The quantum computing industry is advancing, but is it advancing fast enough? While we constantly hear claims like a quantum computer could be 20,000 times faster, the actual physical processors — the brains of these machines — have been stuck at a size of around 100 qubits. Google recently hit 105 with its Willow chip, and IBM is at 120 with its Nighthawk model. But going further appears to be difficult because physically cramming more onto a wafer is no easy feat, mostly due to wiring issues.

Well, a Dutch startup called QuantWare has decided to blow those numbers out of the water. They announced a new processor architecture called VIO-40K that apparently packs 10,000 qubits onto a single chip. That’s notable considering it’s effectively 100 times denser than the best right now. So how did a name you’ve likely never heard of manage to do what Google and IBM have not? The secret is virtually identical to how city planners handle overcrowding. Instead of building out, they built up. Most quantum processors use 2D, horizontal wiring, which limits how many connections you can squeeze onto a flat surface before things start getting messy. QuantWare’s solution is moving to a 3D architecture with vertical wiring.

This might not sound as exciting as the breakthrough that made teleportation a reality, but it’s significant if the promise holds up. The company used vertical connections, which allowed them to utilize “chiplets” – smaller, modular chips stitched together — to create one massive system. They basically managed to work around the usual data bottlenecks. The result is a QPU that supports 40,000 input-output lines. To actually build these, QuantWare is constructing a massive facility in the Netherlands. They claim it will be the first factory dedicated to creating open-architecture quantum devices at an industrial scale.

Naturally, you have to ask if a 10,000-qubit processor is actually “better” or if it is just bigger. Google and IBM appear to be taking a different approach right now. Google’s whole deal appears to be focusing obsessively on error correction with its Willow chip. They want to prove that they can make qubits that last longer and make fewer mistakes. IBM is doing something similar, aiming for reliability and fault tolerance by 2029. Meanwhile, QuantWare appears to be taking a brute-force approach to scaling. At the same time, they also have a very clever business strategy that sets them apart.

See, Google and IBM try to build everything from the bottom up — the chips, the software, the whole stack. But as Live Science points out, QuantWare apparently wants to be the Intel of this industry. They do not want to build the computer. Rather, they want to sell the chips to whoever’s building one. That is why they are pushing the “Quantum Open Architecture.” To that end, they have already ensured this works with Nvidia’s ecosystem. Essentially, the VIO-40K is designed to plug directly into Nvidia’s NVQLink and CUDA-Q platforms. Hypothetically, this means developers could run hybrid systems, where a classical supercomputer would offload specific tasks it can’t handle to QuantWare’s quantum chip seamlessly.

This is a smart play because it makes the hardware more accessible. Currently, if you want a quantum compute, you have to rent time from IBM or Google. Meanwhile, QuantWare says it will sell its chips directly to customers starting in 2028. That’s big if true, considering IBM’s own roadmap has its Starling computer running only by 2029. But with Big Tech’s vast resources, it will be interesting to see if QuantWare can actually hold any leads.