Internet cables that carry light through air

How far away is hollow-core fibre from replacing standard glass fibre?

Hollow-core fibre guides light through a central air channel instead of solid glass. A 2025 paper by Petrovich and colleagues in Nature Photonics reports a measured loss of 0.091 dB/km at 1550 nm in a research fibre. Its preprint compares this with about 0.14 dB/km for existing silica fibre. Microsoft says its version is already in use in some Azure regions and is being deployed in more. The pages reviewed give no production volumes, and IEEE Spectrum says the fibre will need its own connectors, splicers and amplifiers.

AI illustration · Illustrative scene, not a photograph of the specific device or experiment.
WHERE WE AREPilots

Close. Early pilots are running in the real world.

Microsoft says the fibre is in use in some Azure regions and is being deployed in more, and a 2025 test installation links two Azure data centres in Europe over routes of more than 20 km. These are real networks, but they belong to one operator, the field figures come mainly from that operator, and production volumes, splicing and standards are not shown to be mature. That fits the pilot stage rather than scaling.

The tech tree

Everything recorded on Vrulum that this future needs, oldest foundations at the top and the goal at the bottom. Lines show what each step requires.

What it takes: Internet cables that carry light through air Unlocked Being researched Locked
  1. 2017Asteroid miningResearch and prototypes Being researched.
  1. 2014In-space manufacturingResearch and prototypesRequires: Asteroid mining. Being researched.
  1. 1964Reusable launch vehiclesPilot and early commercialRequires: In-space manufacturing. Unlocked.
  1. 2023Space-based solar powerResearch and prototypesRequires: Reusable launch vehicles, In-space manufacturing. Being researched.
  1. 2022Orbital data centresResearch and prototypesRequires: Space-based solar power, Reusable launch vehicles. Being researched.
  1. 2014Li-Fi (light-based wireless networking)Pilot and early commercial Unlocked.
  2. 2019Optical interconnects and co-packaged opticsResearch and prototypesRequires: Orbital data centres. Being researched.
  3. 2022Hollow-core optical fibrePilot and early commercial Unlocked.
  1. 2025Earlier design installed between two data centresDone Unlocked.
  2. 2025Loss measured below solid-glass limit in the labDone Unlocked.
  3. 2026Deployment across Azure regionsIn progress Being researched.
  4. Date not recordedManufacturing at volumeIn progress Being researched.
  5. Date not recordedConnectors, splicers and amplifiers built for the fibreIn progress Being researched.
  6. Date not recordedShared industry standards and use beyond one operatorNot started Locked.
  1. GoalInternet cables that carry light through airGoalRequires: Hollow-core optical fibre, Optical interconnects and co-packaged optics, Li-Fi (light-based wireless networking), Loss measured below solid-glass limit in the lab, Earlier design installed between two data centres, Deployment across Azure regions, Manufacturing at volume, Connectors, splicers and amplifiers built for the fibre, Shared industry standards and use beyond one operator. Locked.

The road so far

2 of 6 steps done, 3 in progress.

  1. Done · 1 Sept 2025Loss measured below solid-glass limit in the lab

    Petrovich and colleagues report a measured loss of 0.091 dB/km at 1550 nm for a hollow-core research fibre. This is a measurement on a research fibre, not a published specification for installed cable. The preprint abstract also reports loss below 0.2 dB/km across a 66 THz window. A 2023 review by Numkam Fokoua and colleagues sets out the loss mechanisms and scaling limits that such fibres face. [1][2][3]

  2. Done · 2025Earlier design installed between two data centres

    Poletti, quoted by IEEE Spectrum, says Microsoft has installed an earlier generation of the design connecting two Azure data centres in Europe. IEEE Spectrum describes this as a test installation with two routes, each over 20 km long. In June 2024, Microsoft said the fibre could extend transmission distance by up to 1.5 times with no impact on network performance, and that it is 47% faster than standard silica glass fibre. These are Microsoft's own statements. [4][6]

  3. In progress · Mar 2026Deployment across Azure regions

    Microsoft says the fibre is already in use in some Azure regions and is being deployed in more of them. Microsoft does not name the regions or give a total length in that article. IEEE Spectrum quotes Poletti saying 1,280 km is deployed and carrying live traffic. Trade press reports that Microsoft said in 2024 it planned to deploy 15,000 km over two years. The Microsoft pages checked here do not repeat that figure. [5][4][8]

