From 6 qubits to 98 in five years: Quantinuum's Helios solves in an instant what supercomputers would need trillions of years to crack

Photo not related to article. A quantum computer. [Getty Images Bank]

Some calculations would take hundreds — or even tens of trillions — of years even if every powerful supercomputer on Earth were enlisted simultaneously. A quantum computer has now completed such a calculation in an instant. It is Helios, built by US firm Quantinuum, which encodes 98 qubits — the smallest unit of quantum information — using individual atoms.

A research team led by Anthony Ransford of Quantinuum published the findings in the 2026 issue of the international journal Nature. Researchers from the Quantum Performance Laboratory at Sandia National Laboratories also participated.

Classical computers use 0s and 1s; quantum computers use atoms

A conventional computer stores information in only one of two states: 0 or 1. This smallest unit is called a "bit." No matter how complex the calculation, it processes combinations of 0s and 1s one step at a time.

Quantum computers work differently. They use a single atom as the smallest unit of information, called a "qubit." Because of the properties of quantum mechanics, a qubit can hold 0 and 1 simultaneously.

As the number of qubits grows, the number of states that can be represented simultaneously increases exponentially. With 98 qubits, the system can in principle handle all the combinations that a classical computer would have to work through one by one — all at once. This is why quantum computers can be overwhelmingly faster than supercomputers for certain types of calculations.

The challenge is maintaining accuracy while increasing the number of qubits. Atoms are extremely sensitive to external vibrations and electromagnetic fields, so error rates tend to rise as qubit counts grow.

A computer built by trapping atoms

There are several approaches to building a quantum computer. Helios does not use superconducting circuits. Instead, it employs an "ion trap" method, in which individual atoms are held in place in a vacuum using electric fields and used directly as qubits.

This approach offers the highest accuracy of any current method, but scaling up the qubit count had long been its weakness. The first computer of this type, introduced five years ago, had just six qubits. Helios raises that number to 98 — while pushing accuracy even higher.

An actual image of 98 atomic ions trapped on a two-dimensional surface in Helios. [Nature, 2026]

The key to Helios lies in a structure that moves atoms along circuits etched onto a chip. The design separates a "storage zone," where atoms wait, from a "computation zone," where operations take place, with X-shaped intersections connecting the two.

These intersections allow any atom in the storage zone to move directly to the computation zone. When 98 atoms pair up to perform calculations, any atom can be matched with any other — a property the researchers call "all-to-all connectivity."

New technology cuts errors to roughly one in 10,000

The research team also changed the type of atom used as qubits. While earlier systems relied mainly on other elements, Helios is the first programmable quantum computer to use barium atoms as qubits.

The lasers needed to manipulate barium atoms operate in the visible-light range — wavelengths the human eye can see. Lasers and optical equipment in this range are more stable and less expensive, and the error rate per operation fell as a result.

The chart shows how overall system reliability decreases as the length and depth of computation circuits grow. As circuits become more complex, the reliability score (g) drops, but the predicted curve (black line) and actual measured values (red dots) align precisely, showing the system remains under control. [Nature, 2026]

The research team found that the error rate for single-qubit operations was about 25 per million. For two-qubit operations, the error rate was somewhat higher, at roughly 8 per 10,000.

To verify this accuracy, the team ran two experiments. The first involved repeatedly executing random operations and comparing the results against predictable reference values. The second was "random circuit sampling" — running random quantum circuits that are difficult even for supercomputers to replicate.

Photo not related to article. The third quantum computer at the Barcelona Supercomputing Center. [EPA]

Classical supercomputers can no longer keep up

The results of the second experiment are the heart of the study. How long would it take a supercomputer to reproduce the outputs Helios generated?

Using the most advanced computational methods currently available, the team estimated that even a modest increase in circuit complexity would require tens of trillions of years for a supercomputer — even one equipped with optimized GPUs — to perform the same calculation. The power required would exceed the capacity of any electrical infrastructure on Earth.

The chart compares the time required (left vertical axis) and power needed (right vertical axis) for top-performing GPUs and the world's best supercomputers to classically simulate quantum operations, plotted against circuit depth (horizontal axis). At a circuit depth of 26 layers, conventional supercomputer systems (green dashed lines) would need tens of trillions of years to complete the calculation. [Nature, 2026]

Matching Helios at the same speed would demand more power and time than any existing supercomputer can provide, the research team said. This means Helios has already surpassed the capabilities of classical computers for certain calculations.

The team said Helios is already being used in active research — including quantum simulations of superconducting phenomena and cryptography work aimed at generating certified random numbers that cannot be compromised by hacking.

"Even at this stage, we are demonstrating world-leading performance at the 100-qubit scale," the team said. "We believe two-qubit gate errors can be reduced by half going forward."

The team added that scaling up the intersection architecture used in Helios "will allow us to advance toward large-scale quantum computers that maintain all-to-all connectivity across far greater numbers of qubits."

Reference

DOI: 10.1038/s41586-026-10676-4

Ransford, A., Allman, M.S., Arkinstall, J. et al. A 98-qubit trapped-ion quantum computer with all-to-all connectivity. Nature (2026).

dbsdn1110@heraldcorp.com