Volantis Raises $88M for AI Memory Bottlenecks
Volantis is targeting one of AI hardware’s hardest constraints: moving data fast enough between compute and memory.
AI hardware is not only limited by how many accelerators companies can buy. It is also limited by how quickly those chips can access memory.
What happened
Volantis raised $88 million to develop laser-based interconnect technology for linking AI compute units with memory chips. The round was led by Lachy Groom and Abstract Ventures, with participation from several deeptech and semiconductor-focused investors.
The company is focused on optical connections that could help AI systems move data between processors and memory with higher bandwidth and lower bottlenecks.
Why it matters
Memory bandwidth is one of the hidden constraints behind AI performance. Even powerful accelerators can sit underused if they cannot access the right data quickly enough. That makes interconnect technology a critical layer in the AI infrastructure stack.
Volantis sits in a category that may become more important as model sizes, inference loads and data-centre scale continue to grow. The funding suggests investors are looking beyond headline GPU supply and into the supporting technologies that make large AI systems efficient.
The bigger picture
The AI infrastructure race is becoming more specialised. Chips, networking, memory, optics, power and cooling are all turning into venture-backed categories. Volantis is part of that broader shift from general compute expansion toward solving specific physical bottlenecks inside AI data centres.