SANTA CLARA: nEye.ai has raised $80 million in a Series C round to scale its optical circuit switching technology for AI data centres. The round was led by Sutter Hill Ventures with participation from CapitalG, Alphabet’s independent growth fund, and M12, Microsoft’s venture fund. Total funding reaches $152 million.
The company builds Optical Circuit Switches, compact hardware that allows GPU, CPU, and memory resources to be pooled and redistributed flexibly across a data centre. As AI workloads grow in complexity, fixed network architectures become a bottleneck.
The bottleneck nobody talks about
This bottleneck is driving a new wave of AI-native networking systems. Where infrastructure is designed specifically for large-scale model workloads rather than traditional data traffic patterns. nEye’s technology lets infrastructure adapt in real time rather than being locked to static configurations.
The problem nEye is solving is increasingly urgent. AI data centres, now described as gigawatt AI factories, need more than raw compute. Data must move between chips fast enough to keep GPUs fully utilised.
This has led to rapid innovation in next-generation data centre networking, with new architectures focused on handling the scale and complexity of AI workloads more efficiently. Without fast, flexible interconnects, expensive hardware sits idle waiting for data.
That wasted capacity translates directly into higher operational costs and slower model performance. This bottleneck is not limited to networking. Memory efficiency is emerging as another constraint, with technologies like TurboQuant aiming to reduce the amount of data AI models need to store and process.
Why optical beats electrical
Optical switching addresses this at the physical layer. Light moves data faster and with less energy than electrical connections. nEye’s switches introduce a reconfigurable optical layer that dynamically connects resources based on real-time workload demand. As AI models grow larger and multi-GPU clusters become standard, that flexibility becomes essential infrastructure rather than optional hardware.
The broader shift is toward composable AI infrastructure. This shift is also visible at the software layer, where systems are being designed to share and retain context across workloads, allowing AI components to operate more cohesively.
Rather than building fixed clusters for specific workloads, operators want systems that can be reconfigured on the fly. nEye’s technology enables that model, letting data centres evolve their architecture as AI model demands change.
The $80 million will accelerate development and high-volume manufacturing of nEye’s proprietary switches. Investors from both Alphabet and Microsoft joining the round signals strategic interest from two of the world’s largest AI infrastructure operators.
Previous
