Quantum Art’s new QPU could be both significantly smaller and also faster than competing quantum architectures.
AFP via Getty Images
How can we reinvent quantum computing? Perhaps by shrinking it down and making it small: really small. And, surprisingly for a field that is all about doing in milliseconds calculations that could take decades, making it faster. Israeli quantum computer startup Quantum Art is promising both: a tiny but incredibly powerful quantum computing unit, and 100 times more simultaneous parallel processes than competing quantum architectures.
While there’s a huge amount of innovation in AI and LLMs right now, there’s also huge progress in quantum computing. Alice & Bob recently announced long-lived qubits, IBM shared a roadmap to a quantum computer with 20,000 times the processing power of today’s quantum machines, and Microsoft recently announced a breakthrough in quantum processing using topological superconductors, essentially a new state of matter. Israel is a hotbed of quantum innovation, with $650 million recently raised among nine quantum startups including Quantum Machines, Classiq and Quantum Art.
Based on my recent conversation with Quantum Art CEO Tal David, the company might be a David to the Goliaths in the industry with a real shot at significant leadership in size, speed, and capacity of quantum computers.
On the size side, two inches by two inches is small. Think a GoPro, or a small orange, or a charging brick for your laptop. But according to Quantum Art, it’s also plenty of space to fit one million physical qubits in an incredibly dense quantum computing unit, or QPU. (That number of qubits, and the compact space to fit them all in, is reminiscent of what New York-based quantum computing company SEEQC is working on.)
Add Quantum Art’s focus on multi-qubit gates which can run hundreds of operations in a single step and dynamic reconfigurability that speeds up processing, and you don’t just get a seriously compact, powerful QPU. You just might also get massive quantum advantage, which Quantum Art promises by 2027.
The dynamic reconfigurability alone is a game-changer for speeding up quantum calculations, CEO Tal David says. For starters, it offers a staggering 50X speed advantage in compiling.
“So what usually people do when they have separate ion traps, they physically move or shuttle qubits from one trap to another in order to connect them and do logical operations between them,” he told me in a recent TechFirst podcast. “But this takes a whole lot of time. It could be 98% of the compiling time spent on moving and recooling these ions instead of doing calculations.”
Ion traps are physical spaces where qubits are held and manipulated. To run an algorithm across multiple traps in traditional quantum computers, you need to physically shuttle ions between them or entangle them remotely. This is slow, error-prone, and energy-intensive.
But it’s not only a huge waste of time, it’s also a huge waste of space. Each ion trap often has its own laser, electrode, and control system.
“It increases the QPU size by a lot because the length scale that determines the size of QPU now is not the length scale of how qubits can be brought close together, which is a few micrometers, it’s rather the size of the traps, which could be even a thousand-fold larger,” says David.
It’s slightly ironic, of course, that we’re talking about making quantum computing small. The core components of quantum computers, the part that makes them quantum, are of course incredibly tiny: qubits. While there are multiple species of qubits from superconducting circuits or photonic qubits or topological qubits or trapped ion qubits, they are all minuscule. As in, less than microscopic: micrometer scale, millionths of a meter. The average human hair is 10-20 times wider than a trapped-ion qubit, for example.
But all the complex machinery around qubits to create, sustain, manipulate, and read them can be massive, something like our original room-scale mainframe computers from the 1950s and ’60s. For example, IBM’s Quantum System Two is about 22 feet square by 12 feet high. Other quantum computers are the size of a refrigerator or bigger. A single qubit may be tiny, but the full system to trap and manipulate those ions, like vacuum chambers, cryogenic shielding, and optical tables, can take up a large fraction of a room.
In comparison, Quantum Art plans to make their entire million-qubit quantum computers in just four or five typical 19″ server racks. Cryo, lasers, electronics … the entire quantum computer, fitting in just like a few typical everyday machines in a server farm, CEO Tal David says.
Two other ways Quantum Art is making quantum computing faster are multi-qubit gates and optical segmentation of ion chains. Multi-qubit gates allow the company to perform the equivalent of ~1,000 two-qubit operations in a single step using specially engineered laser pulses. This drastically reduces the number of sequential operations needed and enables more parallel computation within each core. A side benefit: lower error accumulation.
“Instead of playing my guitar note by note, we play chords,” explains David. “So this saves a whole lot of time because we’re doing many, many operations at once and it saves us a lot of errors … it’s kind of difficult to design, but we’ve found a way of how to do that and implement that, and this is kind of special for us.”
Optical segmentation of ion chains divides a single ion chain into multiple independent cores using laser-defined barriers. Think optical tweezers, if you will, to segment and control the cores. This enables many operations to happen simultaneously, across multiple cores, and avoids bottlenecks that occur when only a single register is active. The result is higher compute density and better parallel execution.
The net result is that Quantum Art can deliver quantum computing systems with 100x more parallel operations and 100x more gates per second than competitors, the company says, all in a footprint up to 50x smaller.
That’s impressive.
While Quantum Art’s timeline for quantum advantage is 2027, just two years out, the company’s roadmap for a million-qubit QPU is a little longer: 2033. By 2029, the company projects having 12,000 physical qubits forming 500 logical qubits, and by 2031, it should have 40,000 qubits making up 1,000 logical qubits. The million-qubit monster QPU Quantum Art projects for 2033 will have 10,000 logical qubits, which, if realized, would be an absolute game changer.
The company’s timeline is aligned with what I’ve heard from other quantum computing CEOs, who predict usable commercial systems starting in late 2026 or 2027.
But, of course, timelines are just timelines. Execution is hard, and reality often breaks our projections about the future.
CEO Tal David thinks his projections are pretty safe, however.
“We’ve spent our time now de-risking all of the conceptual risks of our special building blocks of the architecture and going forward,” he told me. “Now we need to bring all of these building blocks together and go to scale and the engineering challenges, which is… you know, there are challenges there, but they’re not conceptual as much as they are engineering.”
Which means, he feels the company has already solved the hard theoretical and architectural problems, and what remains is execution at scale: still difficult, perhaps, but not as uncertain as the core theoretical operations.
Perhaps he’s right. It will certainly be very interesting to see.