Nokia Oyj (NYSE: NOK) — Priced for the AI Inflection It Hasn’t Booked Yet

Independent Fundamental Equity Research Company: Nokia Oyj (NYSE: NOK ADR; primary listing Nasdaq Helsinki: NOKIA.HE) Sector: Information Technology — Communications Equipment Report date: 2026-06-11 · Price referenced: $13.40 (ADR, 2026-06-10) · Financials: EUR; market data USD (EUR/USD ≈ 1.085)

⚡ Claude’s Take

This block is the author’s own subjective opinion and general information only — not investment advice. The analysis that follows it takes no position and carries no price target; it discusses valuation only as embedded expectations and scenarios.

Verdict: AVOID at $13.40 — a genuinely improving business at a demonstrably wrong price. Not a short. Fair-value zone ≈ $8–11; would turn constructive on weakness toward the high-single digits ($8–9), where you stop paying for the AI story and get the licensing annuity, the net cash, and the optical option for roughly free. Conviction: medium.

The tag: a telecom toaster wearing an AI halo. Nokia is a real operational turnaround — new ex-Intel CEO, a strategically smart Infinera acquisition into the one structurally growing niche (optical for AI data centers), a $1B Nvidia endorsement, a €2.4B AI-and-cloud order book growing 3x, and a high-quality patent-licensing annuity throwing off ~70% of segment profit. None of that is fake. The problem is entirely the price. At $13.40 the stock sits at the ~99th percentile of its own ten-year valuation on every metric, trades ~26% above the Wall Street consensus target of $10.63, and is being valued against the data-center comp set (Arista, Ciena) while ~85% of its revenue still lives in the structurally poor, sub-cost-of-capital telecom-equipment world it has inhabited for a decade (group ROE 3.1%, ROCE 4.3%). The market is paying roughly 30–45% of the equity value for an inflection that exists today as orders, not recognized, profitable revenue — only ~€350M of AI-and-cloud revenue was booked in Q1-2026 against a €1B quarterly order intake, a ~7:1 gap gated by an indium-phosphide fab ramp Nokia has never executed at scale.

What the market is pricing correctly: the order book is real and the strategic repositioning is credible. What it is pricing incorrectly: that a 3%-ROE business hitting its own 2028 plan (group “double-digit” operating-profit growth, network-infrastructure margins reaching only 13–17%) justifies a ~29x EV/EBITDA, ~30x-comparable-earnings multiple. It doesn’t — even management’s success case implies a flat-to-lower price once the multiple normalizes. This is a narrative-driven re-rating of a low-return business and a mean-reversion candidate, not a de-risked compounder. It is not a short, because the licensing crown jewel, €3.4B net cash, a 0.5 beta, low short interest, and genuine momentum make the downside slow and the borrow unrewarding — and because the bull case (optical scales, the multiple holds) is live, not absurd. The single fact that would flip me bullish: AI-and-cloud revenue converting to a €3–4B run-rate at stable-to-rising gross margin by FY2027, with group ROCE crossing ~10% — proof the inflection is earnings, not backlog. The single fact that would flip me bearish (toward a short): optical gross margins compressing as the industry’s indium-phosphide capacity floods in (the classic capital-cycle tell) while the order-to-revenue gap persists into 2027 — narrative reversal on a peak multiple.

1. Executive Summary

Nokia is two companies wearing one ticker. The first is a high-quality intellectual-property annuity — Nokia Technologies — that licenses 26,000+ patent families (8,000+ declared essential to 5G) to 250+ licensees including Apple, Samsung and Mercedes-Benz, earning a ~70% operating margin and generating roughly 70% of the group’s segment operating profit on just 7.5% of revenue, with >€800M of revenue contracted through 2030. The second is a structurally disadvantaged telecom-equipment business — mobile radio, optical, IP routing, fixed access and core software — that sells into a consolidating, capital-starved base of telco buyers with monopsony power, must spend ~24% of revenue on R&D just to stand still, and earns blended low-single-digit margins. The combined entity produced FY2025 revenue of €19,889M (down from €23.3B in 2019), reported IFRS operating profit of €885M (4.4% margin), net income of €651M, and a return on equity of 3.1% — below any reasonable cost of capital.

For most of the past decade this was a stagnant, fairly-valued workhorse. Over the trailing twelve months the stock has roughly tripled (52-week range $4.00–$17.45; +168–208%), driven by three events: (1) a new CEO, Justin Hotard, hired from Intel’s Data Center & AI group in April 2025; (2) the €2.5B acquisition of Infinera (closed February 2025), planting Nokia squarely in the optical-transport supply chain feeding AI data centers; and (3) a $1B equity investment by Nvidia (October 2025) tied to an AI-RAN partnership. The market has re-rated Nokia from a telecom-equipment multiple to a data-center-infrastructure multiple.

