Relativity Networks Raises $22M to Cut Data Center Fiber Latency 30%

Relativity Networks closed a $22 million SAFE note round on August 19, 2026, with participation from Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC, among others. Alongside the funding, the company secured a $40 million follow-on order from an unnamed leading hyperscaler. The product: hollow-core fiber, a physical-layer technology that cuts signal latency by 30% compared to conventional glass-core fiber — a difference that, at data center scale, is starting to matter enormously.

Why Hollow-Core Fiber Is Different

Conventional fiber-optic cable transmits light through glass. Hollow-core fiber replaces that glass core with a vacuum chamber, moving the signal closer to the theoretical maximum speed of light in a vacuum. CEO Jason Eisenholz quantifies the gap precisely: a signal travels one kilometer in approximately five microseconds through standard fiber, versus three and a half microseconds through hollow-core. That 1.5-microsecond-per-kilometer saving is the entire value proposition — at the distances now involved in large-scale AI infrastructure, those microseconds accumulate into a meaningful architectural constraint.

The Data Center Geography Problem

When GPU workloads fit inside a single rack, inter-node fiber latency was negligible. Today, a single data center campus can sprawl across hundreds of acres and dozens of buildings. The largest deployments now stitch together multiple pre-existing campuses to aggregate the power capacity that a single greenfield site cannot supply.

Eisenholz frames this as a third era of AI infrastructure optimization. "The first era of AI optimized for compute," he said. "It was GPU, GPU, GPU. The second era optimized the networking inside the data center to take advantage of that compute. The third era that we see coming is optimizing the geography." His argument: a 30% reduction in per-kilometer latency translates directly into a 30% expansion of the viable radius within which distributed compute can still operate as a coherent, synchronized system. For operators assembling multi-campus clusters to chase available grid capacity, that radius expansion is the difference between a workable topology and one that does not close.

This kind of physical-layer shift has downstream consequences for the hardware and software decisions shaping AI at the frontier, where synchronization overhead compounds across every inter-GPU link.

Funding Structure and Commercial Traction

The $22 million raise is structured as a SAFE note, converting to equity at a specific share count once Relativity Networks completes its first priced round — a standard structure for pre-seed and seed-stage companies. The named investors are joined by additional unnamed participants.

More telling than the SAFE round is the $40 million follow-on order from a hyperscaler that declined to be identified. A purchase order of that size from a hyperscale operator represents a meaningful commercial commitment, not a pilot, and gives Relativity Networks a revenue anchor before a priced round closes.

Deployment Context

Hollow-core fiber is not a new invention — the physics have been understood for years — but it has remained a laboratory and long-haul curiosity rather than a data center staple. Relativity Networks is making a direct bet that the economics of AI infrastructure have finally crossed the threshold where a 30% latency improvement justifies the switching cost. The $40 million hyperscaler order suggests at least one major operator agrees.

AI infrastructure investment is moving down the stack, past GPUs and networking silicon, into the physical transport layer. As Nvidia's supply chain position demonstrates, control of a chokepoint in AI infrastructure can extract substantial value — and the physical medium carrying inter-GPU traffic at multi-campus scale is a chokepoint that has received almost no competitive pressure until now.