  4. In progressManufacturing at volume

    Microsoft says it has agreed to manufacturing collaborations to increase production. Trade press reports that Microsoft is working with Corning and Heraeus Covantis to scale production. Microsoft has not published production volumes, yield figures or cost in the pages reviewed. Francesco Tani, quoted by IEEE Spectrum, says drawing tens or hundreds of kilometres is harder than for solid fibre, and that, as far as he knows, a significant part of fabrication is still manual. [5][4][8]

  5. In progressConnectors, splicers and amplifiers built for the fibre

    IEEE Spectrum says the technology will require its own ecosystem of connectors, splicers and signal amplifiers, and that Microsoft is working with partners to develop one. Microsoft says the fibre is designed to work with existing single-mode equipment. The two accounts differ, and no splicing or amplifier figures were given in the sources reviewed. [4][5]

  6. Not startedShared industry standards and use beyond one operator

    Trade press reports that splicing, standards and technician skills remain open challenges. The sources reviewed do not show an agreed standard for hollow-core fibre or deployment by an operator other than Microsoft. [8]

Why it matters

Standard glass fibre is installed at a very large scale, with IEEE Spectrum citing over 5 billion kilometres worldwide, so any replacement would have to work alongside that base. If hollow-core fibre performs as reported, light could travel further before it needs amplifying, which would mean fewer amplifier sites, less power use and less delay on long links. Most of the field results so far come from the company that is deploying the fibre.

What stands in the way

  • Connectors, splicers and amplifiers for the fibre must be built or adapted, and sources disagree on how much existing single-mode equipment can be reused.
  • Drawing long lengths of hollow-core fibre is harder than drawing solid fibre, and a significant part of fabrication is still manual.
  • Most latency and deployment figures come from Microsoft, which develops and deploys the fibre. The sources reviewed include no independent audit of those figures.
  • The sources reviewed do not show a public cost comparison with solid-glass fibre at volume. A 2026 modelling paper refers to a current cost premium.
  • In direct-detection (IMDD) systems using this fibre, a 2026 IEEE Network preprint names inter-modal interference as the dominant impairment.
  • Standards for splicing and installation are not yet shown to be agreed.

Published forecasts

What others have predicted. These are their estimates, not ours.

WhoForecastTimeframeMade
Md Ghulam Saber and Zhiping Jiang (authors of a 2026 IEEE Network paper)The authors conclude that hollow-core fibre is economically justified now for links inside data centres and on campuses, and that metro data-centre interconnects follow in 2027 to 2030 as manufacturing costs continue to fall. The timing rests on the authors' adoption roadmap and their comparison of transceiver savings with the current cost premium. [7]2027–203022 Jun 2026

Signals to watch

  • Published production volumes, yield or cost figures from Microsoft's manufacturing partners.
  • Hollow-core links carrying live traffic in networks run by operators other than Microsoft.
  • Independent field measurements of loss and latency on installed cable, published in peer-reviewed venues.
  • Agreed standards or commercial connectors and splicers for hollow-core fibre.
  • Microsoft or another operator naming the regions and total kilometres of hollow-core fibre in service.

Sources

  1. peer reviewed paper
    Broadband optical fibre with an attenuation lower than 0.1 decibel per kilometre
    Marco Petrovich, Eric Numkam Fokoua, Yong Chen et al.. Nature Photonics, 1 Sept 2025.
  2. preprint
    First broadband optical fibre with an attenuation lower than 0.1 decibel per kilometre
    Marco Petrovich, Eric Numkam Fokoua, Yong Chen et al.. arXiv, 27 Mar 2025.
  3. review article
    Loss in hollow-core optical fibers: mechanisms, scaling rules, and limits
    Eric Numkam Fokoua, Seyed Abokhamis Mousavi, Gregory T. Jasion et al.. Advances in Optics and Photonics, 20 Jan 2023.
  4. news
    Hollowing Out Fiber Speeds It Up and Keeps Signals Moving
    John Boyd. IEEE Spectrum, 20 Sept 2025.
  5. company announcement
    Using inexpensive MicroLEDs, Microsoft networking innovation aims to make datacenters more efficient
    Chris Welsch. Microsoft Source, 17 Mar 2026.
  6. company announcement
    How hollow core fiber is accelerating AI
    Omar Khan. Microsoft Azure blog, 25 Jun 2024.
  7. preprint
    Hollow-Core Fiber in Direct-Detection Optical Networks: Technology Readiness, Deployment Drivers, and Adoption Outlook
    Md Ghulam Saber, Zhiping Jiang. arXiv (journal reference: IEEE Network, 2026), 22 Jun 2026.
  8. news
    Microsoft Scales Up Hollow Core Fiber Production
    Doug Mohney. Fiber Broadband Association, 2 Apr 2026.

Last reviewed 21 Oct 2026.