Next-Level Quantum Computers Will Almost Be Useful

The goal of the quantum-computing industry is to build a powerful, functional machine capable of solving large-scale problems in science and industry that classical computing can’t solve. We won’t get there in 2026. In fact, scientists have been working toward that goal since at least the 1980s, and it has proved difficult, to say the least.

“If someone says quantum computers are commercially useful today, I say I want to have what they’re having,” said Yuval Boger, chief commercial officer of the quantum-computing startup QuEra, on stage at the Q+AI conference in New York City in October.

Because the goal is so lofty, tracking its progress has also been difficult. To help chart a course toward truly transformative quantum technology and mark milestones along the path, the team at Microsoft Quantum has come up with a new framework.

This framework lays out three levels of quantum-computing progress. The first level includes the kinds of machines we have today: the so-called noisy, intermediate-scale quantum (NISQ) computers. These computers are made up of roughly 1,000 quantum bits, or qubits, but are noisy and error prone. The second level consists of small machines that implement one of many protocols that can robustly detect and correct qubit errors. The third and final level represents a large-scale version of those error-corrected machines, containing hundreds of thousands or even millions of qubits and capable of millions of quantum operations, with high fidelity.

If you accept this framework, 2026 is slated to be the year when customers can finally get their hands on level-two quantum computers. “We feel very excited about the year 2026, because lots of work that happened over the last so many years is coming to fruition now,” says Srinivas Prasad Sugasani, vice president of quantum at Microsoft.

Microsoft, in collaboration with the startup Atom Computing, plans to deliver an error-corrected quantum computer to the Export and Investment Fund of Denmark and the Novo Nordisk Foundation. “This machine should be utilized toward establishing a scientific advantage—not a commercial advantage yet, but that’s the path forward,” Sugasani says.

QuEra has also delivered a quantum machine ready for error correction to Japan’s National Institute of Advanced Industrial Science and Technology (AIST), and plans to make it available to global customers in 2026.

The significance of error correction

Arguably, the main trouble with today’s quantum computers is their propensity for noise. Quantum bits are inherently fragile and thus sensitive to all kinds of environmental factors, such as electric or magnetic fields, mechanical vibrations, or even cosmic rays. Some have argued that even noisy quantum machines can be useful, but almost everyone agrees that for truly transformative applications, quantum computers will need to become error resilient.

To make classical information robust against errors, one can simply repeat it. Say you want to send a 0 bit along a noisy channel. That 0 might get flipped to a 1 along the way, causing a miscommunication. But if you instead send three zeros in a row, it will still be obvious that you were trying to send a 0 even if one gets flipped.

Simple repetition does not work with qubits, because they cannot be copied and pasted. But there are still ways to encode the information contained in a single qubit onto many physical qubits, making it more resilient. These groups of physical qubits encoding one qubit’s worth of information are known as logical qubits. Once information is encoded in these logical qubits, as the computation proceeds and errors occur, error-correction algorithms can then tease apart what mistakes were made and what the original information was.

Just creating these logical qubits is not enough—it’s important to experimentally verify that encoding information in logical qubits leads to lower error rates and better computation. Back in 2023, the team at QuEra, in collaboration with researchers at Harvard, MIT, and the University of Maryland, showed that quantum operations carried out with logical qubits outperformed those done with bare physical qubits. The Microsoft and Atom Computing team managed the same feat in 2024.

This year, these scientific advances will reach customers. The machine that Microsoft and Atom Computing will be delivering, called Magne, will have 50 logical qubits, built from some 1,200 physical qubits, and should be operational by the start of 2027. QuEra’s machine at AIST has around 37 logical qubits (depending on implementation) and 260 physical qubits, Boger says.

Quantum computers made of atoms

It may be no coincidence that both of the level-two quantum computers will be built out of the same types of qubits: neutral atoms. While the classical computing world has long since settled on the transistor as the fundamental device of choice, the quantum-computing world has yet to pick the perfect qubit, be it a superconductor (pursued at IBM, Google, and others), a photon (used by the likes of PsiQuantum and Xanadu), an ion (developed by IonQ and Quantinuum, to name a few), or other.

All of these options have their advantages and disadvantages, but there is a reason some of the earliest error-corrected machines are built with neutral atoms. The physical qubits that make up a logical qubit need to be close to each other, or connected in some way, in order to share information. Unlike, say, superconducting qubits printed on a chip, any two atomic qubits can be brought right next to each other (an advantage shared by trapped ions).

“Neutral atoms can be moved around,” says QuEra’s Boger. “That allows us to build error-correction methods that are just not possible with static qubits.”

A neutral-atom quantum computer consists of a vacuum chamber. Inside the chamber, a gas of atoms is cooled to just above absolute zero. Then, individual atoms are captured, held, and even moved around by tightly focused laser beams in a technique known as optical tweezing. Each atom is a single physical qubit, and these qubits can be arranged in a 2D or even 3D array.

Neutral-atom quantum computers consist of individual atoms that are manipulated and controlled primarily by lasers. Complex optical setups guide the laser beams to their precise destinations. Atom Computing

The computation itself—the sequence of “quantum gates”—is performed by shining a separate laser at the atoms, illuminating them in a precisely orchestrated fashion. In addition to maneuverability, the neutral-atom approach offers parallelism: The same laser pulse can illuminate many pairs of atoms at once, performing the same operation on each pair simultaneously.

The main downside of neutral-atom qubits is they are a bit slow. Computations on atomic systems are about one-hundredth to one-thousandth as fast as their superconducting counterparts, says Jerry Chow, director of quantum systems at IBM Quantum.

However, Boger argues that this slowdown can be compensated for. “Because of the unique capabilities of neutral atoms, we have shown that we can create a 50x or 100x speedup over what previously was thought,” he says, referring to recent work at QuEra in collaboration with Harvard and Yale. “We think that when you compare what some people call time to solution, not just clock speed but how long it would take you to get to that useful result…that neutral atoms today are comparable to superconducting qubits.” Even though each operation is slow, more operations are done in parallel and fewer operations are needed for error correction, allowing for the speedup.

More than one way to skin Schrödinger’s cat

Microsoft’s three-level framing is not accepted by everyone in the industry.

“I think that kind of level framing…is a very physics-device-oriented view of the world, and we should be looking at it more from a computational view of the world, which is, what can you actually use these circuits for and enable?” says IBM’s Chow.

Chow argues that, although a large error-corrected machine is the ultimate goal, it doesn’t mean error correction must be implemented first. Instead, the team at IBM is focusing on finding use cases for existing machines and using other error-suppressing strategies along the way, while also working toward a fully error-corrected machine in 2029.

Whether or not you accept the framing, the teams at QuEra, Microsoft, and Atom Computing are optimistic about the neutral-atom approach’s potential to reach large-scale devices. “If there’s one word, it’s scalability. That’s the key benefit of neutral atoms,” says Justin Ging, chief product officer at Atom Computing.

Both the teams at QuEra and Atom Computing say they expect to be able to put 100,000 atoms into a single vacuum chamber within the next few years, setting a clear path toward that third level of quantum computing.

This article appears in the January 2026 print issue.