AI infrastructure is pushing data centers into a harder physical problem: where to put compute when power, politics, land, and latency all matter at once. Relativity Networks is arguing that the answer may not only be more buildings or more GPUs, but faster connections between them.
The company is focused on hollow-core fiber, a rarely deployed technology that moves data faster than conventional fiber. If it works at scale in the environments Relativity Networks is targeting, it could change how developers think about the distance between data center buildings and campuses.
A funding round tied to a latency bet
On Tuesday, Relativity Networks announced $22 million in SAFE note funding. The funding was drawn by Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC, among others.
A SAFE note is a standard method used for pre-seed and seed rounds. In this structure, an investment transfers into a specific numbers of shares once the company raises its first priced round.
Relativity Networks also secured a $40 million follow-on order from a leading hyperscaler. That customer declined to be named for the piece.
The timing matters because data center developers are expected to spend as much as $4 trillion by the end of the decade. At the same time, they are already heavily constrained by political and power-grid considerations in where they can build.
Most of the industry treats fiber speed as a fixed input. Relativity Networks is taking the opposite view: if the fiber itself gets faster, the map for AI compute can become more flexible.
How hollow-core fiber changes the signal path
Relativity Networks deals in hollow-core fiber. The source describes it as a rarely deployed technology that allows data to be transmitted 30% faster than conventional fiber.
The reason comes down to the medium that carries the light. Traditional fiber transmits light through fiber-optic glass. Hollow-core fiber transmits the same light through a vacuum chamber in the center of the line, bringing it far closer to the theoretical limit of light speed.
The improvement is measured in microseconds, not seconds. CEO Jason Eisenholz estimates that a signal takes roughly five microseconds to travel one kilometer in conventional fiber. With hollow-core fiber, that figure can fall to only three and a half microseconds.
Those small increments become more important when systems are spread across larger spaces. A few microseconds may be easy to dismiss in a compact setup, but the source makes clear that the physical scale of AI compute is changing the value of low-latency infrastructure.
Why AI campuses make distance harder to ignore
When AI compute occurred across a single rack of GPUs, fiber latency was easier to ignore. The compute was close enough that the delay from fiber was not the central constraint described in the source.
That picture has changed as scale has grown. It is now common for a data center campus to sprawl across hundreds of acres and dozens of buildings.
Eisenholz sees a particular opportunity in multi-campus deployments. In that model, pre-existing data centers are connected so they can operate as a single unit.
“The largest systems are distributing the compute across multiple campuses to reach the power that exists,” he tells TechCrunch. “The y’re moving to where the warm shell is, but they still need to operate as one synchronized machine.”
That statement points to the core operational challenge. The compute may need to move toward available power and usable data center shells, but the system still has to behave like a coordinated machine rather than a loose collection of distant buildings.
This is where hollow-core fiber becomes more than a component upgrade. If developers can reduce latency enough, they may be able to connect facilities across wider areas while preserving the synchronization those systems require.
The geographic implication of faster fiber
The source frames the result as a way to partially alleviate the spatial logic restraining many ongoing data center buildouts. Faster fiber does not remove constraints around power or politics. It does, however, change the distance calculation when latency is one of the limiting factors.
In latency terms, reducing time by 30% gives developers an opportunity to span 30% larger distances before latency becomes a problem. That is the key practical claim around Relativity Networks and hollow-core fiber.
For AI data centers, this could matter most when projects are no longer contained within a single building or even a single campus. As compute projects scale larger, the ability to link distributed infrastructure becomes a strategic question, not just a networking detail.
Eisenholz describes the industry as moving through phases of optimization. The first was centered on compute. The second focused on networking inside the data center. The third, in Relativity Networks’ view, is about geography.
“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.”
That is the larger bet behind the $22 million funding announcement and the $40 million follow-on order. If AI infrastructure keeps spreading across larger physical footprints, the fiber between compute locations may become part of the design frontier.
Relativity Networks is not presenting hollow-core fiber as a replacement for the broader work of finding power, buildings, and viable sites. The claim is narrower and more technical: faster fiber can reduce latency enough to give developers more room to connect the places where compute can actually be built.