Hollow-core fiber could let data centers spread farther apart without slowing down

Key points

• Relativity Networks closed $22 million in SAFE note funding on Tuesday, with investors including Rhapsody Venture Partners, Bell Ventures Inc., and Faster Than Glass LLC.
• A major cloud operator, unnamed, placed a separate $40 million follow-on order for the company's product.
• Hollow-core fiber moves light through a near-vacuum rather than glass, delivering a 30% speed advantage over conventional cable.
• AI training now routinely spreads thousands of GPUs across multiple buildings, making inter-campus fiber speed a real bottleneck.
• A 30% latency reduction translates directly into a 30% wider physical footprint before chip-to-chip delay becomes a hard barrier.

Data center developers are on track to spend as much as $4 trillion by 2030, yet land constraints and power-grid limits already shape where those buildings can land. That pressure is the subject we explored in August, when a US lawmaker pushed for communities to have the right to reject new facilities over electricity and water concerns. Relativity Networks is working a different angle: if the cable itself were faster, campuses could sit farther apart without losing performance.

Tuesday's announcement put $22 million in the company's account via SAFE notes, a financing structure in which the investment converts into shares once the company completes a priced funding round. The round drew participation from Rhapsody Venture Partners, Bell Ventures Inc., Faster Than Glass LLC, and others. On top of that, a large cloud operator that declined to be identified placed a $40 million follow-on product order.

What is hollow-core fiber, and why does it matter?

Hollow-core fiber routes light through a vacuum channel at the cable's centre rather than through solid glass. That change matters because light moves faster through open air than through glass.

Conventional fiber-optic cables, which already carry most of the world's internet traffic, push light through a glass strand. CEO Jason Eisenholz puts a number on the difference: a signal takes roughly five microseconds, five millionths of a second, to travel one kilometre in standard cable. Hollow-core brings that down to about three and a half microseconds, a 30% reduction.

Why would a fraction of a second matter?

For ordinary browsing, it probably wouldn't. AI training is different.

A decade ago a single rack of GPUs, the specialised chips that handle AI's heavy computation, could run an entire job, with all the chips sitting close together. Now the largest models require spreading thousands of GPUs across multiple buildings or separate campuses linked by fiber. Every added microsecond of delay between chips risks throwing those systems out of sync.

Eisenholz told TechCrunch: "They're moving to where the warm shell is, but they still need to operate as one synchronized machine." A 30% speed gain lets engineers place campuses 30% farther apart before latency becomes a hard constraint.

His read on where the industry is heading came through plainly: "The first era of AI optimized for compute. It was GPU, GPU, GPU. Networking inside the data center got its turn next. The third era that we see coming is optimizing the geography."

For ordinary people the effect is indirect. Cheaper, faster infrastructure tends to produce cheaper, faster AI products over the long run. The more immediate audience for Relativity's pitch is the small group of cloud giants currently racing to build capacity, the same firms whose investment plans we have tracked across several recent stories, including Nvidia's $1.5 billion stake in the company behind OpenAI's Ohio campus.

What should readers watch for?

Hollow-core fiber has lived in research labs for years with almost no commercial rollout. Whether Relativity Networks can change that is the real question here, not the funding number. Watch for the unnamed hyperscaler customer to go public with a deployment, and for independent confirmation that the 30% speed gain holds outside controlled test conditions at real-world scale.