When a data centre campus sprawls across hundreds of acres and dozens of buildings, the microseconds start to matter. Relativity Networks hollow-core fibre technology is built on precisely that premise, and on 19 August 2026 the company announced it had closed a $22 million SAFE investment (a simple agreement for future equity), more than double its original target, alongside a $40 million follow-on order from a leading hyperscaler.
The funding round drew new investors including Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC. The hyperscaler contract followed a successful test of Relativity Networks’ ChronoCore™ advanced optical networking technology linking two data centres.
What Hollow-Core Fibre Actually Does
Conventional fibre-optic cable transmits light through glass. ChronoCore™ guides it through an air-filled core surrounded by a ring of glass tubes in a honeycomb pattern, which Data Center Dynamics reports also minimises chromatic dispersion and raises capacity. According to the company’s own press release, the result is light travelling approximately 47% faster than through conventional glass. The snippet and at least one secondary source cite figures of 30% and 50% respectively; the company’s issued document is the primary record.
In practical terms, CEO Jason Eichenholz puts it this way: a signal takes roughly five microseconds to travel one kilometre in conventional fibre. In hollow-core, that falls to around three and a half microseconds. A peer-reviewed study published in MDPI Applied Sciences found the latency of hollow-core fibre equals 68–69% of standard single-mode fibre, with a difference of 1.53 µs/km, broadly consistent with Relativity Networks’ figures.
For a single rack of GPUs, that gap was always easy to ignore. At the scale of modern AI infrastructure, it is not.
Relativity Networks’ Hollow-Core Fibre and the Geography Problem
Eichenholz, who holds a PhD in optical sciences and engineering and more than 90 patents in the field, previously co-founded Luminar Technologies, the LiDAR company. He co-founded Relativity Networks in 2023 with Dr. Rodrigo Amezcua Correa, who pioneered independent anti-resonant hollow-core fibre designs at CREOL, the College of Optics and Photonics at the University of Central Florida.
Their thesis is that AI infrastructure has moved through two distinct phases and is entering a third. ‘The first era of AI optimised for compute,’ Eichenholz said. ‘It was GPU, GPU, GPU. The second era optimised the networking inside the data centre to take advantage of that compute. The third era that we see coming is optimising the geography.’
That geography problem is real and growing. As power constraints force operators to distribute workloads across pre-existing facilities rather than purpose-built campuses, the physical distance between GPUs increases. Eichenholz is blunt about the driver: ‘The largest systems are distributing the compute across multiple campuses to reach the power that exists. They’re moving to where the warm shell is, but they still need to operate as one synchronised machine.’
A 47% reduction in signal travel time translates directly into the ability to span proportionally greater distances before latency degrades synchronisation. Multi-campus deployments, in which older data centres are knitted together to function as a single logical unit, are where Relativity Networks sees its clearest commercial opening.
The $22 million SAFE follows a manufacturing milestone with Prysmian, the cable group, in producing the highest-density hollow-core fibre cable the two companies have built together using ChronoCore™ technology. That pairing of a startup’s photonics design with an established cable manufacturer’s production capacity is itself a signal about where the company is in its development: the product works, and it now needs to be made at scale.
The hyperscaler contract suggests at least one major buyer agrees. Whether the $40 million order converts into a broader deployment programme will be the first real measure of whether the third era of AI geography Eichenholz describes is arriving on his timeline, or someone else’s.
