IQM Entangles All 54 Qubits on Its Emerald Chip

IQM Quantum Computers says its IQM Emerald system has entangled all 54 of its qubits at the same time, and that cloud customers can now use the whole chip in a single computation. Until now, the Finnish company says, users could reliably use about 30 of Emerald’s qubits in one algorithm.

The result was announced in a press release and on IQM’s investor site. It is IQM’s own measurement and has not been peer reviewed.

The numbers

IQM used a Greenberger–Horne–Zeilinger (GHZ) state, a standard test in which every qubit in a group is linked into one shared quantum state. The measure is fidelity: how close the prepared state is to the ideal one.

According to IQM’s benchmark documentation, a fidelity above 0.5 is enough to show that the qubits are genuinely entangled with each other rather than just noisy. IQM reported:

0.515 at 40 qubits without readout error mitigation;
0.516 at all 54 qubits with readout error mitigation applied.

Readout error mitigation is a correction applied to the results afterwards to account for mistakes made when the qubits are measured. IQM did not give a 54-qubit figure without it.

The company says the results “rank among the strongest reported for any commercially available quantum computer”. It adds that studies on large quantum chips often use only 10 to 20 qubits, even when far more are available.

How IQM got there

The main change is an automated routine for preparing the state. It picks which two-qubit gates to use and in what order, trading gate quality against the time qubits spend idle, when they tend to pick up errors. IQM’s documentation describes a “tree” method that uses the latest calibration data to find the best path through the chip and recommends it for square-grid processors such as Emerald.

“Entangling the entire chip means every qubit and every connection between them has to work correctly in the same circuit,” said Craig Ciesla, IQM’s chief technology officer. “Getting there came down to being smarter about how we prepare the state itself, choosing the gate order to keep idle time to a minimum.”

IQM says the result builds on about a year of work on calibration, an improved gate set and error-reduction tools such as dynamical decoupling and qubit selection. Together these cut error rates for quantum operations by up to 50%, it says. The routine is part of a wider push to automate how its machines are tuned, alongside AI-driven calibration work.

IQM is listed on Nasdaq, employs more than 450 people and has major operations in Munich. It sells machines for on-site installation as well as cloud access, and is installing a system in Madrid with Telefónica.

Above the line, but only just

The 54-qubit result clears the entanglement threshold by 0.016, and only after readout correction. That is enough to show the whole chip can be entangled at once, but it leaves little margin, and IQM gave no error bars. Its own documentation notes that the tree method depends on daily calibration and can produce different circuits on different days, so the figure may vary between runs.

GHZ fidelity is also a narrow test. It shows qubits can be linked, not that the chip can run useful algorithms or correct its own errors, the focus of other recent error-correction and DARPA test results. The “strongest reported” claim comes from IQM, with no comparison published alongside it.

The practical gain is for cloud users, who can now run experiments across the whole processor. For European rivals such as Qilimanjaro, the bar is full-chip results customers can reproduce. The next step is a peer-reviewed paper or customer runs that repeat the result.