News brief · 12 November 2026

AMD and PsiQuantum widen partnership beyond FPGAs, with an FPGA decoder preprint

According to SDxCentral, AMD plans to invest in PsiQuantum and the pair will jointly design quantum computing infrastructure, building on Versal FPGA work and a machine-learning error decoder.

According to Xenospectrum, PsiQuantum and AMD announced an expanded partnership on 12 November 2026. Xenospectrum says it grew out of the Versal FPGAs that PsiQuantum selected when developing its fault-tolerant platform. According to SDxCentral, the companies say it now also covers Epyc CPUs, Instinct GPUs and the ROCm software stack, alongside PsiQuantum’s silicon photonics. SDxCentral reports that AMD plans to invest in PsiQuantum, and that the two firms intend to jointly design the supporting infrastructure for fault-tolerant quantum machines.

According to SDxCentral, one strand of the collaboration is PRISM, a machine-learning decoder for photonic quantum error correction. Xenospectrum describes it as transformer-based. A 15-author arXiv preprint reports logical error rates comparable to integer-programming decoders, in less time, on surface, colour and quantum low-density parity-check codes. It also explores an implementation on an AMD VP1902 FPGA. Xenospectrum reports 10.11 microseconds for a distance-8 block in a 250 MHz build of the full network, against a 10 microsecond target.

Why it matters

According to Xenospectrum, the paper is a non-peer-reviewed preprint. Xenospectrum also notes the error-rate comparison covers the unquantised model, not the quantised version on the FPGA, so check which figures apply to the hardware.