITIONED

A Note to the Reader

This report is written for a mixed audience: senior technologists who want precise specifications, and investors, policymakers, and strategic planners who need the big picture without a physics degree. To serve both groups, the report weaves together two layers: the main text provides full technical detail with specific numbers, citations, and analysis, while plain-language explanations are interspersed throughout to translate key concepts into everyday terms. If you already know what a qubit is or how error correction works, you can skim the simpler passages. If not, they are your on-ramp.

Key Concepts: A 60-Second Primer

Before diving in, here are the foundational ideas that appear on almost every page of this report:

Qubit — The basic unit of quantum information. A classical computer bit can be either 0 or 1. A qubit can be 0, 1, or a “superposition” of both simultaneously, which is what gives quantum computers their theoretical power. Think of a coin: a classical bit is a coin lying flat, showing heads or tails. A qubit is the coin mid-spin, simultaneously partaking in both states until you look at it.

Physical Qubit vs. Logical Qubit — A physical qubit is the actual hardware device (an ion, a superconducting circuit, a photon). A logical qubit is a collection of physical qubits working together to represent one reliable unit of information, using error correction to compensate for the noisiness of the individual physical qubits.

Gate Fidelity (“Nines”) — The accuracy of a quantum operation (a “gate”). 99% fidelity means 1 error per 100 operations. 99.9% means 1 error per 1,000. 99.99% means 1 error per 10,000. Each additional “nine” is an order-of-magnitude improvement. This report tracks fidelity “nines” because they are the single most predictive measure of when quantum computers become useful.

Coherence Time — How long a qubit retains its quantum information before environmental noise destroys it. Measured in microseconds (μs) for superconducting qubits or seconds for trapped ions. Longer is better: if your computation takes longer than your qubits stay coherent, you lose your answer to noise.

Quantum Error Correction (QEC) — The technique of spreading quantum information across many physical qubits so that errors can be detected and fixed without destroying the computation. It is the single most important enabling technology for practical quantum computing.

Surface Code — The most widely studied error correction code for quantum computing. It arranges physical qubits in a 2D grid and is relatively simple to implement on superconducting chips. Its downside: it requires roughly 1,000 physical qubits per logical qubit at current noise levels—a huge overhead.

Quantum Advantage — The point at which a quantum computer solves a specific problem faster, cheaper, or better than any classical computer can. “Quantum supremacy” (an older term) referred to solving any problem faster, even an artificial one. “Quantum advantage” increasingly refers to solving a useful problem faster.

Introduction: The Year Quantum Got Real

In a clean room in Broomfield, Colorado, ninety-eight barium ions hover in electromagnetic suspension above a chip the size of a thumbnail. Each ion is a qubit—a quantum bit—and together they constitute the computational heart of Quantinuum’s Helios, the most accurate quantum computer^[9] ever built for commercial use. The ions are spaced mere microns apart, held in place by precisely tuned electric fields, shuttled through junctions and logic zones by a choreography of laser pulses that would have seemed impossible five years ago. The machine's two-qubit gate fidelity is 99.921%. Using the [[4,2,2]] Iceberg error-detection code, it produced ninety-four error-detected logical qubits—and through code concatenation, forty-eight logical qubits at a 2:1 physical-to-logical encoding ratio—numbers that, as recently as 2023, most experts in the field considered unattainable within this decade. An important distinction: these are error-detected qubits, not error-corrected qubits. The [[4,2,2]] code is a distance-2 code that detects errors through post-selection (discarding faulty results) rather than actively correcting them—a meaningful intermediate step, but not yet the fault-tolerant error correction that will be required at scale. In March 2026, Quantinuum extended Helios further: 48 logical qubits under active error correction at logical gate errors of approximately 1×10^-4 — below Helios's raw physical two-qubit error rate — and, separately on the error-detection layer, a 94-qubit Greenberger-Horne-Zeilinger state at 94.9% fidelity (parity-oscillation lower bound)^[10]. This is the first commercial demonstration of active error correction below the physical error rate on dozens of logical qubits at once, though underlying peer review is still pending.

Three thousand miles away, in Google’s quantum computing facility in Santa Barbara, a different kind of breakthrough sits inside a dilution refrigerator cooled to fifteen millikelvins—colder than outer space. Google’s Willow chip, a 105-qubit superconducting processor, achieved something in December 2024 that the quantum computing community had been chasing for nearly thirty years: below-threshold error correction. For the first time since Peter Shor proposed quantum error correction in 1995^[2], adding more qubits to a system made it more accurate, not less. Each time the Google team scaled from a 3×3 grid to 5×5 to 7×7, the logical error rate halved. The resulting logical qubit lasted 2.4 times longer than the best physical qubit on the chip. The paper was published in Nature^[1]. It was, in the understated language of the field, a qualitative change.