The evidence supports a genuine operational turnaround but not the valuation. Network Infrastructure orders from AI-and-cloud customers reached €2.4B in 2025 (3x year-over-year) and €1B in Q1-2026 alone; nine of the top ten hyperscalers use Nokia optical. But order intake is running ~7x ahead of recognized revenue, conversion is gated by a semiconductor fab ramp, and management’s own 2028 targets (NI operating margin of only 13–17%, group “double-digit” operating-profit growth, flat mobile) describe improvement, not the hyper-growth the multiple implies. A sum-of-the-parts that fully credits the licensing annuity, the optical optionality and the net cash lands at roughly $38–52B against a ~$74.8B market capitalization — meaning ~30–45% of today’s price is AI option value not yet in the numbers. On a three-year scenario set, the base case (management hitting its plan) still implies a lower price because the multiple does the damage; only the bull case clears today’s level, and the downside scenario is roughly twice the magnitude of the upside.

This report takes no position and sets no price target (the single exception is Claude’s Take, above). It argues that Nokia is a structurally low-return business with one durable franchise, executing a credible but unproven pivot, currently priced as though that pivot were already complete.

2. Business Overview

Nokia Oyj, headquartered in Espoo, Finland and founded in 1865, is one of the two Western “trusted vendor” suppliers of telecommunications network infrastructure (the other being Ericsson). It employs ~78,000 people and reports in euros. The FY2025 20-F (filed 2026-03-05) presents the company in four segments; effective January 1, 2026 Nokia collapsed these into two operating segments plus an exit bucket. Both lenses matter and are reconciled below.

2.1 The four-segment FY2025 structure (the basis of the audited accounts)

(Source: FY2025 20-F, “Results of segments.” Segment operating profits sum to ~€2.4B; group reported operating profit was €885M, the difference being Group Common/Other plus purchase-accounting and restructuring reconciling items — see the Financial Quality section.)

The single most important structural fact in the entire business: Nokia Technologies generates roughly 70% of segment operating profit on 7.5% of revenue. The patent-licensing annuity effectively subsidizes a large, low-return equipment business. To make this concrete: the three equipment/software segments together (NI + MN + CNS) produced ~€18.4B of revenue and ~€1.34B of segment operating profit — a blended ~7.3% margin — while Nokia Technologies produced ~€1.5B of revenue and ~€1.06B of operating profit. The licensing unit earns nearly as much operating profit as the entire rest of the company combined, on one-twelfth the revenue. Strip out licensing, and “hardware Nokia” earns a low-single-digit margin that, after the corporate-center costs and reconciling items that do not show in segment numbers, barely clears breakeven on an IFRS basis.

The composition within Network Infrastructure also matters: Optical (€3,019M, +85%) is the Infinera-inflated growth line; IP routing/switching (€2,594M, essentially flat) is the steady franchise where Nokia holds the #1 edge-routing position; and Fixed/PON (€2,373M, +3%) is the broadband-access business where Nokia has led for six consecutive years. Only Optical is genuinely accelerating, and its acceleration is half acquired.

2.2 The new two-segment structure (from January 1, 2026)

• Network Infrastructure (~€7.8B TTM, ~10% op margin): optical networks, IP networks, fixed networks. The growth engine; the home of the Infinera assets and the AI-data-center narrative.
• Mobile Infrastructure (~€11.6B TTM): combines Radio Networks (~€7.5B, ~36% gross margin), Core Software (the former CNS, ~€2.5B, ~49% gross margin) and Technology Standards (the licensing business, ~€1.4B contracted run-rate, ~€1.1B operating profit).
• Portfolio Businesses (~€0.9B, ~−€0.1B operating loss): an explicit exit bucket — Fixed Wireless Access (CPE), site/outside-plant services, enterprise campus edge, microwave radio. The Inseego/FWA divestiture (April 2026) is the first exit.

2.3 Customers, recurrence, and geography

Customer types span (1) telcos/communication service providers — historically the dominant buyer, but management targets falling below 60% of NI by 2028; (2) AI and cloud customers — hyperscalers and “neoclouds” (CoreWeave, Nscale named), the growth story; (3) mission-critical enterprise and defense (a new “Nokia Defense” unit); and (4) IP licensees. No single customer exceeded 10% of net sales in 2023–2025, though the 20-F warns concentration is rising as operators consolidate.

Revenue quality is bifurcated. The equipment business is project- and frame-agreement-based and lumpy — recognized on customer acceptance — and management has warned that AI/cloud orders are elongating and order flow will get “lumpier” into 2027. The only genuinely recurring, high-quality streams are Nokia Technologies licensing (contracted through 2030) and a growing core-software/SaaS base (125+ 5G standalone contracts, 25+ Core-as-a-Service customers).

Geographically (by customer country), the United States is the largest market at €5,871M (~30% of sales); Finland (€1,615M, 8%) is inflated because all licensing revenue books to Finland; India is €1,528M (8%). Post-Infinera, ~58% of non-current assets are now US-located — a deliberate Americas / AI-cloud pivot. This US concentration is a double-edged sword for the thesis: it positions Nokia in the deepest pool of hyperscaler capex (a tailwind) and aligns it with the trusted-vendor/CHIPS-Act onshoring agenda (a policy tailwind), but it also concentrates exposure to US carrier consolidation (three-way wireless market with hard-nosed procurement), to any US-Europe tariff friction, and to the FX translation drag of a EUR-reporting company earning ~30% of its revenue in dollars. The Infinera deal materially increased this US tilt, which is coherent with the strategy but raises the stakes on a single geography and a single demand driver (AI capex) that is itself cyclical.

