Relativity Networks raises $22 million to bring a faster kind of fiber to data centers

Everyone assumes fiber is fiber — glass, light, done. Relativity Networks is betting that assumption is about to cost the industry real money, given that data center developers are on track to spend as much as $4 trillion by decade's end while fighting over power and political constraints on where they can even build. If the speed of the connection between buildings can stretch, so can the map.

On Tuesday the company announced $22 million in SAFE note funding, drawn by Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC, among other backers. That's a pre-seed and seed-stage instrument, not a priced round, but the company also picked up something harder to fake: a $40 million follow-on order from a hyperscaler that asked not to be named. Someone with real infrastructure budget is already buying.

The product is hollow-core fiber, a technology that's barely been deployed at scale despite existing for a while. Instead of pushing light through solid glass, it routes the same signal through a vacuum channel running down the center of the line, which gets it much closer to light's actual speed limit. CEO Jason Eisenholz puts numbers on it: a signal takes roughly five microseconds to cross one kilometer of conventional fiber, and only about three and a half microseconds through hollow-core. That's a 30% cut in transit time, and it doesn't sound like much until you realize how much distance now separates the machines doing the work.

Compute used to live on a single rack, where a few feet of glass fiber barely mattered. It doesn't anymore. Data center campuses now sprawl across hundreds of acres and multiple buildings, and Eisenholz argues the real opportunity is stitching separate campuses together so they behave like one machine. Operators, he says, are chasing power availability first and then trying to make the resulting patchwork of buildings act in sync — which is exactly the kind of problem that gets easier when your fiber buys you 30% more distance before latency becomes a limiting factor.

Eisenholz frames this as the industry's third act. First came the raw compute race — GPUs stacked on GPUs. Then came networking optimized to actually feed those GPUs efficiently. What he's pitching now is optimization of geography itself, treating the physical layout of a build as something engineerable rather than fixed. Whether that framing holds up, a $40 million order from a hyperscaler suggests at least one buyer isn't waiting to find out.