Nexans calls cable capacity the AI data-centre bottleneck—meaning the copper price trade is now an electrical-equipment order-book story

Nexans CFO Vincent Piquet told Reuters that cable production capacity is emerging as the bottleneck for AI-driven data-centre construction. The core mechanism is timing: demand for cable is rising faster than the industry can expand manufacturing capacity.

The most load-bearing detail is that a new cross-linked polyethylene (XLPE) cable production line—called an industry “CV line”—takes around three years to stand up. In practical terms, projects can be designed and permitted while cable manufacturing slots remain the scarce resource.

The Nexans framing connects AI racks to electrical infrastructure through a chain that investors usually underweight.

First, hyperscale AI campuses concentrate power draw and require heavy low- and medium-voltage cabling to connect into on-site electrical substations and distribute power across buildings/blocks. Nexans’ Reuters-described logic is that larger facilities consume proportionately more medium-voltage cable—so the cable bill of materials scales with campus size.

Second, the cable constraint doesn’t just delay procurement; it changes how aggressively customers can commit to commissioning timelines. That matters because data-centre owners (and their contractors) usually optimize for when capacity can come online—not just what can be purchased today.

Third, Nexans indicates grid constraints can both slow some projects and simultaneously support cable demand. That combination is exactly what you get when projects are rescheduled but electrification spend stays sticky.

Investors often debate AI data-centres through fibre vs copper on the telecom/ingress layer. Nexans’ cable-capacity warning moves the debate earlier—into power distribution.

That is why “copper is the line the market still prices” can remain directionally true even as fibre deployment continues: the gating constraint described by Nexans is about cable manufacturing capacity (including the long-lead XLPE production step), and the Reuters-described data-centre wiring needs are dominated by the campus power and distribution build-out.

Where fibre can matter is in secondary bottlenecks (telecom cabling, structured cabling, and network top-of-rack interconnect choices). But in the Nexans setup, the binding constraint is the shared, electrical-infrastructure layer that sits underneath both fibre and copper cabling systems.

The Reuters report attributes the slow response to long-lead manufacturing capacity expansion: the CV line needs roughly three years. That means the constraints felt today were decided—commercially and operationally—well before hyperscalers’ latest AI build announcements.

This also helps explain why downstream behaviour changes. Nexans told Reuters that Enedis moved from annual cable procurement to multi-year contracts lasting six or seven years to secure supply. When buyers lock in multi-year terms, it’s typically because they expect delivery scarcity—not just price volatility.

And it can trigger project-level friction elsewhere: the Reuters coverage referenced Oracle issuing a force majeure notice to a Blue Owl unit over potential difficulty securing sufficient power for a large New Mexico data-centre development.

None of that requires changing the fundamental AI compute demand story. It only changes the transmission mechanism: from GPU names → to electrical distribution infrastructure → to cable manufacturing slots → to the pace at which campuses can become operational.

Copper still matters because much of the distribution and campus wiring architecture continues to use copper conductors in electrical components and cabling systems, and because cable availability is tied to material-intensive long-lead production. Even if fibre is the communications surface, the constraint Nexans points to is the power delivery backbone.

In the short term, investors should look for signals that customers are front-loading cable procurement and adjusting construction sequencing: multi-year contract language from distribution networks, increased backlog disclosure from cable manufacturers, and schedule changes from large campus projects.

In the long term, the question becomes whether industry expansions—especially XLPE/CV capacity—come online fast enough to prevent “slot scarcity” from persisting across successive build waves. If new line additions take ~three years, the market can remain tight until those expansions fully translate into deliverable shipments.

Peer cable makers and electrical infrastructure suppliers can experience a similar “infra-cycle” translation: when delivery is constrained, pricing power, backlog duration, and contract structure tend to dominate near-term outcomes.

Prysmian reported FY25 revenues and net income in its FY25 results release. Schneider Electric’s data-centre power distribution focus makes it more of an “electrical distribution” beneficiary than a direct conductor bottleneck—still, its order intake can be affected when campus timelines change.