Note from Editor Z: We first covered quantum computing back in February 2025. In the time since, the field has moved from lab curiosity toward something closer to an engineering race — and it has also produced one of the messier scientific disputes in recent memory. This is an update, not a repeat: what changed, what the original piece got wrong, and what a reader should actually take away from the noise.
What Makes a Quantum Computer Different
A classical computer stores information as bits: each one is a 0 or a 1. A quantum computer uses qubits, which can hold a combination of both states at once — a property called superposition. Pair that with entanglement, where the state of one qubit is correlated with another no matter the physical distance between them, and you get a system that can represent an exponentially larger space of possibilities than the same number of classical bits.
That doesn't make quantum computers faster at everything — for most everyday tasks, like loading a webpage or running a spreadsheet, they're irrelevant or worse. Their advantage shows up only on specific problem types: simulating molecules and materials, optimizing across huge combinations of variables, and certain kinds of cryptographic math. Google's Willow chip, one of the most closely watched machines in the field, has 105 superconducting qubits — a modest number compared to the roughly 1,000-qubit chips some vendors now advertise, but Willow's significance is about error rates, not raw qubit count, which is the harder problem.
Where the Major Labs Actually Stand
Google: From "Supremacy" Claims to a Verifiable Result
The original version of this post said Google, IBM, and Microsoft had all achieved "quantum supremacy." That was an overstatement even in 2025, and it's worth correcting now: Google is the only one of the three that has run the specific class of experiments the term refers to. Its 2019 Sycamore demonstration was real but scientifically contested, because the task it solved had no practical use and couldn't be independently checked. Willow changed that. In a December 2024 Nature paper, Google showed that as it scaled its error-correcting code from a 3x3 to a 7x7 grid of physical qubits, the logical error rate dropped by roughly a factor of 2.14 with each step up — the first clear demonstration that adding more qubits to a well-designed system makes it more reliable rather than noisier, a threshold researchers had chased for over a decade.
Then, in October 2025, Google went further with an algorithm called Quantum Echoes, run on a 65-qubit subset of Willow. It solved a molecular-structure problem about 13,000 times faster than the best available classical algorithm on a supercomputer — and, working with UC Berkeley, Google verified the quantum-computed predictions against real NMR spectroscopy measurements on two actual molecules. That verifiability is the detail the original 2025 sampling-based "supremacy" claims lacked, and it's why researchers are treating this one differently.
IBM: A Roadmap, Not a Supremacy Claim
IBM has deliberately avoided "supremacy" language, instead publishing a public hardware roadmap. In 2025 it shipped Nighthawk, a 120-qubit processor, and Loon, a chip built to validate the error-correcting codes IBM plans to use at scale. Its next step, a system called Kookaburra with roughly 4,158 physical qubits connected via a new error-correction scheme (qLDPC codes) and a "quantum long-term memory" module, is the machine IBM has publicly targeted to demonstrate quantum advantage on a genuinely useful workload by the end of 2026 — a claim that, unlike Google's, hasn't been tested yet.
Microsoft: The Most Disputed Claim in the Field
This is the biggest story the original post missed entirely, because it hadn't happened yet: in February 2025, Microsoft announced Majorana 1, a chip it said contained eight "topological" qubits — a fundamentally different, theoretically more error-resistant qubit design based on exotic quasiparticles predicted by physicist Ettore Majorana. The announcement landed badly with much of the physics community. The accompanying Nature paper showed only one of the two measurement types (Z measurements, not X) needed to actually prove a topological qubit exists, and critics — including physicist Henry Legg of the University of St Andrews, who called unverified topological-qubit claims "essentially selling a fairy tale" — argued Microsoft hadn't demonstrated what it claimed. Microsoft followed up in June 2026 with Majorana 2, described as improved using AI-assisted analysis of the underlying physics, but independent physicists say the same core objection still applies: no outside lab has independently confirmed the topological qubit exists as described. This is an open scientific dispute, not a settled milestone, and readers should treat Microsoft's qubit-count claims accordingly.
Quantinuum, IonQ, and the Non-Superconducting Field
Superconducting qubits (Google, IBM) aren't the only approach. Quantinuum, using trapped-ion hardware, went from 12 logical qubits on its H2 system in 2025 to 94 logical qubits on its H-series by March 2026. QuEra, using neutral atoms held in place by laser tweezers, has demonstrated 96 logical qubits. IonQ reported a two-qubit gate fidelity of 99.99% in a prototype in October 2025 — matched by Oxford Ionics on its own trapped-ion hardware — figures that superconducting systems (around 99.7%) haven't caught up to.
