Nokia (NOKIA.HE): The AI Infrastructure Cornered-Resource Hidden in a Finnish Telecom

In 2013, Finland's Nokia sold its phone business to Microsoft for 5.4 billion euros. Most observers called it a humbling end. They were wrong.

$NOKIA.HE is a 74.6 billion dollar Finnish infrastructure company sitting at the center of the AI buildout. The phones were always the visible piece. The infrastructure underneath them was always the more durable business. After divesting the visible piece, Nokia spent a decade quietly rebuilding the unglamorous one. The market is finally catching up.

What Nokia Actually Sells Today

Nokia builds what you never see but always rely on. The optical gear shooting data between hyperscaler campuses. The routers carrying global internet traffic between cities and continents. The 5G core software running calls and data sessions for the mobile carriers you actually pay every month. Its customers are AT&T, Verizon, Microsoft, Meta, Lockheed Martin. Not consumers.

In a world where every additional AI token requires a packet to travel from a user, to a data center, through a training cluster, and back, the demand for this kind of infrastructure is structurally accelerating. Nokia does not sell the GPUs. It sells the wires, glass, and software that connect everything that uses them.

Why Pricing Becomes Secondary

The single most important point about Nokia's carrier business is that pricing is secondary. An operator running Nokia's OLT (the optical line terminal at the heart of fiber networks) typically also runs Nokia ONTs (the customer-side units), Nokia Altiplano management software, and Nokia integrated backhaul.

Ripping out one layer means 12 to 18 months of re-qualification work, custom integration, regression testing, and field engineering. That is not a vendor relationship. That is a marriage. Even when a competitor quotes 15 percent below, the operator's effective switching cost is years of operational disruption. We covered the structure of switching costs in our explainer on what an economic moat actually is.

"The deepest moats are the ones where the customer's switching cost is measured in calendar quarters, not in dollars."

The Indium Phosphide Cornered Resource

Through its 2024 Infinera acquisition, Nokia owns its own indium phosphide (InP) fab in Sunnyvale, California. Indium phosphide is the material substrate of choice for high-speed coherent optical transceivers, the components that move data at multi-terabit-per-second rates between hyperscaler campuses. Ciena and most other rivals design their photonic chips and outsource fabrication to a small pool of foundries.

Nokia controls the wafer, the yield, and the cost curve. A second InP fab in San Jose is ramping in late 2026. That is what Hamilton Helmer would call a Cornered Resource. Not the only photonic capability in the world, but a structurally limited input where competitors depend on third-party foundries while Nokia owns its own.

Catalysts Stacking Through 2027

The catalyst calendar is unusually dense for a company that the market spent a decade ignoring:

• NVIDIA took a 3 percent stake. More important than the dollar amount, NVIDIA and Nokia co-built anyRAN, an AI-native mobile network architecture designed around inference workloads at the radio edge. A cap table that includes NVIDIA is a cap table that attracts the next wave of AI infrastructure customers.
• Second InP fab in San Jose ramps late 2026. Doubles the cornered-resource capacity exactly when hyperscaler optical demand is structurally accelerating.
• Lockheed Martin partnership across 50 percent or more of NATO nations. Sovereign-grade defense communications is a buyer category that pays a premium for trust, traceability, and non-Chinese supply chain. Nokia is one of the very few credible Western options.
• Hyperscaler capex revised from $540B to $700B+ for 2026. The ground beneath the entire thesis just shifted up by more than 30 percent in a single revision cycle.

Capital Returns: R&D, Then M&A, Then Us

Nokia held its dividend flat through the brutal 2022 to 2024 RAN (radio access network) downturn when peers were cutting, then grew it from 0.13 to 0.14 euros per share. Two buybacks executed: 600 million euros in 2024, then 900 million euros in 2025 specifically to neutralise the Infinera acquisition dilution. Roughly 1.5 billion euros returned in two years.

Management's capital priority is explicit and, in our view, correctly ordered: R&D first, then M&A, then us. For a deep-tech infrastructure business, you do not want a CFO who privileges shareholder returns over research investment. The technology curve is the moat. If R&D loses, the moat erodes, and no amount of buybacks rescues you.

The MoatMap Scorecard: Q49 V32 M74, StockRank 74

Here is the Nokia MoatMap StockRank:

• Quality: 49/100. Middling. The legacy carrier business drags the composite down. The infrastructure and Infinera contributions are still maturing.
• Value: 32/100. Stretched. P/E of 81.6x. The market is paying for the future, not the past.
• Momentum: 74/100. Up 36 percent in 30 days, near 52-week highs, RSI at 85. Strong but flashing technically overbought.
• Composite StockRank: 74/100. Buy on the composite. Momentum and the underlying structural story are doing the heavy lifting.

The asymmetry is the story. Quality is middling, value is stretched, momentum is screaming, and the underlying business is at the very early stage of being repriced from a slow-growth carrier-equipment vendor to a strategic AI infrastructure supplier with a cornered photonic resource. These factor profiles can resolve very differently depending on cycle timing. Our guide to factor investing covers how to think about this kind of mix in detail.

Too Late, or Just the Second Inning?

The honest question with any name up 36 percent in 30 days is whether you have already missed it. The base rate argument says yes, RSI 85 is exactly where retail piles in and the move ends. The structural argument says no, because the underlying capex revision (540 billion to 700 billion plus for 2026) only just happened, and Nokia's share of that flow is still being repriced into earnings estimates that are months behind reality.

The memory analogy is worth sitting with. Memory names looked stretched in September after a 2x. They have since 5x'd. The market often underestimates how long structural cycles run, particularly when the cycle is being driven by a category of capex (AI infrastructure) that did not exist on most analysts' models 18 months ago. We are not arguing the next 36 percent is automatic. We are arguing the second inning is a more defensible starting position than the ninth.

The Bottom Line

Nokia is a paradox. By legacy reputation, it is a slow-moving European telecom equipment vendor that lost the consumer phone war. By current cap table and capex exposure, it is one of the most strategically positioned AI infrastructure suppliers in the Western world, with a cornered resource in indium phosphide, a NVIDIA equity anchor, and a Lockheed Martin NATO partnership. The two stories are at war with each other inside the multiple, and the market is in the early stages of resolving the tension.

For investors using Nokia as a single-name AI infrastructure idea, our guide to reviewing your portfolio for weak spots is the right framework for thinking about position size, concentration, and how a momentum-led name fits with the rest of the book.

For the full breakdown including segment economics, Infinera integration risk, the carrier-versus-hyperscaler mix shift, and the management quality assessment, the Nokia Deep Dive is the place to go.

Disclosure: this article is for informational purposes only and is not investment advice.