Industries / Technology / Telecommunications / Fiber optics
Laying and splicing
Laying and splicing is the work of getting fiber cable where it has to go, under streets, onto poles, into buildings and across the sea floor, and of joining the glass ends so that light passes from one fiber to the next with little loss.
5
Milestones achieved, from 1982 to 2025.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1982
Air carries fiber into ducts
On 8 November 1982 British Telecom researchers Stephen Cassidy and Michael Reeve filed a patent for carrying light fiber packages along a small tube on a stream of air, instead of pulling them in with a rope. In 1987 the Netherlands PTT, the Dutch state telephone company, blew whole cables while a machine pushed them in, a method called jetting, and raised the length one machine could install from under 300 m by pulling to 700 m. Plumettaz, the Swiss company that licensed the Dutch method, says blowing is now the standard way to put cable into ducts.
Lessonthe Dutch traced their short pulls to small wiggles in buried ducts, each of which adds friction; air pushes along the whole length of the cable, so the wiggles matter much less.
Credit
Stephen Cassidy and Michael Reeve (British Telecom); Willem Griffioen (Netherlands PTT); Plumettaz
1995
Twelve fibers in one splice
Mass fusion splicing welds all 12 fibers of a flat ribbon at once. In 1995 engineers at Sumitomo Electric Lightwave and AT&T Bell Laboratories, citing industry studies, reported it about four times as productive as splicing fibers one at a time, with mean losses of 0.05 to 0.10 dB per splice on factory-made 12-fiber ribbon.
Lessonthe gain depends on the cable; factory-made ribbon gave almost twice the productivity of loose fibers lined up into a ribbon in the field.
Credit
Sumitomo Electric Lightwave and AT&T Bell Laboratories (reported by Lindsay, Robbins, Mettler, Cottingham and Jackson)
2008
Japan: more fiber than DSL
At the end of December 2008 Japan had 14.42 million fiber-to-the-home contracts against 11.59 million DSL contracts, broadband over copper phone lines, by the count of Japan's Ministry of Internal Affairs and Communications. An NEC engineer writing in Lightwave called Japan the first broadband market where optical access subscribers outnumbered DSL.
Credit
NTT East and NTT West, which held over 70% of fiber subscribers in 2007, and KDDI; figures from the ministry
2023
Fiber in water pipes halted
A UK government-funded project to run fiber inside live drinking-water mains stopped in May 2023, before its build phase. The independent evaluation, published in October 2024, found that the product did not have approval for contact with drinking water under Regulation 31, and that the installation method could stir up sediment in the mains.
Lessonthe evaluators recommended a first trial in a closed test network, such as Ofwat's 5 km National Leakage and Test Research Centre, rather than in mains that people drink from.
Credit
UK Department for Science, Innovation and Technology and the project consortium; evaluation by Frazer-Nash Consultancy
2025
Spain switches off copper
Telefónica closed all 8,532 of its copper local exchanges in Spain. Most retail copper service ended on 19 April 2024 and the final wholesale switch-off came on 27 May 2025, with over 99% of customers moved to fiber, or to fixed wireless or satellite in thinly populated areas, under a process notified to the CNMC, Spain's regulator.
LessonTelefónica told European regulators that "the copper switch-off cannot precede the maturity of fiber.".
Credit
Telefónica España; the CNMC
Watch for
A robot drill for driveways
MicroBORE is a small robotic drill first designed for exploring the Moon and Mars. Its makers say it bores narrow, steered paths under pavements and driveways, so fiber can run from the street duct to a home without an open trench. Early tests at BT's research center in Suffolk "confirmed the operational viability," and Openreach is evaluating wider use; no bore length, speed or cost was published.
Achieved whenOpenreach uses it for routine customer connections, or an outside party publishes measured time and cost per connection against a dug trench.
Claimed by
BT Group, Openreach, Synthotech and the University of Surrey, November 2025
Two fibers per splicer load
AFL says the Fujikura 100S splicer and CT60 cleaver let a technician prepare and load two fibers at once, cutting "splicing preparation time by up to 30%," and gives no time per splice.
Achieved whena contractor, trainer or researcher outside AFL and Fujikura publishes a timed comparison on the same splicing jobs that shows the saving.
Claimed by
AFL, January 2026
Field splicer for hollow-core fiber
Hollow-core fiber carries light in an air channel, and its thin inner glass walls are easily damaged by heat. Furukawa Electric and Lightera say their FITEL S185EVROF splicer avoids that damage by heating the fiber evenly from three electrodes, inspects the fiber's end by camera, and is compact enough for field work. Neither the June 2025 launch nor the March 2026 update gave splice-loss figures.
