Relativity Networks Raises $22M to Deploy Hollow-Core Fiber for AI Clusters
By replacing glass cores with vacuum chambers, the startup reduces data center latency by 30%, allowing hyperscalers to expand the physical footprint of distributed compute.
Relativity Networks has secured $22 million in SAFE note funding to accelerate the deployment of hollow-core fiber (HCF) across AI data center campuses. The funding arrives alongside a $40 million follow-on order from a leading, unnamed hyperscaler, signaling strong industry demand for lower-latency interconnects.
The investment round included contributions from Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC. Founded by Jason Eichenholz and Rodrigo Amezcua Correa, the company builds on research from the University of Central Florida's College of Optics and Photonics (CREOL). The core technology replaces the traditional glass core of fiber-optic cables with a vacuum or air-filled chamber, allowing light to travel closer to its maximum speed in a vacuum.
The Latency Bottleneck
As AI compute scales, the physical distance between GPUs across sprawling data center campuses has become a critical performance bottleneck. Conventional fiber-optic cables transmit light through glass, which inherently slows the signal. According to company data, signal travel time in conventional fiber is roughly five microseconds per kilometer; hollow-core fiber reduces this to approximately three and a half microseconds.
This shift allows data to be transmitted roughly 30% faster than conventional glass fiber. By removing the glass medium, Relativity Networks aims to solve the physical constraints that currently limit how far apart AI compute clusters can be placed while still maintaining the synchronization required for massive model training.
Optimizing Geography
For hyperscalers, a 30% reduction in latency translates directly into geographic flexibility. It allows developers to expand the physical footprint of their AI clusters by 30% before hitting the latency walls that degrade performance. This is particularly vital as developers face constraints regarding power grid availability and political factors, forcing them to utilize "warm shells"—existing buildings that may be further apart but can now be linked as a single unit.
"The first era of AI optimized for compute... The second era optimized the networking inside the data center... The third era that we see coming is optimizing the geography," said Jason Eichenholz, CEO of Relativity Networks. He noted that while companies are moving to where available building shells exist, they "still need to operate as one synchronized machine."
Future Outlook
The immediate focus for Relativity Networks is the fulfillment of its $40 million order and the scaling of HCF production to meet the needs of the AI infrastructure boom. While the technology provides a clear path to expanding cluster footprints, the industry will be watching to see how these deployments perform at scale across multi-campus environments and whether other hyperscalers follow suit in replacing traditional glass fiber for their most latency-sensitive AI workloads.