The rapidly evolving quantum computing industry has quickly spun through two critical stages and now has solidly moved into a third one, according to Yuval Boger, chief commercial officer at QuEra. The first stage was showing that a quantum system – even one with as few as five qubits – could be built.

Once that was proven, the next question facing system makers was whether they could correct errors, a crucial test because a quantum system that produces a lot of errors isn’t useful. As we’ve written, there have been significant advances in error correction, such that companies ranging from Microsoft and Google to QuEra and others now say they can.

“The third and exciting phase is, 'Can you scale it up? Can you get to a large enough number of qubits that can run useful problems?’” Boger told The Next Platform. “For an end customer, the question has always been, ‘How far is it away from something really useful? What's the gap between where I am and where I need to be?’ That gap has been very, very rapidly narrowed in terms of the time to true use.”

Quantum systems continue to evolve with more logical and physical qubits and roadmaps being laid out that stretch into the next decade and preview the coming era of fault-tolerant, scalable commercial computers.

QuEra this week is making its next step public, announcing a new quantum system – Libra – that will be hosted on the Braket service at Amazon Web Services, a fully managed cloud service that serves as a unified platform that developers and researchers can use to build and run quantum algorithms across multiple modalities, from superconducting and trapped ions to neutral atoms. Libra, a neutral-atom system, will be QuEra’s first fault-tolerant quantum computer and will arrive on Braket in 2028.

It follows Aquila, a quantum computer that QuEra launched on AWS in November 2022. Aquila is a 256-physical-qubit system housed in Boston and based on programmable arrays of neutral Rubidium atoms. It’s available on Braket for 130 hours a week.

Libra is the first system in QuEra’s newly expanded partnership with AWS. It’s a megaquop-class system, one that can perform a million reliable logical quantum operations, an important metric that illustrates not only the importance of the number of logical qubits, but also how many operations can be run before errors overwhelm the computation, according to QuEra.

Logical qubits are created by grouping together multiple physical qubits to detect and correct errors. Libra will have more than 256 logical qubits built from 10,000 to 15,000 physical qubits that will be created through QuEra’s neutral-atom process. It also will have a logical error rate of 10⁻⁶, which means that for every million quantum operations performed on a logical qubit, the system will produce an error, on average, only once. This will allow the system to support early practical commercial and research jobs.

The number of physical qubits will be a significant jump from the hundreds now being used by quantum companies, but the industry’s focus on those qubits misses the point, Bogar said.

“The real metric is how many good qubits you have – error-corrected qubits – and how good they really are,” he said. “The number of qubits obviously talks about the breadth of the application that you can make, so more qubits, more variables. The error rate tells you about the depth of the circuit. How long a calculation you can run. In physical qubits, the state-of-the-art is that with every thousand operations, 999 operations get to the correct result and one gives you the wrong result. [That] doesn't seem like a lot, but if you want to run a serious algorithm that has half a million steps, then having an error every thousand steps obviously kills any usability. You're always guaranteed to get junk. Now that we have a computer that can do one in a million – so 999,999 times out of a million will be good – that gives you the ability to run longer, more sophisticated, more serious algorithms.”

Libra also will deliver the economics of neutral-atom computing, Bogar said. The atoms are held in place by lasers, so QuEra’s systems can run at room temperature – other modalities need expensive and power-hungry cryogenic or other cooling systems – and can run in standard 19-inch racks, a change from some of the other chandelier-shaped quantum computers.

However, Bogar said other modalities – like superconducting, trapped ions, and photonics – need to be explored, adding that some workloads may run best on other architectures. Google, which has been focusing on superconducting, in March said it is expanding its quantum work to include neutral atoms.

Libra, which QuEra believes will be particularly useful in simulation calculations, such as those in the chemical and material industries, will be a product of innovations being made in the industry in recent years that are accelerating the run to commercial, fault-tolerant quantum computing, with Bogar pointing to three key advancements that are making that happen.

“The first reason is that quantum computers have gotten better,” he said. “You get more qubits, better qubits, so you're closer to where you need to be. The second thing is, there's been this major, major improvement in quantum error correction. Specifically, the concept behind quantum error correction is that you use multiple physical qubits to create an error-protected logic computer. Only two years ago, the assumption was that you were going to use some code that is called surface code. People were saying that the ratio between physical and logical maybe would be a thousand to one, so if you need a thousand error corrected qubits, then maybe you need 1 million physical qubits.”

That said, recent research is showing that the number of physical qubits needed to build a reliable logical qubit could be much fewer, in the hundreds or thousands. A paper issued by QuEra, Harvard, and MIT in April showed it could take as few as two physical qubits, which Bogar said is “one big overhead improvement.”

The third advancement is that the number of qubits needed for an algorithm also is falling. Initially, Shor’s Algorithm determined that a million qubits are needed.

“Then someone said, ‘I found a trick to do only half a million.’ And then another one came: ‘I need only a hundred thousand,’” he said. “By now it's like I only need 20 or 30,000 qubits. So all of a sudden, computers are getting better. Algorithms require less error correction, which makes it more efficient, so this is converging to the point where we think truly useful applications are maybe only two years away.”

QuEra’s announcement of Libra comes days after Alice & Bob unveiled the Helium Quantum System, which will include the European company’s first logical qubit using as few as 18 physical cat qubits, which have fault-tolerance built into them by encoding information into the quantum states of light rather than typical discrete particles.

The Helium Quantum System is a step up for the vendor, shifting from developing cat-qubit chips toward delivering a complete quantum system – with cabling, control electronics, and a software stack – that can be deployed on premises and is optimized for error correction.

Alice & Bob’s roadmap includes a quantum system that can support an upcoming 48 cat-qubit chips that will include multiple logical qubits. The Helium Quantum System needs about 40 kilowatts of power to run, which the company said will lower the cost of deploying the technology.