Start with the thing everyone in the room already knows but nobody can fix: you cannot buy enough DCI gear right now. DCI — data-center interconnect — is the long-haul optical equipment that stitches data centers together, carrying traffic between buildings and across cities on light over fiber. The AI build-out has turned it into one of the hardest-to-get items in the whole infrastructure stack.
How tight is it? Lead times that used to run one to one-and-a-half years have stretched to two to two-and-a-half. Ciena, the long-time market leader, runs about $3.5 billion a year of capacity — and it is effectively sold out, with 2027 already booked. Meanwhile Google, Microsoft and Meta have all raised their budgets and started tendering early to lock supply in. When demand runs that far ahead of supply, the usual questions (whose product is a hair faster? whose roadmap is prettier?) stop mattering. One question takes over: who can actually deliver? That is the lens for everything below.
Edge 1 — the idle capacity nobody else has
This is the decisive one. When Nokia bought Infinera — a US optical-systems maker — it inherited something priceless in a sold-out market: large, ready-to-run, idle production capacity. Infinera's North American lines, centered in San Jose, were sitting largely unused: on the order of $2.5 billion a year of capacity dark, and up to roughly $4 billion if you run them hard with overtime.
In a market where everyone else is booked solid, that makes Nokia the only supplier who can say yes to a giant new order and ship it on a normal timeline. That is not a small advantage — it is the advantage. It is the reason Google handed Nokia the single biggest slice of its DCI tender, on the order of 50–60%. Not because the product won a feature shootout, but because Nokia was the one vendor that could promise the boxes would actually arrive.
Edge 2 — it owns its own supply chain
Capacity gets you in the door; owning your inputs keeps you there. Nokia is unusually vertically integrated — it makes its own critical parts instead of buying them. Two pieces matter most:
- Its own DSP. The DSP (digital signal processor) is the chip that encodes data onto the light and decodes it at the other end — the brain of an optical link. Nokia designs its own, now at 1.2 terabits per wavelength on a 5-nanometer process. Most rivals buy this chip from merchant suppliers like Broadcom or Marvell.
- Its own indium-phosphide fab. Indium phosphide (InP) is the semiconductor the lasers themselves are built from. Nokia runs its own InP wafer fab plus packaging and test — so it isn't waiting in line for the photonics that go inside every transponder.
Why it matters: the single scarcest part in a DCI system right now is the DSP chip. A vendor that buys it on the open market is exposed to the same shortage as everyone else; a vendor that makes its own is not. That is a genuine supply-chain wall that Ciena and most others — who outsource much of their manufacturing — simply don't have. There's a real cost to doing it in North America: building there runs 70–150% more expensive than building in Asia. But in this window, Nokia is happy to trade margin for speed — spend more, ship now, take share — and optimize cost later once it has the customers. It's backing the bet with money, too: InP capacity is slated to expand tenfold.
Edge 3 — no tech gap, and the share map just flipped
A capacity story only works if the product is good enough, and here the honest read is that it is. There's no meaningful technology gap between Nokia and Ciena. The DCI standards that matter — 400G and 800G (gigabits per second per wavelength) — are well settled, so any product that meets the spec is interchangeable on the wire. Nokia has already shown a full 1.6-terabit DCI solution. Parity is enough; in a sold-out market, the tiebreaker isn't the spec sheet, it's delivery.
And the scoreboard is already moving:
| 2025 | 2026 | |
|---|---|---|
| Ciena | ~90% of DCI — a near-monopoly | capacity-capped, ceding the increment |
| Nokia | a minor player | >55% of Google's DCI tender; the year's main growth story |
The clearest signal came at the top line: in Q4 2025, Nokia's worldwide optical-transport revenue excluding China passed Ciena's for the first time. A market that was a near-monopoly a year ago now has a second structural winner — and it's the one with the spare factory.
Where the capacity actually sits
The whole thesis rests on physical lines in physical buildings, so it's worth being concrete about where they are:
| Site | Role | Scale / note |
|---|---|---|
| San Jose (Infinera core) | main DCI assembly | ~$2.5B/yr idle, up to ~$4B with overtime — the engine of the surge |
| Mexico | companion boards | a low-end line once slated to close, kept alive by data-center demand |
| San Jose (InP fab + DSP R&D) | in-house lasers + chips | the vertical-integration core; InP capacity planned to grow 10× |
| 德科立 (China partner) | co-development + early foundry | a Chinese optical contract manufacturer; supplied the early OEM orders |
That last row carries a wrinkle worth spelling out. Nokia co-developed its DCI solution with a Chinese contract-manufacturing partner (德科立), which exclusively supplied the early foundry orders — a way to lean on Asian cost and know-how. But for the big whole-box orders going into North America, two forces push the work back to San Jose: US–China trade friction, and hyperscaler customers who want a pure North American supplier on the paperwork. So the marquee orders get built on Infinera's San Jose lines, not shipped in from China — which is exactly why the idle US capacity is the asset it is.
The risks
None of this is free of catches, and the bear case writes itself from the same facts:
- The capacity dividend is finite. The whole edge is the idle line. Once that backlog is consumed, the scarcity premium fades and Nokia competes on more ordinary terms.
- North American cost. Building at home costs far more than Asia. The premium is fine while customers will pay for delivery; if competition heats up or buyers push on price, margins get squeezed.
- Integration risk. Absorbing Infinera — its people, culture and technology — is a multi-year job. Botch it and the very advantage being bought erodes from the inside.
The bottom line
The case for Nokia isn't that it built a better box. It's that, in the narrow golden window where DCI demand is exploding and the incumbent is sold out, Nokia grabbed the one resource nobody else has — idle North American capacity, backed by a supply chain it owns end-to-end — and converted it straight into the majority of the year's biggest orders. It isn't just a new entrant in the DCI market. Under these exact conditions, it's the single largest structural beneficiary of the boom.
An analysis piece: the figures here are drawn from research on the DCI supply situation and stated as the thesis, not independently re-derived.