Why the optical fibre shortage lasts into 2027: the preform maths

Ask for a quote on common single-mode cable and the second quarter's answer was 6 to 10 weeks, longer for high-count, ribbon and specialty builds. Some fibre types weren't being quoted at all (M2 Optics, June 2026).

Everyone in the trade calls it a shortage. The better question is when it ends. The answer sits upstream, in glass, and it comes with a date.

The glass comes first

Every kilometre of fibre starts as a preform: a rod of ultra-pure glass, grown slowly, then drawn into strands on a draw tower. Fibre output is capped by preform output, and cable is capped by both.

CRU's February read on that upstream layer: effective preform utilisation is running near its limits once mothballed plant is excluded, and restarting idled capacity is measured in months (CRU, February 2026).

The price signal followed. Chinese G.652.D bare fibre rose more than 80% between November 2025 and January 2026, and for the first time in over 7 years Chinese prices moved above European levels (CRU). When the cheapest glass in the world reprices above Europe's, capacity is doing the talking.

The 18-month clock

New capacity is slow to add. A new preform takes up to 18 months before it's ready to feed a draw tower, so glass started today reaches production around late 2027 or early 2028 (M2 Optics). The same analysis reports manufacturers' order books filled well into mid-to-late 2027, and expects no material improvement for the rest of 2026 or well into 2027.

That's the arithmetic worth pinning above the procurement desk. Relief is a 2027 story at the earliest, and glass that hasn't been started yet arrives closer to 2028.

The queue got longer again

Corning's optical communications business booked $1.85bn in the first quarter, up 36%, and disclosed two further long-term hyperscale agreements similar in size and duration to its up-to-$6bn Meta deal (Qz, April 2026). NVIDIA followed in May with up to $3.2bn toward three new US fibre plants, and Amazon signed its own multiyear agreement in June (RCR Wireless, June 2026).

A buyer holding a signed multiyear agreement stays in the queue at any price. The usual relief valve, demand backing off as prices climb, is bolted shut for as long as those contracts run.

What a date changes

Once the squeeze has a date on it, you can plan against it instead of hoping through it.

Two decisions look different against late 2027. Volume you know you'll need in the first half of 2027 is worth locking now rather than shopping in spring. And a second source is worth qualifying before you need it, while there's still time to run the process properly instead of buying from whoever has stock.

Start where qualification is fastest

Qualifying a new cable design into a route is slow work. Assemblies move faster: patch cords, pigtails, MPO trunks and pre-terminated drops carry a shorter test cycle, and nothing about the buried plant changes. In a market where the raw glass is the constraint, the assembly layer is where a second source gets qualified in weeks instead of seasons.

It's also where factory termination gives you back field hours, and splicing crews are the other input in short supply.

Where I stand in this

I should be plain about my interest. Jupiter Enterprises, our agency in Amsterdam, represents fibre-optic manufacturers commercially in Europe. One of them, PDR World, has built fibre optics to spec for over 40 years, for Siemens, ABB, Alstom, Hitachi and Colt among others, and builds to order: custom assemblies air-freighted in days, where the established names quote months.

When a buyer wants a second source qualified properly, we carry the work: a cross-reference of your top 10 lines against mapped equivalents with test reports, a paid sample through the normal process, then a trial run with a real deadline. Replies within one working day.

Run the maths on your own book

Take the part numbers you'd least like to be waiting on next spring and price them against a mapped equivalent. If the equivalent holds on spec and lands sooner, you have your answer. If it doesn't, you've spent an afternoon and learned exactly which lines you're genuinely single-sourced on, which is worth knowing before the spring rather than during it.