Verdict: A complex, multi-segment hardware-and-IP conglomerate whose economic center of gravity (licensing) is hidden inside ~7.5% of its revenue. The headline “telecom-equipment company” framing materially misrepresents where the profit actually comes from — in both directions: the equipment is worse than the blended numbers suggest, and the licensing is better.

3. Industry Dynamics

Telecom network infrastructure is a structurally poor industry that looks like it should be good. On paper it is a tight oligopoly: global RAN/wireless is dominated by five players (Nokia, Ericsson, Huawei, ZTE, Samsung), and optical/IP adds Ciena, Cisco, Arista and Juniper (now part of HPE). Oligopolies are supposed to earn high returns. This one does not, for two structural reasons.

First, demand is set by a tiny, consolidating buyer base with monopsony power. A handful of large telcos buy most of the world’s equipment. They run joint procurement, share network assets, and hire consultants specifically to squeeze vendor pricing. The 20-F is blunt: “Competitive intensity remains high… creating a risk of persistent high price erosion,” and Nokia “may increasingly be required to agree to less favorable contractual terms.” This is the textbook condition for an absence of pricing power — a fragmented, price-taking supplier base selling to a concentrated, price-making buyer base.

Second, R&D intensity is ruinous and non-optional. Nokia spends €4.9B, or 24.4% of revenue, every year just to remain current across successive 3GPP generations (5G, 5G-Advanced, 6G) and optical roadmaps. Mobile Networks alone spends ~26.6% of its revenue on R&D to earn a 2.8% operating margin. This is a fixed cost that does not scale down; it is capital being consumed to defend position, not compounded.

The result is that profit pools are wildly skewed away from the hardware. Nokia’s own segment data is the cleanest possible illustration: the money is in IP licensing (~70% margin) and software (core, ~13% operating / ~50% gross), not in the radio/equipment business (2.8%). The carrier-capex equipment business is where capital goes to die; the toll booth on the standards is where it accumulates.

Where is this in the capital cycle (Marathon lens)? The industry splits into two very different neighborhoods:

• RAN / mobile = late-bust, flat trough. A multi-year carrier-capex downcycle; global 5G is only ~35% penetrated, but carriers are not earning their cost of capital and are deferring spend. Crucially, supply has not cleared — Huawei and ZTE remain dominant outside Western markets, and Samsung is taking Western share (notably winning RAN at AT&T from Nokia). Ericsson’s own management has framed this market as structurally “flattish/cyclical, not growing.” This is a no-pricing-power trough. Structurally negative.
• Optical / IP data-center = mid-boom. Hyperscaler capex is pulling a genuine wave of orders (Nokia booked €2.4B of AI-and-cloud orders in 2025, 3x year-over-year; €1B in Q1-2026 alone). But classic capital-cycle warning signs are flashing: a land-rush in indium-phosphide capacity (management speaks of scaling “100–1,000x”), with every vendor — Ciena, Coherent, Lumentum, Cisco, Arista, Nvidia/Mellanox — racing to add supply. The CEO openly concedes it is “a long-cycle investment for short-cycle demand.” High current returns are attracting a flood of capital; the capital cycle argues strongly against extrapolating today’s optical economics over a 3–5 year horizon. Attractive now, structurally suspect later.

The competitive set, with numbers. In RAN, Dell’Oro-style share rankings place Huawei first globally (~30%+ ex-North-America), Ericsson and Nokia battling for second/third in the West, Samsung rising (notably the AT&T win), and ZTE rounding out the Chinese share. Outside China and a handful of restricted markets, the Western buildout is effectively a three-horse race (Ericsson, Nokia, Samsung) for a flat pie — the definition of a margin-destructive standoff. In optical, the systems market is led by Huawei, Ciena and Nokia (now including Infinera), with Cisco and the Chinese vendors behind; in the components layer that the AI build is straining, Coherent, Lumentum and the Nokia/Infinera indium-phosphide assets are the supply-constrained chokepoints. In IP/data-center switching — the market Nokia is entering late — Arista and Cisco are entrenched incumbents with deep software moats (EOS, IOS-XR) and Nokia is, by its own admission, a challenger winning early design wins rather than displacing share. The takeaway: Nokia holds genuine #1/#2 positions in its legacy strongholds (fixed access, edge routing, optical systems) but is a new entrant in the highest-growth, highest-multiple data-center-switching pocket the bull case leans on.

China exclusion and Open RAN are the two policy/technology wildcards. The Western “trusted vendor” exclusion of Huawei/ZTE artificially shrinks the competitive set in NATO/Five-Eyes markets and underpins “Nokia Defense” — a real but policy-dependent tailwind (the EU “toolbox” has been “talked about for a long time, little implemented,” and a reversal is possible). Open RAN is double-edged: it commoditizes the radio (eroding incumbents’ lock-in) but Nokia is leaning in defensively (AT&T, Deutsche Telekom/Fujitsu trials) to avoid being disrupted — a margin-dilutive posture, not an offensive moat.

