TL;DR · 30-second read
The Short Version
Light moves faster through air than through glass. That is why cable maker Prysmian and fiber developer Relativity Networks are building cable with hollow strands, which carry data nearly 50% faster.
They have now shown it can be reliably blown through plastic tubes at up to a brisk walking pace.
Why care? Buildings that run artificial intelligence together must sit close enough to work in step. This cable could stretch that limit from about 37 miles to 56, making land and electricity easier to find. The companies have not said how much they can make or what it costs.
Prysmian said in a July 22, 2026 press release that it and Relativity Networks had completed high-density cable deployment tests with conduit manufacturer Dura-Line. At Dura-Line’s test facility in Clinton, Tennessee, a cable using Relativity’s ChronoCore hollow-core fibers was installed at jetting speeds of up to 350 feet per minute. Jetting uses compressed air to carry a cable through a narrow duct.
Prysmian also said its Claremont, North Carolina plant has made a 10mm-diameter cable containing 24 hollow-core fibers that meets customer attenuation requirements, meaning limits on signal loss. The company said it is scaling fiber production, including through investments in its Eindhoven facility, and that hollow-core cables are available now for AI data center deployment.
Executive Summary
Hollow-core fiber guides light through an air-filled channel instead of solid glass, and the companies say this moves data nearly 50% faster than conventional fiber. They argue the speed gain could extend the maximum distance between data centers in a synchronized cluster from 60 km to 90 km (56 miles).
The July announcement addresses a more practical concern: whether this specialized fiber can be packed into a dense cable, manufactured in a commercial plant and installed with the microduct methods used in network construction. Microducts are small protective tubes that carry fiber cable. The reported results point toward yes on each count, although the published evidence includes few performance details.
That shifts attention to commercial questions. A single 24-fiber cable on a 90 km route would require 2,160 km of hollow-core fiber, and Prysmian and Relativity have not disclosed production volumes, pricing or named customers. For AI infrastructure planners, those figures will help determine whether the extended reach can be built into real site plans.
Why 24 Fibers Turns 90 km Into a Supply Question
The reach argument comes down to timing. Data centers working as a synchronized cluster can only be so far apart before signals take too long to arrive. The companies put that limit at 60 km with conventional glass fiber. Because hollow-core fiber carries data nearly 50% faster, they say the limit could extend to 90 km, a 50% increase in reach.
Relativity CTO Rodrigo Amezcua Correa set out the test for the technology: “A breakthrough fiber technology only matters if it can be manufactured, cabled, and deployed at scale.” The announcement offers evidence on each step, at differing levels of maturity. Production is being scaled, a commercial plant has put 24 fibers into a 10mm cable, and installation reached up to 350 feet per minute, or about 107 meters per minute. At that top speed, a continuous 90 km pull would take roughly 14 hours. Real builds proceed in segments and at varying speeds, but the figure suggests that placement speed is unlikely to be the main limit on a project of this length.
Fiber volume is a harder constraint to assess. By our arithmetic, one 24-fiber cable over 90 km uses 2,160 km of hollow-core fiber, while two separate routes for resilience would use 4,320 km, before spares or additional cables. An operator planning a multi-site AI cluster needs to know whether that quantity can be delivered on schedule and at what price. Those are now the most consequential unknowns for buyers, even though the companies describe the cable as available now. Prysmian’s Eindhoven investment signals an intent to add capacity, but no output figure accompanies it.
What the Clinton Test Does and Does Not Establish
The jetting result has practical value. Compressed-air installation is a demanding step for a cable, and carrying it out with a commercial conduit manufacturer moves the work closer to field conditions than a fiber measurement alone. Combining Relativity’s fiber with Prysmian’s existing Sirocco cable architecture also builds on an established cable design instead of an entirely new one.
The broader claims go further than the disclosed data. The headline says the companies “deploy” the cable, while the body describes tests at a dedicated facility. Those results are presented as showing readiness for real-world deployment, yet no duct length, route profile or post-installation signal-loss figure has been published. “Highest density” is also a comparison with the companies’ own hollow-core cables, not an industry benchmark. The cable “delivers the expected optical performance” and meets “customer attenuation requirements,” but neither phrase comes with a number.
None of this undercuts the milestone, which is a reasonable step 15 months into the partnership. It does separate what has been shown, a successful installation test and plant production, from what still needs proof: sustained performance on long operating routes carrying live traffic.
A Wider Map, Not More Power
The siting implication is the strongest commercial argument. On a simple geometric view, raising the radius around an existing campus from 60 km to 90 km expands the potential search area by about 2.25 times. Actual fiber routes follow roads and rights-of-way rather than straight lines, so the usable gain would be smaller and would vary by location. Even so, more candidate sites matter to operators that the companies describe as competing for scarce land, power and water.
A larger search area does not create any of those resources, however. A site 90 km away still needs its own electricity supply, permits and water plan, as well as a fiber route that can be built. Hollow-core fiber makes more locations eligible for a synchronized cluster without making any given site viable. The economic test is whether a site in the 60-to-90 km band offers enough advantage to justify any added cost for specialty fiber on that route.
If the economics work, likely beneficiaries include operators with campuses near the current distance limit, landowners and power-rich areas in that wider band, and suppliers of the cable and conduit. Prysmian’s Claremont production also matters for a market it describes as the major US data center market, since cable made in the US may simplify logistics for domestic projects. How much advantage that offers depends in part on where the fiber itself will be produced, which the companies have not fully laid out.
Background
Prysmian, a Milan-based cable maker, formed its partnership with Relativity Networks to industrialize hollow-core fiber 15 months before this announcement, according to the company. In the arrangement described, Relativity contributes its ChronoCore fiber technology, while Prysmian provides high-density cable design and manufacturing and is scaling fiber production.
Hollow-core fiber has drawn interest because AI data center networks are sensitive to latency, the time data takes to travel between points. If operators can place synchronized facilities farther apart, they can look beyond the immediate area of an existing campus for land, power and water, provided the specialized fiber can be produced and installed economically. Source: Hollow Core Fiber: Prysmian and Relativity Networks deploy their highest density HCF cable to date, Prysmian’s July 22, 2026 press release on high-density hollow-core cable testing with Dura-Line.Sources