Home » Quantum Computing » IQM Says It Achieved Full-Chip Entanglement on 54 Qubits

IQM Says It Achieved Full-Chip Entanglement on 54 Qubits

IQM’s new fidelity figure clears the bar for real entanglement by just 0.016, and the two numbers behind that claim were measured under different test conditions, not a clean apples-to-apples…

Abstract network of glowing connected points illustrating full-chip entanglement in a quantum processor

Here’s a number worth sitting with: 0.516. That’s the score IQM says its newest quantum chip got on a standard test for genuine entanglement, where anything at or below 0.500 is statistically indistinguishable from noise. The company announced on October 29, 2026, that it achieved full-chip entanglement across all 54 qubits on its Emerald processor at once, not just a handful of them, and says the capability is already live for people renting time on the machine over the cloud.

  • What happened: IQM’s cloud-accessible Emerald processor entangled all 54 of its qubits in one circuit, a result the company calls full-chip performance.
  • The number that matters: A GHZ-state fidelity of 0.516, measured against a classical noise threshold of 0.500.
  • Who gets it: IQM’s cloud customers, starting now, according to the company. No sign-up details or pricing changes are mentioned.
  • What’s missing: A peer-reviewed paper. Every figure here comes from IQM’s own announcement, not outside testing.

IQM is a Finnish quantum computing company based in Espoo that builds superconducting processors, the same broad approach Google and IBM use. Its Emerald chip is one of the systems customers can already book cloud time on. Before this result, the company says, someone running an algorithm on Emerald could reliably put around 30 of its 54 qubits to work in a single run. The rest sat idle or added too much noise to trust.

What Full-Chip Entanglement Actually Means

The test IQM ran is called a GHZ state, named for the three physicists who first described it. Treat it as quantum computing’s stress test for entanglement: link every qubit on the chip into one shared quantum state, then measure how cleanly that state holds together. A score at or below 0.500 means the result is no better than random guessing. IQM reports 0.515 at 40 qubits without any cleanup of its onboard reading errors, and 0.516 across the full 54 qubits once it applied a fix for those reading errors, a step called readout error mitigation.

Getting there took roughly a year of calibration work, per the release. The real fix was software, not new hardware: an automated routine that picks which two-qubit operations run, and in what order, to keep every qubit as busy as possible so none of them sit idle and lose their quantum state while waiting on the others. IQM’s chief technology officer, Craig Ciesla, framed the challenge this way:

“Entangling the entire chip means every qubit and every connection between them has to work correctly in the same circuit,” said Craig Ciesla, CTO of IQM Quantum Computers.

That’s the crux of the problem. A quantum chip isn’t one part, it’s dozens of fragile components that all have to behave correctly at the same time. IQM also reports up to a 50 percent cut in the error rate of its quantum operations generally, though the release doesn’t say what that figure is measured against.

Why Using the Whole Chip Is the Real Bottleneck

Adding more qubits to a chip is the easy headline. Keeping all of them accurate while they run together is the part that actually limits what a quantum computer can do. RIKEN hit a version of this same wall last month with silicon spin qubits, where driving several qubits at once knocked down their accuracy through a different kind of interference. IQM’s problem and fix are specific to its superconducting hardware, but the underlying story repeats across the industry: a quantum computer’s real capacity isn’t its qubit count, it’s how many of those qubits you can actually trust in one run.

That distinction is why going from “about 30 usable qubits” to “all 54, together” is a bigger deal than it sounds. It doesn’t make the chip bigger. It makes more of the chip usable for any single problem, which is a different, and arguably harder, kind of progress. Other companies are chasing that same goal from different angles. Quantinuum, for instance, has focused on cutting the overhead its error correction needs, rather than squeezing more performance out of every qubit on one chip.

Coverage of IQM’s announcement from The Quantum Insider and Quantum Zeitgeist both ran within hours of the release. Neither outlet added independent verification or quoted anyone outside IQM, which is worth knowing before treating this as confirmed from multiple angles.

Metric Before This Result After, Per IQM
Qubits usable in one algorithm Around 30 of 54 All 54
Full-chip GHZ fidelity Not previously reported 0.516 (threshold: 0.500)
Error rate on quantum operations Baseline not stated Up to 50% lower, per IQM
Cloud access Partial-chip performance Full-chip performance, live now

Source: IQM press release, Oct. 29, 2026.

Why It Matters to You

  • This isn’t a consumer product. The people who benefit directly are researchers and developers renting time on IQM’s cloud, not someone shopping for a laptop or a phone.
  • The indirect payoff is real but still a few steps removed: more usable qubits per run means scientists testing drug-discovery, materials, or logistics algorithms on Emerald can try bigger problems without hitting a performance wall at 30 qubits.
  • Access is immediate, per IQM, for existing and new cloud customers. The company hasn’t published pricing or said whether new signups face a waitlist.
  • None of this touches encryption risk. Entangling 54 noisy qubits for a lab benchmark is a long way from the millions of clean, error-corrected qubits that would take.

The Catch

  • This is a company press release, not a peer-reviewed paper. IQM cites no preprint, no journal, and says no outside lab checked the numbers.
  • The headline fidelity, 0.516, sits just 0.016 above the 0.500 line that separates real entanglement from statistical noise. That’s a thin margin to hang a “full-chip” claim on.
  • The two fidelity figures weren’t measured the same way: 0.515 at 40 qubits came without readout error mitigation, while 0.516 at 54 qubits came with it applied. Comparing them side by side overstates how cleanly the result actually scales.
  • The “up to 50 percent” error-rate cut has no stated baseline, so there’s no way to judge how large that improvement really is.
  • IQM calls this one of “the strongest reported” results for any commercial quantum computer, but cites no benchmark or outside ranking to back that up.

None of that means the result is wrong. It means it’s unverified, which is a different thing, and worth remembering the next time a company’s own press release doubles as the only source for a record-setting claim.

Watch for two things from here: whether IQM backs this up with an actual paper, and whether a competitor answers with a full-chip entanglement claim of its own at a higher qubit count. Until one of those happens, file this under promising but unverified, the same drawer where most quantum computing headlines belong.

Sources

Related Reading