The hollowing-out of fiber optics just passed an important milestone: a research team tied to Microsoft and the University of Southampton has reported a hollow‑core optical fiber design that both lowers signal loss below the long‑standing practical floor for silica fiber and propagates light substantially faster — changes that could materially reshape data‑center interconnects and long‑haul transport if the engineering and supply‑chain hurdles are cleared. The work centers on a double‑nested anti‑resonant nodeless hollow‑core fiber (DNANF) that pushes measured attenuation to about 0.091 dB/km at 1,550 nm, claims a usable low‑loss window spanning tens of terahertz, and reports a propagation‑speed advantage on the order of 45% versus conventional silica single‑mode fiber. These are headline numbers that deserve careful scrutiny, but they also mark a step toward hollow‑core fiber (HCF) moving from laboratory curiosity toward real‑world deployment.
Hollow‑core fiber replaces the solid silica glass core used in traditional single‑mode fiber (SMF) with an air‑filled core and guides light using an engineered microstructure of thin glass capillaries around that core. Because light travels through air far closer to vacuum speed than through silica, and because many glass‑dependent loss mechanisms (notably Rayleigh scattering and nonlinear effects) are greatly reduced, the physical promise of HCF has been recognized for decades.
Until recently, however, practical losses for HCF designs were orders of magnitude worse than SMF — typically measured in dB/km well above 1 dB/km — which made the technology impractical for mainstream communications. The new DNANF design substantially narrows that gap by combining nested anti‑resonant structures to suppress three dominant loss mechanisms: confinement leakage, surface scattering, and microbend‑induced coupling. The result reported by the authors is a fiber in which more than 99.99% of the optical energy propagates in air, producing both lower attenuation and lower effective group index than silica‑core fiber.
Nesting adds concentric rings of these tubes — the DNANF reported here implements a double nested ring, which further reduces leakage and improves confinement without dramatically increasing glass volume in the optical path. The layered approach also reduces the coupling to higher‑order modes and lowers the sensitivity to microbends. That careful geometric tuning is why the team reports both ultralow loss and high modal purity.
However, reporting varies on the precise scale of deployed HCF in production: some outlets report ~1,200 km of installed HCF carrying live traffic in pilot configurations, while other claims cite 1,280 km as a figure presented in certain deployments or internal briefings. These deployment numbers appear in corporate and press materials rather than in the peer‑reviewed technical paper; they should therefore be treated as operational claims that are meaningful but not independently audited by the research publication itself. The distinction matters when assessing how close HCF is to mass commercial adoption.
That said, several real‑world considerations remain before HCF becomes a mainstream replacement for silica SMF: manufacturing scale and yields, splice and connector ecosystems, componentization beyond the fiber itself, and independent long‑distance field validations. Competing vendor claims — from YOFC, Linfiber and others — show the field is advancing rapidly, but they also underscore that multiple architectures and manufacturing approaches are in play. Some vendors report continuous draws of dozens of kilometers and low per‑km loss in certain conditions; those achievements are promising, but these are not yet the same as a commoditized global supply chain for 100,000+ km network rollouts.
In short: DNANF and the 0.091 dB/km milestone are real technical milestones that materially raise the probability hollow‑core fiber becomes a mainstream infrastructure option. The path from milestone to mass adoption runs through manufacturing scale‑up, standards and ecosystem maturity, and independent operator‑grade deployments. For IT teams and network architects, the immediate task is pragmatic: begin pilots in high‑value corridors, insist on independent measurement and production guarantees, and track component and tool availability. If the manufacturing and systems engineering pieces fall into place, hollow‑core fiber could be one of the more consequential infrastructure technologies of the next decade — especially for hyperscalers and anyone building networks to feed a future of distributed AI.
Hollow‑core fiber has moved from a theoretical advantage into a field of credible engineering milestones. The DNANF record and parallel industry efforts put HCF on the short list of technologies that could relieve both the capacity ceiling and latency constraints that modern AI and distributed systems now face. The next chapter will be written in factories, splice labs, and operator testbeds — and in those domains the combination of rigorous independent testing, standards work, and realistic production roadmaps will decide how quickly hollow core becomes part of the everyday network fabric.
Source: IEEE Spectrum Could Hollow-Core Fiber Be the Key to Faster Data Centers?