Achieved whenan operator or lab outside the two companies publishes loss and strength figures for hollow-core splices made with it in the field.
Claimed by
Furukawa Electric and Lightera, June 2025 and March 2026
Open
01
An open cable-blowing calculator
An open-source program that predicts how far a fiber cable will blow into a given duct route, and works backwards from a blowing test log to estimate the friction between cable and duct, including tests in which the cable stopped short of the end, and the "fill effect," in which the cable's own bulk changes how air pressure falls along the duct.
The briefCredit · unclaimed02
A public stopwatch for splicing
A written, repeatable method and an open, labeled video set that times every step of a fusion splice on a bench (strip the coating, clean, cleave the glass to a flat end, load, fuse, slide on and heat the protective sleeve, and store the splice in its tray) for single fibers and 12-fiber ribbons, with a small model that marks where each step starts and ends in new video. The same clips could serve as training episodes for the robot splicing job on the Telecommunications page.
The briefCredit · unclaimed
Open 01 · the brief
An open cable-blowing calculator
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open-source program that predicts how far a fiber cable will blow into a given duct route, and works backwards from a blowing test log to estimate the friction between cable and duct, including tests in which the cable stopped short of the end, and the "fill effect," in which the cable's own bulk changes how air pressure falls along the duct.
Why it is still open
Willem Griffioen of Plumettaz and Jonker of Prysmian wrote in 2021 that the fill effect "has not yet been implemented in the software" and that "no software exists to 'back-calculate'" friction from a test in which the cable did not reach the end. A GitHub repository search on 19 November 2026 for "cable blowing," "cable jetting" and "blown fiber" found 8 results and no calculator; the one named Cable-Blowing is empty.
Built with what exists
The pushing-force and air-pressure equations printed in Griffioen and Jonker (2021); their rule of thumb that wiggles in a duct laid in a trench act like a bend of radius about 2,000 times the duct's inner diameter, about 16 m for a 10/8 mm microduct; the IEC 60794-1-21 Method E24 test route they cite; Python with SciPy. No new AI model is needed; the job is open because nobody has published the code.
First step
Code equations (1) and (3) of the 2021 paper for one 10/8 mm microduct and one cable size, plot the Fig. 2 curves, and post the notebook.
Done looks like
Public code under an open license, with tests; a forward model that matches the curves of Fig. 2 in the 2021 paper within 5% at every point read off the figure; on 100 simulated test logs with known friction, where the cable stops between 50% and 90% of the route, the friction coefficient recovered within ±0.02 in at least 95 cases; and, fitted on one published test with cable, duct, pressure and route data, a predicted blowing length for two other published tests within 20%.
Open 02 · the brief
A public stopwatch for splicing
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A written, repeatable method and an open, labeled video set that times every step of a fusion splice on a bench (strip the coating, clean, cleave the glass to a flat end, load, fuse, slide on and heat the protective sleeve, and store the splice in its tray) for single fibers and 12-fiber ribbons, with a small model that marks where each step starts and ends in new video. The same clips could serve as training episodes for the robot splicing job on the Telecommunications page.
Why it is still open
Makers publish speed claims without times or a method: AFL's January 2026 release claims "up to 30%" less preparation time and gives no time per splice, and the 1995 figure that ribbon splicing is "four times more productive" rests on unnamed industry studies. The Fiber Broadband Association and Cartesian name labor and materials as the top drivers of US build costs, with medians of $18 per foot underground and $8 per foot aerial in 2025. Searches of GitHub and Hugging Face on 19 November 2026 found course materials and a teaching simulator, but no timed or labeled splicing video.
Built with what exists
A phone camera on a stand over a bench splicing station; any core-alignment fusion splicer and cleaver; CVAT, an open-source tool for labeling video; a small video classifier fine-tuned on the labeled frames. Scrap fiber only, offcuts in a sealed bin, eye protection, and nothing connected to a live network.
First step
Write one-line start and end rules for each step, film 20 bench splices on scrap fiber, label them in CVAT, and post the clips and labels.
Done looks like
The method published; at least 300 splices timed, by at least 5 people from novice to experienced, on at least 2 splicer models; the median and spread reported for each step; the model placing at least 90% of step boundaries within 2 seconds of a person's label on unseen video; and a separate team repeating the method on 50 splices with step medians within 15% of the first team's.