The expansion of AI data centers is being constrained by power availability and site selection, with network transmission latency emerging as a new limiting factor. Relativity Networks has announced the completion of a $22 million SAFE financing and stated that it has secured a $40 million follow-on order from a major hyperscale cloud provider, aiming to enter this market with hollow-core fiber.
Financing and orders are executed simultaneously.
This financing round involved institutions such as Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC. A SAFE is a common early-stage financing instrument for startups and typically converts into equity during a subsequent priced round.
In addition to financing, the company secured a $40 million order. The buyer is described as a leading hyperscale cloud provider, though the specific name was not disclosed in the report.
Mainly features hollow-core fiber solution
Relativity Networks develops hollow-core optical fibers. This technology is uncommon and differs from traditional fibers that transmit light signals through glass; instead, hollow-core fibers guide light through an air-filled cavity at the center, bringing transmission speeds closer to the theoretical upper limit of the speed of light.
The company states that this solution can increase data transmission speeds by approximately 30% compared to traditional optical fiber. According to their calculations, signal transmission delay over 1 kilometer in traditional optical fiber is about 5 microseconds; switching to hollow-core fiber reduces this to approximately 3.5 microseconds.
- Traditional fiber optic: approximately 5 microseconds per kilometer
- Hollow-core fiber: approximately 3.5 microseconds per kilometer
- Latency reduction: approximately 30%
Targeting AI computing power interconnection across campuses
As AI training and inference systems scale, computing power is no longer confined to a single GPU rack. Today, a data center campus may span hundreds of acres and be distributed across dozens of buildings, increasing the physical distance between GPUs.
Relativity Networks believes the main opportunity lies in multi-site deployment scenarios—connecting existing data centers to operate as a unified system. This trend arises because developers often need to distribute computing resources based on existing power conditions and available server room locations, rather than concentrating all equipment in a single location.
Company CEO Jason Eisenholz said that larger AI systems are distributing computing power across multiple campuses to leverage existing electrical resources, but these facilities still need to work together like a synchronized machine.
Following this logic, if transmission latency is reduced by approximately 30%, developers can accept a corresponding 30% increase in physical distance before reaching the same latency threshold. This means the geographic flexibility of data center placement could improve, partially alleviating site selection pressures in the expansion of AI infrastructure.