Hollow‑core fiber replaces the solid silica glass core used in traditional single‑mode fiber (SMF) with an air‑filled core and guides light using an engineered microstructure of thin glass capillaries around that core. Because light travels through air far closer to vacuum speed than through silica, and because many glass‑dependent loss mechanisms (notably Rayleigh scattering and nonlinear effects) are greatly reduced, the physical promise of HCF has been recognized for decades.Until recently, however, practical losses for HCF designs were orders of magnitude worse than SMF — typically measured in dB/km well above 1 dB/km — which made the technology impractical for mainstream communications. The new DNANF design substantially narrows that gap by combining nested anti‑resonant structures to suppress three dominant loss mechanisms: confinement leakage, surface scattering, and microbend‑induced coupling. The result reported by the authors is a fiber in which more than 99.99% of the optical energy propagates in air, producing both lower attenuation and lower effective group index than silica‑core fiber.
What the measurements actually report
Key lab/test numbers (verified across independent reports)
- Measured minimum attenuation: ~0.091 dB/km at 1,550 nm (headline figure reported by the research team). Multiple journalistic and trade outlets corroborate this number, and the authors present it as a record low for HCF configurations.
- Broadband low‑loss window: the fiber is reported to maintain <0.2 dB/km across a ~66 THz spectral band, providing a much wider usable spectrum than modern silica SMF typically exploits. Several reports (and the team’s published materials) emphasize this wide window as a platform reason why HCF could increase per‑fiber capacity.
- Subband performance: press and test summaries point to an 18 THz subband where loss stayed below 0.1 dB/km on spool‑length tests. That subband detail helps transport engineers map HCF into WDM designs but should be validated for specific operational wavelengths on a per‑network basis.
- Propagation speed: the effective group index of the air‑filled core gives a ~45% reduction in propagation delay (reported as “up to 45% faster”) relative to silica SMF for the designs tested. This is a physical consequence of the refractive‑index difference between air (~1.0) and silica (~1.44 around 1,550 nm) and is consistent across independent technical summaries.
- Practical spool length tested: measured low losses were validated on multi‑kilometre spools — tests up to 15 km are repeatedly noted in the technical descriptions — though achieving the same low loss consistently on production reels and in long continuous draws remains a manufacturing challenge.
What those numbers imply in practice
Lower per‑km attenuation compounds over distance. For example, at 0.091 dB/km a 1,000‑km span would lose ~91 dB, compared with roughly 140–160 dB for conventional silica at 0.14–0.16 dB/km — a nontrivial difference in amplifier count, power consumption, and repeater station economics for long‑haul links. The lower group index reduces one‑way latency per kilometer, an outcome with clear value for distributed AI training, high‑frequency finance, and other latency‑sensitive workloads. Finally, the broad low‑loss spectral window opens the technical possibility of exploiting bands outside the crowded C‑band — assuming photonic components and amplifiers evolve to support them cost‑effectively.Design and physics: why nested rings matter
Anti‑resonant guidance and nesting
The DNANF design is an evolution of anti‑resonant hollow‑core concepts that use thin glass membranes (capillaries) to create a high‑contrast microstructure around the core. The tubes act as anti‑resonant reflectors for the wavelengths of interest, bouncing the optical energy back into the hollow center and strongly reducing escape into the cladding.Nesting adds concentric rings of these tubes — the DNANF reported here implements a double nested ring, which further reduces leakage and improves confinement without dramatically increasing glass volume in the optical path. The layered approach also reduces the coupling to higher‑order modes and lowers the sensitivity to microbends. That careful geometric tuning is why the team reports both ultralow loss and high modal purity.
Loss mechanisms the team targeted
- Confinement leakage — reduced by stronger anti‑resonant stacks and nested geometries.
- Surface scattering — minimized by improving the smoothness and consistency of inner glass surfaces and optimizing wall thickness.
- Microbending coupling — mitigated through geometry and improved mechanical designs that reduce stress coupling to mode fields.
Real‑world deployment: pilots, promises, and competing claims
Microsoft / Lumenisity activity
Microsoft acquired Lumenisity (a University of Southampton spin‑out) in December 2022 and integrated its Romsey, UK manufacturing site and HCF IP into Azure’s networking efforts. Public statements from Microsoft and trade reporting indicate pilot deployments, a company‑level ambition to deploy thousands of kilometers within Azure’s backbone, and field trials connecting Azure data centers with hybrid cable where HCF strands are routed alongside standard single‑mode fiber. Microsoft and Azure engineering have framed HCF as a strategic enabler for lower‑latency, higher‑capacity data‑center interconnects.However, reporting varies on the precise scale of deployed HCF in production: some outlets report ~1,200 km of installed HCF carrying live traffic in pilot configurations, while other claims cite 1,280 km as a figure presented in certain deployments or internal briefings. These deployment numbers appear in corporate and press materials rather than in the peer‑reviewed technical paper; they should therefore be treated as operational claims that are meaningful but not independently audited by the research publication itself. The distinction matters when assessing how close HCF is to mass commercial adoption.