Verdict: structurally bad industry, with one good neighborhood and one toll booth that sits outside the industry’s economics entirely. Telecom/RAN equipment combines oligopolistic supply with monopsonistic, consolidating, capital-starved buyers, mandatory ~24% R&D, chronic price erosion, and ROIC at or below cost of capital. The optical/data-center adjacency is genuinely better today but is a mid-cycle boom drawing capital from every direction. Patent licensing — the only structurally attractive pool — is not really a “telecom equipment” business at all; it is a standards-essential IP franchise that happens to be owned by an equipment vendor.

4. Competitive Position

Applying Greenwald’s taxonomy (the only genuine competitive advantages are supply/cost, demand/captivity, and economies of scale tied to captivity) segment by segment yields a barbell: one durable franchise, one narrow moat, and two weak-to-absent ones.

Nokia Technologies — intangibles (patents/SEPs). The one real, durable moat. This is a genuine intangible-assets barrier to entry. Nokia holds 26,000+ patent families, 8,000+ declared essential to 5G, with ~70% of patents carrying >10 years of remaining life. The duration is the moat: standards-essential patents must be licensed on FRAND terms, and Nokia’s portfolio is irreproducible — you cannot buy or build 30+ years of cumulative standards R&D. A 70.6% operating margin and 250+ licensees (Apple, Samsung, Lenovo, Mercedes-Benz) confirm a true franchise; economic earnings power vastly exceeds asset value, and that gap is explained by an identifiable advantage (the Greenwald test for a real moat). Durability is high but not permanent. Two real risks: (a) periodic re-licensing and litigation cliffs — the 20-F devotes pages to the reality that renewals sometimes require litigation or arbitration and may or may not come on acceptable commercial terms; the lumpy >€400M FY2024 catch-up and the optical-22% FY2025 “decline” show how renewal timing whipsaws reported revenue; and (b) patents expire — but the moat is self-replenishing through 6G (8,000+ 6G families already declared) so long as Nokia stays at the standards table, which its €4.9B R&D buys. Verdict: durable franchise, the crown jewel, carrying a re-licensing-volatility tax.

Cloud & Network Services / Core Software — switching costs (narrow but genuine). Cloud-native core (packet core, subscriber data management, voice) is embedded in operator networks, where replacement is high-risk and slow — real switching costs. Nokia’s core is present in ~65% of 5G standalone networks, earns a ~50% gross / 13% operating margin, and the margin has been expanding on cost-out. Verdict: a modest, genuine switching-cost moat; the quietly good operating business.

Network Infrastructure (optical/IP/fixed) — weak scale, with a nascent vertical-integration cost call-option. The growth hope; moat unproven. Nokia holds real share positions: #1/#2 in fixed PON (six years running), #1 in IP edge routing, #2 in optical. But these are markets where the prized new customers — hyperscalers — have near-zero switching costs, deliberately design to multiple vendors, build their own silicon (and SONiC software), and are explicitly “customers, partners and potential competitors.” Scale without captivity is not a durable barrier. The one thing that could become a cost moat is Nokia’s vertical integration into indium-phosphide fabrication (Infinera’s San Jose fabs, CHIPS Act-supported, with its own DSPs and FP5 routing silicon) — if the fab ramp succeeds at acceptable yield, owning the photonic supply chain into an 800G-pluggable shortage is a genuine advantage. As of this report it is unproven, early, and a semiconductor-manufacturing risk Nokia has never run at this scale. Verdict: good positions in a crowded, commoditizing market with no proven durable moat; the indium-phosphide integration is a cost-advantage option, not yet a fact.

Mobile Networks (RAN) — structurally weak, eroding. Essentially no moat. No captivity (operators multi-source and force rip-and-replace), no scale advantage (three Western plus two Chinese players, Nokia ranked #3 and losing — it conceded RAN share at AT&T to Ericsson while Samsung rises). The 2.8% operating margin on €7.8B of revenue after €2B of R&D is the financial proof the moat does not exist. Management’s own stated strategy — to “commoditize hardware” and migrate value to software and general-purpose silicon — is an admission that the hardware franchise is gone. Verdict: crowded commodity, share-losing, sub-cost-of-capital — the clearest “bad business” in the portfolio.

Pressure-testing the “AI-RAN with Nvidia” claim — narrative, not yet revenue. This is the headline of the new strategy and must be marked, as of this date, as largely promotional. Evidence of substance: a real GTC demonstration, a first live RF call with T-Mobile US, and portability across Nvidia’s ARC platform. Evidence it is narrative: (a) zero disclosed AI-RAN revenue — every reference is a trial or research collaboration (SoftBank, KDDI, Indosat), with commercial production guided only to “late 2027” and 5G-based; (b) the underlying RAN market is flat for the foreseeable future on management’s own telling, so AI-RAN’s payoff is pinned to 6G (~2030); and © it is built on Nvidia’s portable CUDA stack, which is equally available to Ericsson and Samsung — failing the Greenwald “available to all confers advantage to none” test. The genuine strategic logic (shift RAN from custom ASIC to general-purpose silicon plus software, lifting margin and avoiding Intel-style ASIC misexecution) is sound but margin-defensive, not moat-creating. The Nvidia partnership is non-exclusive; its value is signaling and endorsement, not a proprietary lock. Treat AI-RAN as an option, not a thesis pillar.

