In June 2026, Microsoft said its new Majorana 2 chip held quantum states for an average of 20 seconds, up from milliseconds on the previous version. A few months earlier, IBM’s chief executive told analysts that the company’s partners would show the first real examples of quantum advantage before the year was out. Toronto’s Xanadu rang the opening bell as a public company on both Nasdaq and the TSX. If you only read the headlines, quantum computing in 2026 looks like a field that has finally arrived.
It hasn’t, quite. What has changed is subtler and, frankly, more interesting. For the first time, several companies can show that adding more hardware makes their machines more reliable instead of less. That is the hinge the whole industry swings on. The useful machines everyone keeps promising are still a few years out, and some of the loudest claims remain contested.
Here is where quantum computing in 2026 actually stands, and why Canada has an unusually large stake in it.
Qubits are cheap. Reliable qubits are not.
For most of the last decade, quantum companies competed on qubit counts. More qubits sounded like more power, and the numbers made for tidy press releases. The problem is that physical qubits are fragile. Heat, stray radiation and tiny control errors knock them out of their quantum state, and every operation adds a little more noise. Run a long enough calculation and the answer turns to static.
The fix is quantum error correction. You spread one “logical” qubit across many physical ones and constantly check for errors without destroying the information you care about. It works only if your physical qubits are already good enough, a point physicists call the threshold. Below it, bigger error-correcting codes suppress errors faster than they create them. Above it, adding qubits just adds trouble.
So when you read about a quantum milestone in 2026, the useful questions are not “how many qubits?” but:
- How many logical qubits, and how low is their error rate?
- How many operations can the machine run before errors win?
- Can a classical computer check the answer, or reproduce it?
- Does the result solve a problem anyone outside a physics lab cares about?
Google Willow: the below-threshold moment
Google set the tone in December 2024 with Willow, a 105-qubit superconducting chip. The headline was a random circuit sampling benchmark finished in under five minutes that Google said would take a top supercomputer 10 septillion years. Impressive, and useless.
The more important result was quieter. Google encoded logical qubits on grids of 3×3, then 5×5, then 7×7 physical qubits, and the logical error rate roughly halved at each step. That was the first convincing demonstration that a superconducting system could operate below threshold. In the company’s own announcement, it called this a historic step toward scaling.
In October 2025, Google followed up with an algorithm it calls Quantum Echoes, which runs a computation forward, nudges the system, and runs it backward to learn about the structure of molecules. Google said it ran about 13,000 times faster on Willow than on the best classical supercomputers and, crucially, that the result can be verified. The hardware numbers it published alongside were strong: 99.97% fidelity for single-qubit gates, 99.88% for two-qubit gates, and a trillion measurements over the course of the project. Verifiable is the key word. Earlier “supremacy” claims were hard to check and easy to chip away at with clever classical algorithms.
IBM’s roadmap: Nighthawk now, Starling in 2029
IBM takes a different path. Instead of the surface code Google uses, it is betting on a family of codes called qLDPC, which need far fewer physical qubits per logical qubit but require longer-range connections between qubits. That is hard to wire on a flat chip.
In November 2025, IBM unveiled two processors. Nighthawk has 120 qubits linked by 218 tunable couplers, and IBM expects it to handle circuits of about 7,500 gates by the end of 2026, rising to 15,000 by 2028. Loon is an experimental chip built to test the multi-layer routing and qubit reset tricks that qLDPC codes need. According to The Next Platform’s coverage, IBM expects the first verified cases of quantum advantage in 2026 using Nighthawk plus better software.
The bigger target is Starling, a fault-tolerant machine IBM says will run 100 million operations on 200 logical qubits, delivered from its Poughkeepsie data centre in 2029. After that comes Blue Jay, pitched at around 2,000 logical qubits sometime after 2033.
Here’s the catch with the 2026 advantage talk. When IBM’s Arvind Krishna repeated the promise on the company’s first-quarter call, the evidence he leaned on included work with Japan’s RIKEN, pairing a 77-qubit Heron chip with the Fugaku supercomputer to model iron-sulfur clusters. Analysts noted the quantum results beat one standard classical method but still trailed the best classical techniques. IBM uses a broad definition of advantage (cheaper, faster or more accurate), and that leaves a lot of room for argument.
Microsoft’s Majorana bet and the skeptics
Microsoft has spent close to two decades on a more exotic idea: topological qubits built from quasiparticles called Majorana zero modes. If they exist and behave as theory predicts, they would be naturally protected from many errors, which would slash the overhead of error correction.
In February 2025, Microsoft announced Majorana 1, an eight-qubit chip made from what it calls a “topoconductor” of indium arsenide and aluminum, and said it had a clear path to a million qubits. The physics community was not convinced. The peer-reviewed Nature paper that accompanied the launch stated that its measurements could not, on their own, prove the states were topological. At a major physics conference that spring, Henry Legg of the University of St Andrews described Microsoft’s evidence as “incredibly unconvincing,” and Pittsburgh’s Sergey Frolov said the key data looked like noise, as Science News reported. History adds to the wariness: a 2018 Majorana paper involving Microsoft-backed researchers was later retracted.
Majorana 2, announced on June 3, 2026, raises the stakes. Microsoft says it is 1,000 times more reliable than its predecessor, has 12 qubits, swaps aluminum for lead as the superconductor, and puts the company on track for a scalable machine by 2029. What it has not yet produced is independent confirmation that Majorana modes are really there. Until that arrives, treat Microsoft’s timeline as a hypothesis, not a forecast.
