nEye.ai, a startup based in Silicon Valley, is working on a bold idea: as AI systems grow bigger, the bottleneck won’t be the chips themselves but the data connections between them. The company’s innovation involves integrating optical switches directly into data center hardware, making data transfer faster and more efficient. Recently, they announced raising $80 million to push this technology further, bringing their total funding to $152 million.

The Challenge of Growing Data Centers

Data centers worldwide are facing a surge in energy use, especially as AI models become more complex. Projections show that global electricity consumption for data centers could more than double by 2030. The US and China are responsible for most of this increase. As AI clusters expand, the internal networks connecting thousands of processors and memory units are under immense pressure.

Traditional electrical switches, which rely on copper wires, generate heat, consume a lot of power, and cause delays. These issues make scaling up difficult and costly. To keep pace with AI demands, new solutions are needed that can handle data faster and more efficiently. That’s where optical switching comes into play, offering a way to sidestep some physical limits of current systems.

The Rise of Optical Circuit Switching

Optical circuit switching has been used mainly in telecommunications for decades. However, its application inside AI data centers is relatively new. As AI workloads grow, the need for faster, denser, and more power-efficient connections has become critical. Market research predicts that the optical circuit switching market will surpass $2.5 billion by 2029, showing strong interest from big tech companies beyond just Google.

Many companies are now exploring optical switching not just for training AI models but also for managing back-end networks and hardware pools. The idea is to create a flexible infrastructure where resources like GPUs and CPUs can be reorganized on the fly through software-controlled optical links. This approach, called composable infrastructure, allows data centers to adapt quickly to changing AI workloads without physical rewiring.

nEye.ai’s technology centers around an “OCS-on-a-chip,” a tiny component that switches data using light instead of electricity. This innovative chip is much smaller and more efficient than traditional hardware. Its development aims to unlock new levels of performance for AI training and inference at scale. The company believes optical switching is a key step toward more sustainable and faster data centers in the future.

Inspired by

• https://justainews.com/companies/funding-news/neye-ai-raises-80m-to-put-optical-switching-inside-ai-data-centers/

Sources

• iea.org