Volantis, which aims to use vertical-cavity surface-emitting lasers, like those used by the iPhone's Face ID, to connect AI chips and memory chips, raised $88M

AI hardware cost, previously a barrier to running large models, sees a potential reduction as new architectures emerge.

Volantis, a San Francisco-based semiconductor company, has secured $88 million in Series A funding. The round was co-led by Lachy Groom and Abstract Ventures, with contributions from notable investors including John Doerr, VXI Capital, Triatomic, Susa Ventures, Dwarkesh Patel, Naveen Rao, and Sholto Douglas. The company plans to utilize these funds to advance the development and commercialization of its AI inference system, codenamed A-1, and its photonic memory architecture. Additionally, the capital will be used to expand its engineering team. Volantis's innovative architecture merges photonics and semiconductor design to enhance memory capacity and bandwidth, aiming to enable larger AI models to operate at significantly faster speeds. Their system employs a photonic interconnect using custom micro-VCSELs to link compute and memory chips, targeting models exceeding 20 trillion parameters and achieving processing speeds of up to 10,000 tokens per second per user.

This funding round highlights a significant investment in photonic interconnects for AI hardware, signaling a push towards overcoming current memory bandwidth limitations that hinder the performance of massive AI models. Volantis's approach, utilizing VCSELs akin to those in consumer electronics, suggests a potential pathway to more integrated and efficient AI chip designs, moving beyond traditional electronic interconnects.

The successful financing of Volantis, with its ambitious goal of enabling trillion-parameter models to run at high speeds, indicates a strong market appetite for solutions that can dramatically lower the cost and increase the speed of AI inference. This development could accelerate the deployment of more sophisticated AI applications, impacting cloud providers, AI model developers, and the end-users who will benefit from faster and more powerful AI services.