Head-to-head with Ericsson — the cleanest read on the absence of a moat. Nokia and Ericsson are near-mirror-image businesses: both Western trusted-vendor RAN suppliers, both with large IP-licensing units, both selling into the same consolidated carrier base. Ericsson is the larger radio player (notably stronger in North America, where it won the AT&T Open RAN displacement at Nokia’s expense) and runs a tighter mobile margin. The market’s verdict on the pair is instructive: Ericsson trades at ~14.5x earnings and ~8x EV/EBITDA on flat-to-declining revenue, and is not re-rated. If Nokia had a genuine, defensible competitive advantage that Ericsson lacked, you would expect it to show up in relative margins or share gains — instead Nokia is the one losing RAN share. The ~2x valuation premium Nokia carries over its twin is therefore not a moat premium; it is an Infinera-plus-Nvidia narrative premium. The two companies’ near-identical economics, with the valuation gap appearing only after the AI repositioning, is itself strong evidence that the re-rating is about story, not structural advantage.

Applying the Greenwald market-share-stability and ROIC tests. Greenwald’s diagnostic for a real moat is (1) stable market shares over time and (2) persistently high returns on invested capital. Nokia fails both at the group level: shares are not stable (RAN share lost to Ericsson and Samsung; optical share gained via acquisition, not organically), and ROIC/ROCE (4.3%) is persistently below, not above, the cost of capital. The tests pass only for the licensing sub-business, where share is structurally fixed by patent ownership and returns are ~70% margins. This is the quantitative confirmation of the barbell: one Greenwald-qualifying franchise, surrounded by businesses that fail the tests.

Verdict: a barbell — one durable IP-licensing franchise bolted onto a structurally disadvantaged equipment business. Roughly 70% of the company’s economic value (operating profit) comes from the 7.5% of revenue that is patent licensing. The other ~92% of revenue earns blended low-single-digit margins in a price-eroding, monopsony-buyer industry. The optical/AI pivot is real demand but unproven economics in a mid-boom capital cycle; the AI-RAN/Nvidia story is, today, narrative awaiting revenue.

5. Growth History and Forward Opportunities

History: six years of erosion, masked in 2025 by an acquisition. Group revenue fell from €23.3B (2019) to €19.9B (2025). FY2025’s +3.5% headline is almost entirely inorganic — Infinera added ~€1,273M; FX was a ~4% headwind; organic constant-currency revenue was roughly flat. This is a no-growth base ex-acquisition. The composition tells the story: Network Infrastructure grew 23% (Infinera), but its largest organic line, IP, was flat (+0.4%); Mobile Networks shrank 4%; Cloud & Network Services grew 1%; and Nokia Technologies fell 22% purely on the absence of a >€400M FY2024 catch-up payment (the underlying run-rate is steadier).

The forward story is bifurcated. Roughly 60% of the company (Mobile Infrastructure) is a flat-to-declining harvest business. Management explicitly says of mobile: “our focus is not on making the business necessarily a growth business because the underlying market is not growing, but to make it one that’s much more profitable.” Half the company is, by its own design, ex-growth — managed for margin, not revenue.

The other ~40% (Network Infrastructure) is the entire bull case, and the demand signal is genuine:

• AI-and-cloud orders of €2.4B in FY2025 (3x year-over-year) and €1B in Q1-2026 alone (a ~3x book-to-bill in that category);
• Optical Networks grew 17% in Q4-2025 and 20% in Q1-2026, with AI-and-cloud reaching ~30% of optical;
• Nine of the top ten hyperscalers use Nokia optical; named emerging customers include CoreWeave and Nscale (Nokia co-invested in Nscale alongside Nvidia); Microsoft is a SONiC switching customer;
• Management twice raised the addressable AI-and-cloud market CAGR — from 16% at the November Capital Markets Day to 27% by Q1-2026 — lifting the implied NI-market CAGR from 9% to 14%.

But order-to-revenue conversion is the crux, and it is back-loaded. Against €2.4B of FY2025 plus €1B of Q1-2026 orders, only ~€350M of AI-and-cloud revenue was recognized in Q1-2026 — a ~7:1 gap. Optical lead times are 12–18 months; orders are “elongating” into 2027; IP grew only 3% (design wins are “in pipeline, not in orders”); switching is early. The binding constraint is explicitly supply, not demand (“if we had additional supply, we could probably fulfill that”), gated by the indium-phosphide fab ramp (San Jose Fab 2, ramping late-2026, material in 2027). The growth is real but the market is paying today for 2027–28 revenue that has not converted, and conversion depends on Nokia executing a semiconductor ramp historically outside its competency.