Comparing the Hardware Approaches
There is no single "best" qubit technology yet — each makes a different trade-off, and the field genuinely doesn't know which will scale best.
| Technology | Leading builders | Two-qubit fidelity | Operating conditions | Main open problem |
|---|---|---|---|---|
| Superconducting | Google, IBM | ~99.7% | ~15 millikelvin, dilution refrigerator | Wiring and cooling complexity at large scale |
| Trapped ion | IonQ, Quantinuum, Oxford Ionics | 99.91–99.99% | Room-temperature vacuum chamber | Gate speed (microseconds) and linking multiple traps |
| Neutral atom | QuEra, Atom Computing, Pasqal | ~99.48% | Room-temperature vacuum, optical tweezers | Atom loss and reloading between runs |
| Topological | Microsoft | Not independently verified | Cryogenic, semiconductor-based | Whether the qubit exists as claimed at all |
| Photonic | PsiQuantum, Xanadu | ~93.8% | Room temperature | Photon loss; fidelity currently trails other approaches |
What This Can Actually Do Right Now
The original post listed five industries quantum computing would "transform," without a single real example. Here are two that exist today. In March 2026, Cleveland Clinic and IBM published the first quantum-classical hybrid simulation of the electronic structure of Trp-cage, a 303-atom miniprotein, run on an IBM Quantum Heron r2 processor — a genuine, published result, though on a small protein chosen specifically because it's tractable, not a drug-discovery breakthrough by itself. The same month, quantum-software firm Kvantify, Atom Computing, and Aarhus University launched a project called EarlyBIRDD to tackle a much harder, commercially relevant question: predicting how strongly a candidate drug molecule binds to its target protein.
In finance, banks are running early pilots on portfolio optimization and options pricing, where quantum algorithms can in principle evaluate more market scenarios than classical Monte Carlo methods in the same time. None of this is a deployed production system replacing classical infrastructure yet — it's pilots and published proofs of concept. The honest way to describe 2026's state of the art: quantum computers can now do a few genuinely useful things marginally better or faster than classical machines, on narrow, carefully chosen problems, and can be independently verified doing so. That's a real change from 2025's demonstrations, which were mostly benchmarks with no practical output at all.
Correcting the Cybersecurity Claim
The original post said quantum computing enables "new encryption methods [to] protect sensitive information more robustly." That has it backwards, and it's worth being direct about the correction: quantum computers are a threat to most of today's encryption, not a tool for strengthening it. RSA and elliptic-curve cryptography, which secure most of the internet's traffic, rely on math problems that a sufficiently large, fault-tolerant quantum computer could solve efficiently using an algorithm devised by mathematician Peter Shor in 1994. The resource estimate for breaking RSA-2048 has fallen sharply as algorithms improved — from around 20 million qubits in a 2019 estimate to under 1 million noisy qubits in 2025 projections — and a 2025 Google study suggested fewer than 500,000 physical qubits might eventually crack elliptic-curve encryption in minutes. No existing machine is remotely close to that scale yet; current chips top out in the low thousands of physical qubits.
The actual defensive response is called post-quantum cryptography: new classical (non-quantum) algorithms designed to resist attack by a future quantum computer. NIST finalized its first three standards in August 2024 — FIPS 203, 204, and 205 — and added a fourth algorithm, HQC, in 2025. Most security experts now put "Q-Day," the point a quantum computer could actually break current encryption, in the early-to-mid 2030s, but the real deadline is earlier than that: adversaries can harvest encrypted data now and decrypt it once the hardware catches up, so organizations with long-lived sensitive data are already being told to migrate to the new NIST standards, not wait for Q-Day to arrive.
What It Costs to Actually Try
None of this requires owning a quantum computer. IBM's Open Plan gives free cloud access to smaller quantum processors for learning and experimentation, and IBM also bills a pay-as-you-go rate by the minute for larger jobs. Amazon Braket charges per shot and per task — rates as low as $0.000425 per shot on Rigetti hardware — while Microsoft Azure Quantum offers both per-program pricing and monthly subscriptions that can run over $125,000 a month for access to Quantinuum's higher-end trapped-ion systems. For a mid-size organization wanting to run a real pilot rather than a toy example, current estimates put a meaningful project at $20,000 to $100,000 in cloud costs for the year — a fraction of what building or owning hardware would cost, and the reason cloud access, not ownership, is how virtually every company touching quantum computing today actually uses it.
What's Still Unsettled
IBM, Google, Quantinuum, and Microsoft have all pointed to around 2029 as a target for a genuinely useful, fully fault-tolerant quantum computer — but that's a shared target, not a shared certainty, and the Majorana dispute is a live reminder that not every announcement in this field survives scrutiny. The more interesting open question isn't when quantum computers arrive, but what shape they'll take: nothing in the current data suggests one qubit technology will win outright. Superconducting, trapped-ion, neutral-atom, and photonic systems each have real advantages, and the more likely outcome is several approaches coexisting, suited to different problems, the way GPUs and CPUs coexist today rather than one replacing the other.
The pace since early 2025 has been real, but so has the gap between press releases and peer-reviewed verification — and that gap is the thing worth watching most closely as this technology moves out of the lab.
-EditorZ
Post a Comment