MagiQware Raises $656K, Cuts Quantum Circuits Up to 40%

Dutch startup MagiQware raised €575,000 (around $656,460) in pre-seed funding to build AI software that makes fault-tolerant quantum computers cheaper to run. Graduate Ventures backed the round, joined by Delft Enterprises and LUMO Labs, and the company says its software has already cut circuit length by up to 40% for the magic state factories it targets. That result goes at one of the biggest cost problems sitting between today’s error-prone qubits and quantum machines that do useful work.

What MagiQware is actually optimizing

Error correction is the tax on every serious quantum computation. Because qubits pick up errors from noise and their surroundings, future machines will lean on error-correction schemes that spend many physical qubits to build a few reliable logical ones. That spending balloons the hardware and compute needed for any useful calculation. Magic state factories are one of the priciest parts of that setup, producing the specialized quantum states a fault-tolerant machine needs to run most operations.

MagiQware builds software that sits in the quantum compiler and software stack rather than the hardware. It uses reinforcement learning, where an algorithm improves through repeated trial and feedback, to tune how those factories run. The company reports up to a 40% cut in circuit length for the factories it targets, and shorter circuits mean fewer chances for errors to pile up mid-computation. The pitch to full-stack quantum firms is that they can buy this optimization layer as opposed to building it in-house.

Why a small round is worth a look

All things considered, $656,460 is a small check, even for pre-seed. The reason it’s worth noticing is the layer MagiQware picked. Big hardware players would rather not staff magic state optimization themselves, and a specialist that shaves measurable overhead off error correction has an obvious buyer. The Netherlands has a strong quantum bench to recruit from, and this team spans quantum computing, AI, software engineering, and physics, under CEO Arash Ahmadi and CTO Shakeeb Majid, with Sahar Hejazi on device and Ali Moghaddam on theory.

The caution is the usual one for numbers this early. The 40% figure comes from the company’s own results in the LinkedIn announcement, without an independent benchmark or peer review attached. The round is tiny, the product is pre-commercial, and magic state factory optimization is a research-heavy field with plenty of competing academic work. Even so, the bottleneck is genuine, and a software-only fix that customers can drop into an existing stack is the kind of thing that scales if the results hold. That last part is what the next year has to prove.