The conversion mechanics deserve a closer look, because they are where the bull and bear cases physically meet. An order in Nokia’s optical/IP business is recognized as revenue only on shipment and customer acceptance, 12–18 months after booking. So the €2.4B of FY2025 AI/cloud orders and €1B of Q1-2026 orders are, mechanically, a 2026–2027 revenue event, not a 2025–2026 one. That is not inherently bearish — a growing backlog with a lag is exactly what you’d expect from a genuine ramp. The bearish wrinkles are three: (1) the lag is lengthening (“orders elongating into 2027”), which pushes the payoff further out and raises the risk of cancellation or repricing in the interim; (2) the binding constraint is Nokia’s own supply (the indium-phosphide fab ramp), so conversion depends on manufacturing execution, not just demand; and (3) management has explicitly guided that the 2028 NI operating margin tops out at 17%, meaning even successful conversion drops a limited amount to the bottom line — the revenue scales faster than the profit. An investor paying a data-center multiple is implicitly assuming both that the backlog converts on schedule and that it converts at a margin management is not promising.

Nokia Technologies remains a high-quality but flat annuity — ~€1.4B contracted run-rate, 6G standardization extending duration, with growth optionality in automotive, consumer electronics, IoT and multimedia (a >€200M expansion run-rate). It is a profit engine, not a growth engine. The one underappreciated forward lever here is the expansion categories: as connectivity proliferates into cars, wearables and industrial IoT, the universe of devices that must license Nokia’s cellular SEPs widens, and each new vertical (automotive has been the most active) is a multi-year negotiation that, once signed, adds durable recurring revenue. But these are measured in tens-to-low-hundreds of millions, not the billions that would move the group needle — useful ballast, not a growth engine.

Management’s quantified targets (hypotheses, not evidence):

• 2026 guidance: comparable operating profit €2.0–2.5B (tracking “somewhat above midpoint”); FCF conversion 65–75%; NI growth raised to 12–14% (from 6–8%); capex €900M–1B; USD/EUR assumption 1.18.
• 2028 Capital Markets Day targets: group “double-digit operating-profit growth” (2025–28 CAGR); NI revenue CAGR 6–8%; NI operating margin 13–17% (from 9.5%); Mobile Infrastructure operating profit growing from a €1.5B base on roughly flat net sales; Optical Networks reaching a double-digit operating margin.

Verdict: low-quality growth today, bifurcated, with a credible but unproven path to higher quality by 2027–28. Roughly 40% of the company has a real, accelerating, structurally supported growth vector but unproven margin leverage and supply-gated, back-loaded revenue; ~60% is flat-to-declining harvest. Headline growth is heavily acquired and order-led rather than revenue-realized. Critically, the 2028 NI margin target tops out at 17% — implying surging revenue is being reinvested, not dropped to profit — and 2026 operating-profit guidance (€2.0–2.5B) is barely above 2025’s ~€2.0B comparable despite the optical boom. This is order-quality, not earnings-quality, growth.

6. Financial Quality

The seven-year arc. The single most important context for Nokia’s financials is that this is a business that has shrunk through the entire 5G build-out — the very capex supercycle that was supposed to be its golden era:

The pattern is damning for any structural-growth thesis: revenue peaked in 2019/2022 and has declined ~15% to 2025; the FY2022 net-income spike (€4,250M) is a tax artifact (the €2.5B Finnish DTA re-recognition), not operations; FY2020 was a €2.5B loss (impairments); and operating profit has oscillated between €485M and €2,444M with no durable upward trend. R&D has risen in absolute terms even as revenue fell — the defining squeeze of this industry. Six years spanning the entire 5G cycle produced no revenue growth and no margin progression. That is the base rate the AI/optical narrative must overcome.

Revenue and margins. FY2025 revenue €19,889M (+3.5%, almost all Infinera). Gross margin 43.5%; reported IFRS operating margin 4.4%. The segment margins (NI 9.8%, MN 2.8%, CNS 13.0%, Technologies ~70.6%) show a business whose blended economics are dominated by the mix between a tiny high-margin annuity and a large low-margin equipment base. Note the direction of travel within segments in 2025: Network Infrastructure’s margin fell from 11.7% to 9.8% despite 23% revenue growth (Infinera dilution and FX), and Mobile Networks’ margin halved from 5.5% to 2.8% — so even the “good” 2025 revenue print arrived with deteriorating unit economics in the two largest segments.

The central quality-of-earnings issue — the comparable-vs-reported gap. Nokia guides, reports, and compensates management on a non-IFRS “comparable” operating profit. The bridge for FY2025:

The €1,242M gap is 57% of comparable operating profit being excluded — an unusually heavy adjustment load. And much of what is excluded is recurring: the €478M restructuring charge belongs to a program running since October 2023 (with another ~€450M of cash restructuring guided for 2026), and the €444M of acquired-intangible amortization will persist for years following Infinera. A skeptical reading treats a large portion of these “exclusions” as the ongoing cost of running a serial restructurer/acquirer. The same gap appears quarterly: Q1-2026 comparable operating profit was €281M (6.2% margin, genuinely improved), but reported IFRS operating profit was a fraction of that after restructuring and amortization. The economic earnings power of this business is far closer to the €651M net income / 3.1% ROE / 4.3% ROCE reality than to the €2.0B “comparable” headline.

