Microsoft and the University of Southampton have published what the teams describe as a watershed result in optical communications: a hollow‑core optical fiber with measured attenuation of 0.091 dB/km at 1,550 nm, a performance level that — if reproduced in production volumes and field conditions — would beat the long‑standing practical minimum for silica glass single‑mode fiber and reopen design possibilities for long‑haul and data‑center networks. (arxiv.org, 104887[/ATTACH]Background: why attenuation and medium matter[/HEADING]
Conventional telecom fibers use a solid silica glass core and have for decades sat near a practical floor in attenuation: the best silica fibers run roughly 0.14–0.16 dB/km at 1,550 nm, a figure driven by intrinsic material scattering and absorption. That practical floor shaped network architectures — repeater spacing, amplifier sites, wavelength choices and even submarine cable economics — for the last 30–40 years. ([url="]mdpi.com[/url], [url="]nature.com)
Hollow‑core fiber (HCF) replaces the glass core with an air (or vacuum) core and confines light using a carefully engineered glass microstructure around that core. Because light propagates mostly through air rather than silica, Rayleigh scattering and many glass‑dependent loss mechanisms are dramatically reduced, and the effective group index can be closer to 1.0 — meaning lower latency and reduced nonlinear penalties at high launch power. The tradeoffs historically have been manufacturing complexity, higher early‑generation losses and constraints around bending and field deployment. (azure.microsoft.com)
Industry reporting also states that a pilot involved over 1,200 km of installed hollow‑core fiber that is carrying live traffic, and that Microsoft announced ambitious deployment targets in presentations at Ignite. Those deployment figures appear in trade coverage; the underlying scientific paper is focused on lab and test‑bed characterization and modeling, while corporate announcements and interviews provide the deployment framing. The deployment quantities reported in the press should be read as Microsoft’s operational claims rather than independent, peer‑reviewed measurements of every installed kilometer. (azure.microsoft.com)
Cautionary note: a number widely quoted in some outlets — 15,000 km of hollow‑core fiber to be deployed across Azure — is attributed to a company presentation, but a direct, independent confirmation (for example a Microsoft press release or regulatory filing listing precise routes and build schedules) is not publicly available in the scientific paper itself. That figure should be treated as a corporate deployment target reported in press coverage, not an independently audited installation statistic. (blogs.microsoft.com)
Conventional telecom fibers use a solid silica glass core and have for decades sat near a practical floor in attenuation: the best silica fibers run roughly 0.14–0.16 dB/km at 1,550 nm, a figure driven by intrinsic material scattering and absorption. That practical floor shaped network architectures — repeater spacing, amplifier sites, wavelength choices and even submarine cable economics — for the last 30–40 years. ([url="]mdpi.com[/url], [url="]nature.com)
Hollow‑core fiber (HCF) replaces the glass core with an air (or vacuum) core and confines light using a carefully engineered glass microstructure around that core. Because light propagates mostly through air rather than silica, Rayleigh scattering and many glass‑dependent loss mechanisms are dramatically reduced, and the effective group index can be closer to 1.0 — meaning lower latency and reduced nonlinear penalties at high launch power. The tradeoffs historically have been manufacturing complexity, higher early‑generation losses and constraints around bending and field deployment. (azure.microsoft.com)
The breakthrough: what the paper reports
Key measured numbers
- Measured minimum attenuation: 0.091 dB/km at 1550 nm — reported in the team’s preprint and presented as having been accepted for publication in Nature Photonics.
- Broadband low‑loss window: the authors report the fiber remains below 0.2 dB/km across a ~66 THz spectral window, substantially wider than the practical low‑loss bandwidth of conventional silica fibers.
- Reported low‑loss subband: multiple industry outlets report the new fiber maintains loss <0.1 dB/km across an 18 THz band, a detail attributed to the experimental characterization; this specific subband claim appears in press coverage and should be considered supported by the team’s paper and subsequent reporting, but it is worth noting that the published preprint highlights the 0.091 dB/km point and the 66 THz <0.2 dB/km window most prominently. (phys.org)
Design and physics in brief
The fiber uses a nested anti‑resonant, micro‑structured tubular geometry — the evolution of a design lineage (Kagome, single‑ring tubular, nested anti‑resonant nodeless fibers) aimed at minimizing confinement loss and surface scattering. By carefully tuning capillary wall thicknesses and the nested geometry, the team reduced three dominant loss terms: leakage (confinement loss), surface‑scattering from glass roughness, and microbending‑induced coupling to cladding modes. The result is an air‑core guided mode with very low net attenuation across a very wide spectrum. (arxiv.org, arxiv.org, networkworld.com, arxiv.org)Lower latency for time‑sensitive applications
Because the group index in hollow‑core fibers approaches that of air, one‑way propagation delay per km falls. For latency‑sensitive workloads — high‑frequency trading, distributed AI training loops, AR/VR, remote‑control systems — smaller latency adds up. The Microsoft Azure engineering team has framed HCF as an enabler for more distributed data‑center architectures and faster inter‑site paths. (phys.org)Wider spectral choices and higher per‑fiber capacity
A broad, low‑loss window (the 66 THz claim) opens the possibility of using different amplifier technologies and wavelength bands than the constrained C‑band focus of today’s long‑haul links. That extra spectral real estate could be exploited to multiply transmission capacity per cable or to shift some traffic to bands with cheaper or more efficient photonics. The authors argue this is a pathway toward much larger aggregate capacity per fiber.The Microsoft angle: acquisition, factory and trials
Microsoft’s acquisition of Lumenisity (a University of Southampton spin‑out) in December 2022 gave the company both the ORC heritage and a Romsey, UK manufacturing facility. Microsoft has invested in scaling that facility and in R&D on the hollow‑core designs, and its Azure engineering organization is a co‑author organization on the preprint. Microsoft blog posts and Azure materials emphasize HCF’s role in accelerating AI and low‑latency networking. (azure.microsoft.com)Industry reporting also states that a pilot involved over 1,200 km of installed hollow‑core fiber that is carrying live traffic, and that Microsoft announced ambitious deployment targets in presentations at Ignite. Those deployment figures appear in trade coverage; the underlying scientific paper is focused on lab and test‑bed characterization and modeling, while corporate announcements and interviews provide the deployment framing. The deployment quantities reported in the press should be read as Microsoft’s operational claims rather than independent, peer‑reviewed measurements of every installed kilometer. (azure.microsoft.com)
Cautionary note: a number widely quoted in some outlets — 15,000 km of hollow‑core fiber to be deployed across Azure — is attributed to a company presentation, but a direct, independent confirmation (for example a Microsoft press release or regulatory filing listing precise routes and build schedules) is not publicly available in the scientific paper itself. That figure should be treated as a corporate deployment target reported in press coverage, not an independently audited installation statistic. (blogs.microsoft.com)
Technical caveats and deployment risks
High‑profile lab numbers are necessary but not sufficient to guarantee wide commercial impact. The following practical engineering and commercial risks merit careful attention.1) Manufacturing tolerance and scale
- The nested anti‑resonant designs require sub‑micron control of capillary wall thickness and very tight geometry tolerances. Achieving lab performance in production reels requires process control beyond standard silica draws, including control of gas contamination and long‑length uniformity. The preprint and Microsoft materials highlight that removing manufacturing impurities expanded the low‑loss window; this implies manufacturing quality is a gating factor for large‑scale rollout. (arxiv.org, nature.com, arxiv.org, networkworld.com, arxiv.org, mdpi.com, arxiv.org, Microsoft’s hollow core fiber delivers the lowest signal loss ever
