Volantis raises $88 million for tech to connect AI, memory chips

San Francisco-based semiconductor startup Volantis has successfully raised $88 million in venture capital funding to tackle a significant challenge faced by AI chips. The primary obstacle is the limited communication between a chip's computing parts and the memory chips that store AI models and data. Nvidia and AMD have addressed this issue by surrounding their computing chips with high-bandwidth memory chips, but even their top offerings can only accommodate eight memory chips per GPU due to the restricted reach of electrical wires.

Volantis is pioneering a novel approach by utilizing vertical-cavity surface-emitting lasers (VCSELs) to transmit data between computing chips and memory chips via beams of laser light. This method circumvents the reach limitation and enables the company to integrate 220 memory chips around a GPU.

The company plans to capitalize on existing components, such as VCSELs, which are already employed in facial recognition systems in numerous devices, including Apple products. Volantis aims to circumvent supply chain bottlenecks and deliver an AI chip next year that significantly enhances tasks like AI coding, according to Tapa Ghosh, CEO and co-founder of Volantis.

The funding round was spearheaded by Stripe veteran Lachy Groom and Abstract Ventures, with participation from John Doerr, who backed Google and Amazon.com in their early stages, along with contributions from VXI Capital, Triatomic, and Susa Ventures. Angel investors including AI podcaster Dwarkesh Patel, AI chip veteran Naveen Rao, and Anthropic researcher Sholto Douglas also contributed to the round.