Competitors and independent advances
China’s Yangtze Optical Fibre and Cable (YOFC) and companies such as Linfiber have reported comparable progress on length, loss, and manufacturability — but with different emphases. YOFC announced a demonstration 21.7‑km HCF (ST‑HCF structure) and publicized an attenuation claim as low as 0.05 dB/km in certain tests and a continuous single‑fiber draw length over 20 km; those claims were made in OFC presentations and company PR. Linfiber has reported continuous draws and in‑field tests, and company briefings cite continuous drawing of 47.5 km of hollow core fiber at 0.1 dB/km, a figure that is widely reported in trade coverage but is primarily presented by the company at conferences and on its website. These parallel efforts show that the field is competitive and that manufacturing approaches vary between players.What to watch for in vendor claims
- Whether loss figures were measured on short spool samples versus long continuous reels and whether the values include connector/splice/cabling penalties.
- Whether reported numbers are single‑sample bests, average production figures, or guaranteed specifications for shipment‑lengths.
- Whether deployment claims (kilometers installed) are pilot, demonstration, or audited carrier‑grade installations.
Strengths and immediate benefits for networks
- Lower link loss per km — even small per‑km savings compound across hundreds or thousands of kilometers, reducing amplifier counts and possibly enabling longer amplifier‑free spans in terrestrial and subsea links. This can drive tangible OPEX and CAPEX savings for operators.
- Lower propagation latency — a ~45% speed improvement on the optical propagation term implies meaningful reductions in round‑trip times across distributed systems and multi‑site AI clusters, enabling faster synchronization and potentially altering system design choices for latency‑sensitive applications.
- Wider usable spectrum — a 66 THz low‑loss window opens possibilities for expanded WDM architectures and the use of non‑C‑band wavelengths, giving operators a way to scale capacity per fiber beyond current practical constraints.
- Reduced nonlinear penalties — moving most of the optical energy into air greatly reduces nonlinear effects that limit capacity at high launch powers, benefiting high‑power and high‑capacity links.
Risks, unknowns, and engineering hurdles
- Manufacturing scale and yield: producing hundreds or thousands of kilometers of HCF with the geometric tolerances needed for sub‑0.1 dB/km loss is not the same as making a few 10‑km spools. Drawing stability, contamination control (especially gas in the hollow core), and consistent wall thicknesses are all harder at volume and require process automation improvements. Independent experts warn that a significant portion of fabrication remains manual in many labs.
- Splicing, connectors and cabling ecosystem: HCF requires new connectors, splicing methods, and protective cabling approaches. Field splicing loss, robustness to bending and mechanical stresses, and long‑term reliability (including moisture and particulate control in the hollow core) remain active development areas. Existing fiber plant and ODF hardware are not drop‑in compatible with all HCF variants.
- Component and amplifier ecosystem: exploiting the broader low‑loss spectrum will require amplifiers, EDFAs/PAs (or new amplifier tech), tunable lasers, and transceivers that are optimized for bands beyond the C‑band — a nontrivial supply‑chain and standards effort. The per‑fiber capacity gains are only realizable if the rest of the photonics stack follows.
- Field performance variance: lab spool performance may not directly translate to long, cabled, environmentally exposed deployments. Splice losses, cable stress, and real‑world conditions can degrade the metrics versus controlled testbed numbers. Until multiple independent operator deployments confirm consistent production yields and installed‑length performance, caution is warranted.
- Proprietary formats and standards: a fragmented ecosystem with multiple HCF geometries risks locking operators into particular vendor solutions. Industry harmonization on connectors, splicing procedures, and qualification standards will be essential for broad adoption.
How HCF might be deployed in practice
Immediate, near‑term use cases (0–36 months)
- Data center interconnect (DCI) short links — metro and campus runs between co‑located facilities where the HCF’s latency and dispersion benefits can be used to reduce DSP complexity and improve synchronous training loops for AI. Microsoft has already piloted HCF in DCI contexts.
- Point‑to‑point low‑latency routes — trading links and specialty network segments where every microsecond counts may adopt HCF to gain a latency advantage.
- Specialized high‑power laser delivery and sensing — HCF’s reduced nonlinearity and higher damage threshold make it suitable for long‑range, high‑power laser delivery applications.
