Quantum Computing in 2026: What's Real, What's Still Roadmap

In November 2025, IBM said something bold in public: it expected the industry to reach a verified quantum advantage — a task a quantum computer does better than any classical computer, checkable well enough that skeptics can't wave it away — sometime in 2026. Eight months later, that happened. Not everything in quantum computing moved that fast this year. But that one claim is worth taking seriously, and understanding why it's different from the "advantage" claims that came before it is the most useful thing to take from the state of the field in 2026.

The claim that actually landed

On July 30, 2026, IBM and the Finland-based quantum software company Algorithmiq announced they'd used an IBM Quantum Heron processor to simulate heterogeneous quantum matter — materials with irregular, localized variations in how information, energy and particles move through them, the kind of structure that shows up in real catalysts and battery electrolytes, not textbook toy systems. The result itself isn't the headline. The headline is what happened next: the team put the result on a public "Quantum Advantage Tracker" and open-sourced Algorithmiq's own classical simulation method, called monoprop, specifically so other researchers could try to match or beat it classically. Eight months on, nobody has, across the full range of problems the work covers.

That's the part that separates this from most of what's been called "quantum advantage" or "quantum supremacy" in prior years. Earlier claims — including some of Google's — were often contested precisely because there was no good way to check them: the whole point of the demonstration was that a classical computer couldn't reproduce the answer, which also meant nobody could independently confirm the quantum computer got it right. IBM and Algorithmiq built the challenge into the announcement instead of leaving it to be litigated afterward. That's a meaningfully more defensible way to claim a milestone, even if — worth saying plainly — it's still IBM and a commercial partner grading their own homework, on a public leaderboard other researchers are free to attack.

The 2025 groundwork this built on

The 2026 result didn't come out of nowhere. In October 2025, Google's Quantum AI team published a Nature paper describing its own "Quantum Echoes" algorithm, run on the 105-qubit Willow chip, which it also called verifiable — the result could be reproduced on the same or a similar-caliber machine. Google's team reported it running roughly 13,000 times faster than the best classical algorithm on a leading supercomputer, and demonstrated a proof of concept using it to measure molecular geometry via nuclear magnetic resonance techniques on molecules with 15 and 28 atoms, matching conventional NMR results while surfacing information standard NMR doesn't provide. A few weeks later, in November 2025, IBM unveiled two new chips of its own — Nighthawk, a 120-qubit processor built specifically as IBM's platform for demonstrating quantum advantage, and Loon, an experimental ~112-qubit chip built to test the error-correcting codes fault tolerance will eventually need — alongside the public target of verified advantage in 2026 that the Algorithmiq result went on to meet.

Worth being precise about the dates here, since it's easy to round all of this into one undifferentiated "quantum had a big year" story: Quantum Echoes and Nighthawk/Loon are both late-2025 milestones. They're the setup, not the 2026 payoff — but they're also why the 2026 payoff wasn't a surprise to anyone following the field closely.

What's still genuinely roadmap, not real

None of this means fault-tolerant quantum computing — a machine that can run long, complex programs without errors piling up faster than it can correct them — is here. It isn't, and nobody serious is claiming it is. The clearest marker of how far that milestone still is: in January 2026, researchers at ETH Zurich and the Paul Scherrer Institute, working with theorists at RWTH Aachen and Forschungszentrum Jülich, published the first demonstration of "lattice surgery" on superconducting qubits — a technique for performing operations between logical qubits (qubits built from many physical qubits working together for error correction) while continuing to correct errors as the operation happens. Their setup split a logical qubit built from 17 physical qubits into two entangled logical qubits, mid-surgery, without losing error correction. It's a genuine first, and a necessary building block for scaling error-corrected quantum computers. It's also, by the researchers' own account, a small one: they note that full protection against both major error types would need 41 physical qubits per logical qubit in this scheme, not 17 — and useful fault-tolerant machines will need many logical qubits working together, not one. IBM's own public target for a fault-tolerant machine is 2029, three years past the advantage milestone it just hit.

That's the honest shape of "roadmap" right now: real, peer-reviewed progress, published on a normal research cadence, still multiple hardware generations away from the thing that would let quantum computers run useful, general-purpose programs unsupervised.

The money doesn't care about that distinction

Investment isn't waiting for fault tolerance to arrive. By McKinsey's Quantum Technology Monitor 2026, more than 300 companies worldwide are now actively adopting quantum computing in some form, up sharply from a landscape that was almost entirely public-sector and research-lab a few years ago — 72% of current quantum computing use now sits with privately owned companies. Investment in quantum startups hit $12.6 billion in 2025, a 6.3x jump from 2024, and quantum computing companies collectively crossed $1 billion in revenue that year, a figure McKinsey projects will reach $4.4 billion by 2028. A third of large companies with active quantum programs put more than $10 million behind them in 2025; 7% spent over $50 million. McKinsey's own framing is careful, not breathless: it calls 2026 a "commercial tipping point" while explicitly acknowledging that near- to medium-term return on investment remains hard to quantify for most companies in the space, and that most current applications are still experimental or hybrid rather than fully deployed. The firm's long-range number — up to $2.7 trillion in economic value by 2035 — is a decade-out projection, not a 2026 result.

What this means if you're not a physicist

For nearly everyone reading this, quantum computing in 2026 is not yet a tool you'll touch directly — it's still infrastructure being built by a small number of very well-capitalized labs, for problems (novel-materials simulation, drug discovery, certain optimization problems) that don't look like the AI tools covered elsewhere on this site. What changed this year isn't that it became usable. It's that one of its central claims — that a quantum computer can beat a classical one at something real, in a way other scientists can actually check — stopped being a promise and became a verified, still-being-challenged fact. That's a smaller step than the headlines around it suggest, and also a genuinely different kind of milestone than the field has produced before. Worth knowing the difference the next time a "quantum breakthrough" headline crosses your feed.