Fujikura Rides the AI Buildout

TYO:5803

For most of its 140-year history Fujikura was a sensible, unexciting maker of electrical cable, optical fibre and wire harnesses. Then artificial intelligence data centres began consuming optical connectivity at a rate nobody had planned for, and a Tokyo cable company became one of the most spectacular re-ratings on the Japanese market.

From 1885 copper to 2020s photonics

Fujikura was founded in 1885 by Zenpachi Fujikura, initially producing insulated electrical wire — the infrastructure product of Japan’s industrialisation. The company grew alongside the national grid and telephone network, and then made the transition that decided its modern fate: into optical fibre, fibre cable, connectors and the tools used to join them.

Today the group runs four broad businesses:

The product that caught the AI wave

The specific technology that put Fujikura in front of the AI build-out is high-density optical cable — in particular its rollable ribbon fibre design, marketed as Spider Web Ribbon. Conventional fibre ribbon is a flat, rigid array of fibres. Fujikura’s approach connects fibres intermittently so the ribbon can roll and fold, letting far more fibres fit into the same duct diameter.

That matters because the binding constraint in modern data-centre and metro network construction is not the fibre itself; it is physical space in conduits and the labour hours required to splice thousands of individual strands. Rollable ribbon addresses both: more fibres per duct, and mass fusion splicing of twelve or more fibres in a single operation using the company’s own splicers.

Why AI changed the demand curve

• Fibre count per facility exploded. Training clusters interconnect tens of thousands of accelerators, and the east-west traffic between them is optical. A single AI campus can consume fibre counts that once served a whole metropolitan area.
• Campus-scale builds replaced single buildings. Hyperscale operators now build multi-building campuses that must be linked with very high fibre-count trunk cable.
• Speed of deployment became the constraint. When a data centre’s economics depend on being energised months earlier, splicing productivity has direct financial value — which sells splicers alongside cable.

The parts of the story investors underweight

Two cautions belong in any assessment. First, the automotive wire harness business is large, competitive and structurally low-margin, and it dilutes the group’s exposure to the optical story. Management’s direction of travel has been to improve or contain it rather than let it absorb capital.

Second, the optical boom is a capital-spending cycle, and capital-spending cycles digest. Fibre and cable have a long history of violent overbuild-and-hangover — the early 2000s telecom bust being the definitive example. Current demand is real, but the multiple assigned to it embeds an assumption that the build continues at pace.

Competition is also serious and domestic: Sumitomo Electric and Furukawa Electric supply overlapping fibre and splicing product lines, alongside Corning in the United States and Chinese manufacturers with substantial capacity.

What it means for investors and partners

• It is physical-layer AI exposure. Most AI-linked investment is concentrated in chips and power. Optical interconnect is the third bottleneck, and it is far less crowded.
• Watch splicer sales as a leading indicator. Fusion splicers are bought before large cable deployments; equipment orders lead cable volume.
• For network builders, Fujikura is a design partner. Choosing rollable ribbon architecture affects duct sizing, splice enclosure design and installation crew planning — decisions made at network design stage, not procurement.

The takeaway

Fujikura did not pivot into artificial intelligence. It kept solving the same problem for 140 years — how to get more signal through a physical route at lower installed cost — and the AI build-out turned that unglamorous discipline into the scarcest commodity in data-centre construction.