And in Redmond, Washington, Microsoft unveiled Majorana 1^[3] in February 2025— an eight-qubit chip built on a “topoconductor,” a new class of material engineered to host Majorana quasiparticles. If the physics holds, topological qubits could offer built-in error protection at a fraction of the overhead required by other architectures. Microsoft’s CEO called it “quantum’s transistor moment.” Physicists at the APS Global Physics Summit, however, were considerably more skeptical, and the scientific debate remains unresolved (see Section 1.1). It may be the most consequential open question in quantum physics today.

These three events—Quantinuum's Helios, Google's Willow, Microsoft's Majorana 1—occurred within a thirteen-month window spanning late 2024 through late 2025. Together with a cascade of supporting milestones—Princeton's transmon coherence time exceeding one millisecond^[4] (fifteen times the industry standard), IBM’s demonstration of all hardware elements needed for fault-tolerant computing, QuEra’s techniques for reducing error correction overhead^[6] by up to 100×, IonQ's reported 12% speedup over classical HPC in a medical device simulation^[5] (a contested, unverified result), and Google's October 2025 "Quantum Echoes" demonstration — an out-of-time-ordered-correlator algorithm on 65 of Willow's 105 qubits at a reported 13,000× the speed of the best known classical algorithm running on Frontier — the first claim of quantum advantage that is reproducible on another quantum device (in contrast to the 2019 and 2024 random-circuit-sampling results)^[11]—they constitute the most concentrated period of progress in the history of quantum computing.

The Thesis

Between 2024 and 2025, quantum computing crossed two thresholds that had stood for decades—below-threshold error correction and near-2:1 physical-to-logical qubit encoding via error detection—and produced a contested early signal of quantum advantage over classical high-performance computing. These are not incremental improvements. They are individually modest in absolute terms, but collectively demonstrate that the fundamental physics works at small scale. The question is shifting from if to when and who—but significant open questions remain about whether the physics continues to cooperate at commercially relevant scale.

This report is an attempt to answer those questions with precision. It is not a hype piece, and it is not a dismissal. It is a bottoms-up accounting of where quantum computing actually stands—measured in gate fidelity nines, logical qubit counts, error correction overhead ratios, and dollars invested—and where the trendlines point. The analytical method borrows from Leopold Aschenbrenner’s Situational Awareness (June 2024)^[7], which decomposed AI progress into measurable axes of improvement to make an extrapolative case about the trajectory of artificial intelligence. The quantum equivalent of “counting the OOMs” (orders of magnitude of effective compute) is counting the nines—tracking each additional nine of gate fidelity (99.9% → 99.99% → 99.999%) because each additional nine exponentially reduces the overhead required for error correction, and therefore exponentially increases the useful computational power of a quantum machine.

The Information Asymmetry

Most mainstream coverage of quantum computing oscillates between two poles: uncritical hype (“quantum will break all encryption tomorrow”) and dismissive skepticism (“quantum is always ten years away”). The reality, visible to those tracking the primary literature, is more nuanced and considerably more interesting. The hype is wrong because: current quantum computers cannot break any encryption in use today; cryptographically relevant quantum computers remain 10–25+ years out under most hardware trajectories; many “quantum advantage” claims have been challenged or matched by improved classical algorithms; and quantum machine learning—the application category with the largest projected market value—remains largely theoretical. The caveat on the timeline tightened in 2025–2026: Craig Gidney's May 2025 analysis reduced the physical-qubit cost of factoring RSA-2048 from ~20 million to under 1 million noisy qubits—a roughly 20× algorithmic compression since 2019, holding the physical-gate-error assumption (~10⁻³) and surface-code cycle time (~1 μs) constant^[12]—and Google Quantum AI's March 2026 cryptocurrency-oriented analysis showed that breaking 256-bit elliptic-curve cryptography (ECDSA-256, the signature scheme underpinning Bitcoin and Ethereum) could be accomplished with fewer than 1,200 logical qubits in a runtime of minutes on fewer than 500,000 physical qubits—again under the same ~10⁻³ / ~1 μs superconducting assumption^[13]. These are algorithmic compressions of the threat, not new hardware; they shift the harvest-now-decrypt-later calculus without changing when a machine of that scale will exist, and the real-world timeline still depends on the underlying hardware reaching those error rates at scale.

The skepticism is wrong because: the pace of error correction progress in 2024–2025 exceeded what the field's own researchers predicted; the physical-to-logical qubit ratio has improved from ~1,000:1 to ~2:1 in trapped-ion systems within two years; investment has tripled year-over-year; DARPA is spending real money^[8] evaluating whether utility-scale quantum computing is achievable by 2033; and eleven companies across five different qubit modalities have advanced to the second stage of that evaluation.