Volantis Raises $88 Million for Optical Chips Targeting AI Memory Bottlenecks

Volantis raised $88 million in a Series A round to develop optical technology that aims to ease AI memory bottlenecks and support faster inference for very large models.

The approach would equip chips with large amounts of high-speed memory, increasing per-chip memory bandwidth and capacity by several orders of magnitude. It is designed to support models with more than 10 trillion parameters at speeds of up to 10,000 tokens per second per user.

Volantis plans to initially use the technology to accelerate AI agents. The systems could shorten programming tasks that take hours to minutes or seconds.

The optical approach could improve memory systems by several orders of magnitude between product generations, compared with roughly twofold gains each year. Volantis has taped out its next-generation product and is targeting a bandwidth-density record at that distance.

The team has worked on CoWoS products, early high-bandwidth memory, silicon photonics co-packaged optics systems, tunable VCSELs and processors that communicate directly with light. Its current technology achieves more than 10 times the transmission distance of equivalent very thin copper wires inside a chip package.