Returns are structurally poor. ROE 3.1% (2025), ROCE 4.3% — below cost of capital. The equity base (~€20.8B) is itself inflated by tax accounting: Nokia derecognized €2.9B of Finnish deferred tax assets in 2020 and re-recognized €2.5B in 2022, a non-cash gain that produced the FY2022 €4,250M net income spike and bloated book equity ever since (dragging the ROE denominator). The 20-F explicitly warns the DTA “may be materially reduced” if Finnish taxable income disappoints. Reported equity and ROE are both distorted by tax accounting, not operations.

Other quality flags: (1) a 2025 reclassification of venture-fund fair-value gains/losses out of operating profit into financial income — convenient timing that removes volatility from the operating line exactly as that metric becomes central to the new strategy; (2) rising stock-based compensation — €337M in 2025, up from €239M (2024) and €201M (2023), +41% year-over-year, a real and growing dilutive cost the comparable metric does not strip out; (3) lumpy licensing catch-up payments (>€400M in FY2024) that flatter or depress year-over-year comparisons.

Cash generation and the balance sheet are the genuine strengths. FY2025 free cash flow was €1,465M (operating cash flow €2,071M; capex a capital-light €606M), a ~72% comparable-FCF conversion. The balance sheet carries net cash of €3,378M (total cash and interest-bearing investments €6,791M vs. €3,413M debt; ~56% equity ratio) — no liquidity risk, and optionality to keep investing through a downturn. The caveats: capex is stepping up sharply to €900M–1B in 2026 (optical/InP fabs), pulling conversion toward 65–75%; and net cash fell €1.5B in 2025 (Infinera cash, the Shanghai Bell buy-in, buybacks and dividends, partly offset by the €859M Nvidia inflow).

Verdict: economics do not improve with scale — arguably the opposite. Network Infrastructure added 23% of revenue (via Infinera) yet its margin fell ~190bps; Mobile Networks shrank and earns under 3%. Returns sit below cost of capital. Nokia is a strong-balance-sheet, genuinely cash-generative, but low-return collection of mediocre equipment businesses cross-subsidized by a high-margin patent-licensing annuity. The only credible path to scale economics is the AI/optical ramp delivering the 13–17% NI margin target by 2028 — entirely unproven and front-loaded with investment.

7. Capital Allocation

Management’s stated framework, in priority order: (1) organic R&D, (2) bolt-on / minority M&A, (3) a stable and growing dividend, (4) buybacks only with excess cash. The record is mixed, leaning cautious.

M&A. Infinera (closed February 2025, total consideration €2,496M = €1,066M cash + €785M assumed/settled convertible notes + 127.4M new Nokia ADSs (€584M) + €61M replacement awards) is the most consequential move — it created the AI/optical thesis by buying into the one structurally growing niche. The purchase-price allocation booked €833M of goodwill (~33% of price, the “workforce and synergies” soft kind), but integration is running ahead of plan: the synergy target (€200M) was raised in ambition and accelerated 9–12 months, with the mix flipping to two-thirds COGS as revenue outran expectations. The price looks defensible given the demand inflection, and Nokia used stock when its shares were cheaper — reasonable. The caveat is Nokia’s own M&A history: Alcatel-Lucent (2016, ~€15.6B) remains the value-destruction benchmark, and management itself concedes systems integration issues that “go back to the Alcatel acquisition” and that “one acquisition does not a trend make.” The Nokia Shanghai Bell buy-in (€501M for the remaining 50%) sensibly removes a minority drag. The Inseego transaction (April 2026) is a divestiture — Nokia is selling its sub-scale Fixed Wireless Access business as the first Portfolio-Businesses exit — correct pruning of a margin-dilutive line.

R&D at €4.9B (~24.4% of sales) is the stated #1 priority and structurally necessary; the issue is not the level but that this much spend produces sub-5% ROCE — it defends position more than it compounds value. Consider the arithmetic of the last six years: Nokia spent roughly €26B of cumulative R&D between 2019 and 2025, during which revenue fell €3.4B and operating profit went nowhere on a normalized basis. That is the single most important capital-allocation fact about the company — its largest and most-defended use of capital has, over a full technology cycle, failed to grow the business or lift returns. The bull rebuttal is that this R&D bought the patent portfolio (real) and the optical/AI position (partly acquired, not built), and that the next cycle (6G, AI-RAN, optical) will convert R&D into growth where 5G did not. That is possible, but it is a hope, and it is asking investors to extrapolate the opposite of the realized track record. The honest framing: Nokia’s R&D is a toll it must pay to remain in the game, not an engine that compounds shareholder capital — and a business whose mandatory reinvestment earns sub-cost-of-capital returns is, in Greenwald’s terms, destroying value at the margin even as it stays technologically relevant.

