How to read a fibre capacity announcement if you buy in Europe

Four capacity announcements have landed since May. If you buy fibre or cable in Europe, you've read at least one of them and felt a small amount of relief.

Hold onto that feeling for a minute and read the postcodes first.

In May, Corning and NVIDIA announced a long-term partnership to expand US optical connectivity manufacturing tenfold and US fibre production by more than 50%, with three new plants going up in North Carolina and Texas (NVIDIA newsroom).

In July, Prysmian signed a ten-year agreement with Molex worth up to €5.5 billion for optical cables inside data centres, with €550 million paid upfront. To serve it, Prysmian committed €1.25 billion to 2031 for a capacity increase that more than doubles its fibre capacity in the United States (Prysmian, 20 July 2026).

Corning's Q2 came a week after that. Optical Communications at $2.07 billion, up 32%, sales into AI data centres roughly doubled, and a multiyear, multibillion-dollar agreement with Amazon covering its US data centres (Qz, 28 July 2026).

Emtelle went upstream instead and concluded a deal to produce 5.5 million fibre kilometres of core in Indonesia. Their own wording on where it goes: the Emtelle portfolio, plus "select customers in the United States and Europe" (Emtelle).

Read those four one at a time and they sound like the shortage easing. Read them together and something else shows up.

Why the new capacity doesn't reach the European mid-market

The short version: most of the fibre capacity announced in 2026 is being built in the United States, and almost all of it was sold to a named customer before construction started. A European distributor or contractor buying in normal commercial quantities isn't in the queue for any of it.

Every one of those announcements arrived attached to a customer. NVIDIA's demand behind Corning's plants. Molex's €550 million behind Prysmian's expansion. Amazon's US data centres behind Corning's optical growth. Emtelle's own portfolio, plus a set of select customers they chose not to name, behind the Indonesian fibre.

That's ordinary industrial logic, and in their seat I'd do the same. Nobody spends €1.25 billion on capacity and then goes looking for buyers. The offtake gets signed first, and the plant follows.

The consequence for everyone else is the part worth sitting with. Announced capacity is not free capacity. By the time it's producing, most of it already has a name on it, and the name usually isn't a European distributor buying 12 drums at a time.

CRU's read on the underlying supply layer says much the same from a different angle: constraints in preform availability are feeding through into fibre supply, and capacity additions are not catching up with global demand. European fibre prices are rising even where regional demand is soft, because pricing here is being set upstream rather than by what European operators are actually buying (CRU, April 2026).

Prices going up in a market with softer demand is the clearest signal I know that the constraint sits somewhere other than your order book.

Three questions to ask about any fibre capacity announcement

I've watched supply cycles in telecom for a long time and the headlines rhyme every time. These three questions strip one down in about five minutes.

Where is the plant?

Corning's three new sites are in North Carolina and Texas. Prysmian's doubling is in the United States, with European plants extended rather than doubled. A tonne of glass drawn in North Carolina doesn't shorten your lead time in Rotterdam.

Is the output already committed?

Look for the offtake in the same press release. Ten-year agreement, upfront payment, named partner. When those words are present, the capacity has an owner. When they're absent, ask why the manufacturer didn't mention one.

Does it make the thing you actually buy?

Most people skip this one. Fibre kilometres, cable, and finished connectorised assemblies sit in different queues with different constraints. An announcement about drawing more glass says very little about who can build you 400 custom-length pre-terminated drops in three weeks.

I'd be careful with that gap, mind. Anyone building assemblies still buys glass from the same tight market, so it's a different queue rather than a shorter one. Different queue is often enough.

What I'd do with the answers

If all three answers come back unhelpful, and for a European mid-market buyer they usually do, the move is to qualify a second source now, while your current one is still delivering and you're working to a normal deadline rather than an emergency.

Qualification takes weeks: cross-referencing specs, reading test reports, running a paid sample, placing a trial order small enough that being wrong costs you very little. Every one of those weeks is cheaper to find before the slip than after it.