Neurophos CIM promises super high performance at a fraction of the power

AI processing at the speed of light.

February 9, 2026

Jon Peddie

Neurophos, a 2020 Duke University spinout, is developing metamaterial‑based optical AI chips aimed at high‑performance, low‑power inference. Its optical processing units perform in‑memory matrix multiplications using tunable metasurface modulators, enabling claimed 50–56 GHz operation and analog‑like scaling without RC delay limits. Neurophos targets 0.47 EFLOPS at FP4/INT4 with 1–2 kW power, 768 GB HBM, and 20 TB/s bandwidth, vastly exceeding GPU compute and efficiency targets on paper. The company has not yet shipped product but raised $110 million and plans first systems around 2028. Neurophos is not directly comparable to Nvidia’s NPU in terms of design and capabilities. However, Neurophos’

...

Enjoy full access with a TechWatch subscription!

TechWatch is the front line of JPR information gathering service, comprising current stories of interest to the graphics industry spanning the core areas of graphics hardware and software, workstations, gaming, and design.

A subscription to TechWatch includes 4 hours of consulting time to be used over the course of the subscription.

Already a subscriber? Login below

Neurophos CIM promises super high performance at a fraction of the power

Enjoy full access with a TechWatch subscription!

This content is restricted

Login