tech
January 22, 2026
From invisibility cloaks to AI chips: Neurophos raises $110M to build tiny optical processors for inferencing
Neurophos is taking a crack at solving the AI industry's power efficiency problem with an optical chip that uses a composite material to do the math required in AI inferencing tasks.
TL;DR
- Neurophos has developed an optical processing unit (OPU) using "metasurface modulators" for AI computing.
- The technology aims to solve the problem of scaling computing power while controlling power consumption in AI labs and hyperscalers.
- Neurophos claims its OPU is significantly faster and more energy-efficient than current silicon GPUs and TPUs used for AI inferencing.
- The startup raised $110 million in a Series A funding round led by Gates Frontier, with participation from Microsoft and others.
- Neurophos's "metasurface" is reportedly "10,000 times" smaller than traditional optical transistors, enabling higher density and efficiency.
- The company claims its chip can perform calculations at 56 GHz, yielding 235 Peta Operations per Second (POPS) while consuming 675 watts, compared to Nvidia's B200 AI GPU.
- Neurophos's chips can be manufactured using standard silicon foundry materials and processes.
- The company plans to release its first chips by mid-2028 and is expanding its engineering sites in Austin, Texas, and San Francisco.