Volantis raises $88 million for optical-memory AI inference hardware

Volantis raises $88 million for optical-memory AI inference hardware

San Francisco-based Volantis raised an $88 million Series A co-led by Lachy Groom and Abstract Ventures, with participation from John Doerr, VXI Capital, Triatomic and Susa Ventures. Founded in 2022 by Tapa Ghosh, Roy Meade and Inderjit Singh, the company is developing A-1, an inference system designed to increase memory capacity and bandwidth together. Volantis targets models exceeding 20 trillion parameters and speeds of up to 10,000 tokens per second per user. Those figures are development targets; the article provides no measured results. The financing will support engineering expansion, A-1 development and preparations for first customer deployments in 2027.

The investment adds an optical-memory architecture to the AI infrastructure competition over inference speed and cost. Volantis argues that storing larger models and their context requires more memory, while generating responses quickly requires enough bandwidth to move data between memory and compute. Its proposed use of optics addresses both constraints, placing the company among hardware challengers seeking to change inference economics. Ghosh identifies coding agents as an initial application, where faster inference could shorten task completion times. That potential depends on delivering the stated performance and lower cost per token.

For builders and investors, the concrete checkpoint is the planned 2027 customer deployment. Evaluation should distinguish the announced parameter capacity and token-speed targets from demonstrated operation, then examine whether the integrated system delivers lower inference costs on customer workloads. The round funds development toward that checkpoint; it does not establish commercial performance.

#Volantis #AIInfrastructure #AIInference #Photonics #Semiconductors