Mid‑term (2–5 years)
- Backbone augmentation — operators may deploy HCF in high‑value backbone corridors where amplifier spacing and power costs are a major part of OPEX. Here HCF could displace or complement existing SMF, but only after connector/splice ecosystems mature.
- Hybrid cable strategies — combining HCF fibers with standard SMF in the same cable sheath to enable progressive migration without replacing entire plant networks during early adoption. Microsoft’s hybrid pilots follow this pragmatic approach.
Long term (>5 years)
- Widespread replacement or parallel rollout if manufacturing scales, field reliability is proven, and photonic components evolve to use the wider spectrum. Standardization and economies of scale would be required; the transition will likely be incremental and route‑selective rather than wholesale.
Practical guidance for operators and IT architects
- Inventory current interconnects and identify high‑value, latency‑sensitive corridors where HCF could deliver measurable ROI (e.g., AI cluster replication, financial links, cross‑region synchronous services).
- Run controlled pilots using hybrid cables and monitor splice losses, environmental sensitivity, and long‑term stability. Use multi‑point tests and independent measurement labs to validate vendor claims.
- Engage with vendors and standards bodies to track connectorization and splicing tool development. Demand production guarantees for attenuation vs. reel length and defined environmental stress tests.
- Quantify amplifier and transponder lifecycle impacts: fewer amplifiers also shift where risk and capital are expended, so model the full systems‑level economics rather than only per‑km attenuation numbers.
Balanced assessment and concluding view
The DNANF results reported by the Microsoft/University of Southampton team represent a meaningful technical advance that reduces attenuation and propagation delay in hollow‑core fibers to levels that — if reproduced consistently at commercial scale — could change how networks are engineered for latency‑sensitive and capacity‑hungry workloads. Multiple independent outlets summarize the same headline figures (0.091 dB/km, ~45% faster, wide low‑loss windows) and Microsoft’s acquisition of Lumenisity gives a clear industrial pathway to commercialization.That said, several real‑world considerations remain before HCF becomes a mainstream replacement for silica SMF: manufacturing scale and yields, splice and connector ecosystems, componentization beyond the fiber itself, and independent long‑distance field validations. Competing vendor claims — from YOFC, Linfiber and others — show the field is advancing rapidly, but they also underscore that multiple architectures and manufacturing approaches are in play. Some vendors report continuous draws of dozens of kilometers and low per‑km loss in certain conditions; those achievements are promising, but these are not yet the same as a commoditized global supply chain for 100,000+ km network rollouts.
In short: DNANF and the 0.091 dB/km milestone are real technical milestones that materially raise the probability hollow‑core fiber becomes a mainstream infrastructure option. The path from milestone to mass adoption runs through manufacturing scale‑up, standards and ecosystem maturity, and independent operator‑grade deployments. For IT teams and network architects, the immediate task is pragmatic: begin pilots in high‑value corridors, insist on independent measurement and production guarantees, and track component and tool availability. If the manufacturing and systems engineering pieces fall into place, hollow‑core fiber could be one of the more consequential infrastructure technologies of the next decade — especially for hyperscalers and anyone building networks to feed a future of distributed AI.
Quick reference: the most load‑bearing claims and supporting evidence
- 0.091 dB/km at 1,550 nm — reported by the research team and covered by multiple independent outlets.
- ~45% faster propagation vs. conventional SMF — physical consequence of air‑core group index and reported by the authors and trade press.
- Broad low‑loss bandwidth (~66 THz) with an ~18 THz subband <0.1 dB/km — reported in the technical materials and press coverage; exploiting this window requires ecosystem upgrades.
- Microsoft/Lumenisity production and pilot deployments — Microsoft acquired Lumenisity (Dec 2022) and public statements describe pilot routes and installed kilometers, but deployment counts vary between sources and should be treated as corporate operational claims pending independent audits.
- Competitor advances (YOFC, Linfiber) — both companies report long continuous draws and low losses in conference presentations and company releases; these claims corroborate rapid progress but are vendor‑reported and require independent cross‑validation.
Hollow‑core fiber has moved from a theoretical advantage into a field of credible engineering milestones. The DNANF record and parallel industry efforts put HCF on the short list of technologies that could relieve both the capacity ceiling and latency constraints that modern AI and distributed systems now face. The next chapter will be written in factories, splice labs, and operator testbeds — and in those domains the combination of rigorous independent testing, standards work, and realistic production roadmaps will decide how quickly hollow core becomes part of the everyday network fabric.
Source: IEEE Spectrum Could Hollow-Core Fiber Be the Key to Faster Data Centers?