Everyone else is not standing still
Superconducting chips get the most attention, but other approaches are posting serious numbers. Quantinuum’s Helios, launched in November 2025, uses 98 trapped barium ions and reported two-qubit gate fidelity around 99.92%. Because every ion can interact with every other one, its error correction can be very efficient; the company has shown logical qubits encoded in as few as two physical qubits each, and it promises a fully fault-tolerant system called Apollo by 2029.
Neutral-atom firms such as QuEra and Atom Computing are scaling up too. Washington is hedging across all of them. DARPA’s Quantum Benchmarking Initiative, which asks whether any design can reach “utility scale” by 2033, moved 11 companies to its second stage in November 2025. And in May 2026, the US government handed out roughly $2 billion in CHIPS Act grants to nine quantum firms, taking equity stakes in return, with IBM getting about half.
Canada’s quantum companies, by the numbers
Canada punches well above its weight here, and three of DARPA’s 11 Stage B picks are Canadian: Nord Quantique, Photonic and Xanadu. Each is chasing fault tolerance in a different way.
Xanadu (Toronto)
Xanadu builds photonic quantum computers that use light and can run at room temperature for much of the system. Its 2022 Borealis experiment was one of the few non-superconducting quantum advantage demonstrations, and a January 2025 Nature paper laid out a modular way to network photonic machines together. The company is also behind PennyLane, a widely used open-source quantum programming library. After merging with the SPAC Crane Harbor Acquisition Corp., Xanadu began trading on March 27, 2026 under the ticker XNDU on Nasdaq and the TSX, raising roughly US$302 million in gross proceeds.
Photonic (Vancouver area)
Photonic uses spin qubits in silicon, called T-centres, that can be linked with light over ordinary optical fibre. The pitch is a distributed machine: instead of cramming everything into one fridge, you network modules together. Investors bought it. In May 2026, Photonic closed a US$200 million round at a US$2 billion valuation, led by Planet First Partners, with BDC, Export Development Canada, InBC and Bell Ventures joining earlier backers such as Microsoft, RBC and TELUS.
Nord Quantique (Sherbrooke)
Nord Quantique stores each qubit in microwave photons bouncing inside a superconducting cavity, a “bosonic” approach. In 2025 it showed a multimode version, using what it calls the Tesseract code, that protects against several types of error at once and showed no measurable decay across 32 rounds of correction. One caveat worth noting: about 12.6% of data per round was discarded through post-selection. The company aims for machines with more than 100 logical qubits by 2029, and argues it can get there with far fewer physical components than rivals.
D-Wave (born in B.C.)
D-Wave, founded in 1999, still has Canadian operations but is moving its corporate headquarters to Boca Raton, Florida. It sells annealing machines, which tackle optimization problems rather than running general algorithms, and in January 2026 it paid US$550 million for Quantum Circuits Inc. to add gate-model technology. Revenue is still small (about US$3.1 million in the second quarter of 2026), though first-half bookings jumped to US$35.5 million.
Ottawa’s play: keep the champions at home
Canada’s National Quantum Strategy, launched in 2022, committed $360 million over seven years across three missions: computing hardware and software, quantum-safe communications, and quantum sensing. It was modest next to US and European spending, and the worry in the sector was obvious. Canadian labs incubate the companies, then American capital and customers lure them south.
The federal answer, announced in December 2025, is the Canadian Quantum Champions Program, which commits $334.3 million over five years. Its first phase offers up to $23 million each to four companies building fault-tolerant machines: Anyon Systems, Nord Quantique, Photonic and Xanadu. The stated aim is to anchor talent and keep future systems built and operated in Canada. It’s a sensible bet, though $23 million is small change next to what US agencies now hand out. Whether it keeps a Nasdaq-listed Xanadu rooted in Toronto is an open question.
Quantum computing in 2026: how to read the next headline
Quantum computing in 2026 sits between two eras. The physics of error correction now works in the lab. The engineering of doing it at scale, cheaply and reliably, has barely started. If you’re a founder, investor or just a curious reader, a few rules of thumb will save you from both hype and cynicism:
- Ask for logical qubits and error rates. A press release that leads with physical qubit counts is usually hiding something.
- Check who verified it. Google’s Quantum Echoes is notable because the answer can be checked. Microsoft’s Majorana claims are notable because, so far, they can’t be independently.
- Watch the 2029 cluster. IBM’s Starling, Quantinuum’s Apollo, Microsoft’s scalable machine and Nord Quantique’s 100-logical-qubit target all land around the same year. Not all of them will hit it.
- Prepare for the security side now. A machine capable of breaking today’s public-key encryption is still distant, but data stolen today can be decrypted later. Moving to post-quantum cryptography is a project you can start this year.
- Don’t buy “quantum advantage” without a definition. Faster on a contrived benchmark, cheaper than one classical method, and useful for a real customer are three very different claims.
The honest summary: no one has a commercially useful, fault-tolerant quantum computer yet, and nobody will in 2026. But for the first time, the roadmaps are grounded in demonstrated physics rather than hope. And a surprising share of the most credible bets are being placed by companies from Toronto, Sherbrooke and Vancouver.
Sources and further reading
- Google: Meet Willow, our state-of-the-art quantum chip (Dec. 2024)
- Google: Our quantum hardware, the engine for verifiable quantum advantage (Oct. 2025)
- The Next Platform: IBM lets fly Nighthawk and Loon QPUs
- Science News: Physicists are mostly unconvinced by Microsoft’s topological chip
- DatacenterDynamics: Microsoft unveils Majorana 2 quantum chip
- MIT Technology Review: A new ion-based quantum computer makes error correction simpler
- DARPA: Quantum Benchmarking Initiative Stage B selections
- The Quantum Insider: Canadian government backs four quantum computer developers
- Quantum Computing Report: Photonic reaches $2B valuation
- Government of Canada: National Quantum Strategy
