Advancements in quantum computing continue to pile up in areas from error correction and infrastructure to software and algorithms, sharpening the view of what the once-theoretical compute paradigm will look like when it reaches its full fault-tolerant, useful, and practical potential. Architecture will play a central role in all of this, and as we have seen in past months, the focus on this aspect of quantum computing is accelerating.
Recently, that’s included the work that D-Wave scientists – they of the annealing quantum systems – are doing with the vendor’s dual-rail superconducting architecture, as well as QuiX Quantum’s introduction of the Dedalo architecture for its fault-tolerant photonics-based systems, taking a significant step forward with its Carina commercial system.
The effort being put into developing the architecture around quantum computing shouldn’t be a surprise, says Nadav Katz, co-founder and chief technology officer for Qarakal Quantum, a startup founded in 2024 in Israel. The architecture work signals a growing maturity in the quantum space, “a realization that quantum computers are computers, not a pile of qubits that you make and then hope that by some sort of magic programming, it will converge to a full computational system,” Katz tells The Next Platform.
Architecture will be key in shifting quantum computing from institutional and vendor labs and cloud environments into the commercial space, creating modular, scalable, and efficient systems that can fit in datacenters alongside classical enterprise and supercomputers. Focusing on architecture and hardware also touches on a deeper question of how information is handled and how the industry will need to think about how to program a real quantum system.
“The current boogeyman of quantum computing is this sort of monolithic architecture in which you just say, 'Let's just have this huge sort of planar thing that grows maybe by chiplets being connected, but ultimately, it's just a huge array of qubits that you just program after the fact to do something,” he says. “The way we see it, it’s analogous to how classical computing was derived. Initially, if you go way back to Turing, then it's a pile of bits – not qubits – that you have, just very inefficient programming going around doing stuff on that pile of bits. That's a good theoretical framework, but it's not a real system. What really worked was the Von Neumann architecture, which basically is – with some very interesting recent modifications – what we use to this day, eighty years later.”
The comparison of classical computing’s evolution with what’s needed for quantum computing to move forward snakes it way throughout Qarakal’s vision. Katz notes the Von Neumann architecture is based on specialization and communication in systems. It’s the vision that Qarakal has for its superconducting quantum systems, and a foundational driver of the Pangaea architecture the vendor recently unveiled. Panagea itself is part of a larger push by the company toward specialization that allows heterogeneous code and functionality, and a step forward in that is the introduction of the quantum bus, a communication bridge that helps drive the development of modular and scalable quantum computer.
In the classical computing world, modularity and scalability came after the components of the systems – processors, memory, storage, and various peripherals – were able to work together in a unified manner. That came about with standardized communication interfaces, such as motherboards, backplanes, and buses, Qarakal executives argue. Quantum architecture is now at the point where similar advances are needed.
The Qarakal quantum bus is a foundational element of Pangaea, paving the way for more modular and scalable systems. The idea of a quantum bus isn’t new, but the flexibility in the Qarakal’s design – the ability to reach fault-tolerant logical operations between logical qubits with different codes and modules of various types, and to remove the need for logical qubits to be physically adjacent – is what drives scalability beyond simply adding more qubits.
Current quantum systems are complex monoliths that demand local connectivity and physical adjacency between logical qubits, according to Qarakal executives, who argue that such a design makes quantum computing costly and difficult to scale. Qubits are replicated across a single layout, which requires every qubit to perform every function, which tightly ties computation with communication and error correction and means components can’t scale individually.
The quantum bus Qarakal has created plays the same role as a bus in a classical system. It is a universal interface for quantum information and can handle qubits that won’t need to be adjacent. It can adapt to the error-correction method of each qubit without them having to share the same code. Sheir Yarkoni, Qarakal’s director of quantum software, says that with superconducting qubits, the need is to find the best error correcting code and use it to build the quantum system. Some codes have a high encoding rate and use few physical qubits per logical qubit, but come with costs like physical connectivity and a lot of operations. Others codes, such as planar codes, are simpler to operate and have more scalable logic. However, they are two-dimensional and require more physical qubits for each logical one.
“What we're trying to bring to the table through the invention of the quantum bus is to alleviate this necessity of choosing between different codes,” Sheir tells The Next Platform. “If you have a unit that you want to build that has a high density of memory and slower operations – like what we would think of as a hard disk in a classical computer – and you have another code that's very lightweight and has fast operations, but it doesn't have a lot of memory – kind of the equivalent of RAM in classical computing – then we don't want to build a computer where you have to choose between those.”
As with traditional machines, the goal is to “build a [unified] system that is greater than the sum of its parts. That's what the quantum bus allows you to do,” he says. “When we say architecture, when we say quantum bus – modularity, composability – when we use these words, we are truly viewing them through the lens of the same computing principles that brought the Von Neumann breakthrough to classical computing and morphing them to adapt and enable the same breakthroughs in quantum computing.”
The quantum bus has an immediate tangible benefit: the Pangaea architecture requires a 10th of the physical qubits that other approaches need, which Katz says translates into significantly less infrastructure, leading to not only fewer qubits, but also less wiring, lower energy consumption, and less noise.
The quantum bus provides a slower accumulation of errors, which means the logical qubits needed to store the logical information in the quantum system can be smaller, he says.
“In the conventional architectures – the monolithic architecture which we're comparing to and over which we get this factor of 10 – a lot of the [qubit] effort is just shuttling things around,” Katz says. “That creates an overhead and reduces the threshold and makes it harder to achieve fault tolerance. Here you get more logical operations out of the qubits that are there, so it's just that much more efficient and converges to fault tolerance faster.”
Sheir adds that due to a “much slower accumulation of errors, we can expect that the size of the logical qubits that are needed in order to store the logical information in the quantum computer can be much smaller.”
What does accumulate are the benefits, he says.
The introduction of the Pangaea architecture and quantum bus is part of a three-step roadmap that Qarakal has laid out, with the first focus this year being on integrating components and capabilities (below). System deployments are targeted for 2028 and expanding the systems will come two years later.