Buybacks are the weakest point. Nokia repurchased €624M in 2025 (€680M in 2024; ~€1.6B since 2023). In principle this ranks last in the framework. In practice the optics are poor: Nokia simultaneously issued 127.4M shares for Infinera and 166.4M shares to Nvidia while buying back stock — net share count rose — and it is repurchasing low-return equity at a ~99th-percentile valuation, which is, on its face, value-destructive timing. The buyback reads as dilution-offset/EPS management, not opportunistic value capture.

Dividend is the cleanest leg: €759M paid in 2025, a proposed €0.14/share for FY2025, ~€2.1B since 2023, covered by FCF and sustainable; policy is “recurring, stable, over time growing.”

Incentive alignment carries a flag. CEO Justin Hotard’s long-term incentive metrics are comparable operating profit (constant currency) and a cash-flow measure — i.e., management is paid on the very number that excludes the €478M restructuring and €444M amortization that represent the recurring cost of its own strategy, rather than on IFRS profit, ROIC, or total shareholder return. That is a misalignment worth watching, partly offset by Hotard’s modest personal co-investment (€2.8M with 2:1 matching).

Verdict: not yet an “intelligent allocator” on the evidence. Positives: a disciplined dividend, a strategically sound and (so far) well-integrated Infinera buy, correct portfolio pruning, and capital-light capex. Negatives: buybacks at a peak valuation while net-diluting, rising SBC, incentive metrics tied to the adjustment-heavy comparable measure, and a corporate history that earns Nokia M&A skepticism by default. The ultimate scorecard is a decade of sub-cost-of-capital returns: capital has been allocated to defend a low-return position, not to compound value. Hotard’s tenure is too short to re-rate the verdict; he gets the benefit of the doubt on Infinera and none yet on the buyback.

8. Changes and Headwinds — Last Two Years

The transformation (net thesis-strengthening on strategy, unproven on delivery):

• CEO transition. Pekka Lundmark → Justin Hotard (ex-head of Intel Data Center & AI), announced February 2025, started April 2025, with a near-wholesale leadership refresh (CTO Pallavi Mahajan ex-Intel; David Heard, ex-Infinera CEO, running Network Infrastructure; a new CCO and CPO). This is an “outside-in” reorientation toward AI/data-center.
• Infinera acquisition (closed February 2025, ~€2.5B) — created the optical/AI thesis; integration tracking ahead of plan.
• Nvidia $1B investment (October 2025) — 166.4M shares at $6.01 (€5.16), ~2.9% stake, non-exclusive partnership; proceeds for general corporate purposes (not earmarked to buy Nvidia GPUs). Signal value exceeds near-term financial value; it re-rated the stock.
• Two-segment reorganization (January 1, 2026) — NI + Mobile Infrastructure + Portfolio Businesses exit bucket; Group Common targeted down from ~€370M to ~€150M by 2028.
• Nokia Shanghai Bell buy-in (€501M, Q4-2025) and the Inseego/FWA divestiture (first Portfolio exit, April 2026).
• Restructuring — the October-2023 program raised to ~€1.2B gross savings (2023–26), ~14,000 headcount reduction, with ~€450M of further cash outflow guided for 2026.
• Capital returns — dividend and buyback continuity, supported by €3.4B net cash.

Headwinds:

• Order-to-revenue conversion gap (~7:1 in AI/cloud) and order “lumpiness” elongating into 2027.
• Supply constraints — indium-phosphide capacity ramp, TSMC leading-edge node access, and memory-price inflation across routers/CPE/radios/servers (a gross-margin risk, only partly passed through).
• Fab-ramp execution risk — San Jose Fab 2 (late-2026), with larger photonic die sizes harder to yield; peers have stumbled here.
• RAN downcycle and the AT&T loss — flat mobile market, North America RAN headwinds in 2026.
• FX — a ~4% FY2025 revenue headwind; a €0.02 USD/EUR move is ~€50M of operating profit (about half hedged).
• Tariffs — US/Finland exposure not directly quantified; US fabs and CHIPS Act support provide partial mitigation. (Open question.)
• Restructuring cash (~€450M in 2026) and rising SBC.

On the Nvidia investment specifically — separate the signal from the substance. The $1B is, financially, a rounding error for both parties and was dilutive to Nokia (166.4M new shares at $6.01, below where the stock now trades). Its value was never the cash; it was the endorsement — Nvidia, the most important company in AI infrastructure, publicly anointing Nokia as an AI-RAN and data-center partner, which is precisely what re-rated the stock from a telecom to a data-center multiple. But the partnership is non-exclusive: Nvidia is free to partner with Ericsson, Samsung, or anyone else, and the underlying technology (a portable CUDA/ARC software stack running RAN workloads on general-purpose GPUs) is available to all comers. So Nvidia did not hand Nokia a proprietary asset; it handed Nokia a head start and a halo. The substance — a first live RF call with T-Mobile, lab trials, commercial production guided only to “late 2027” — is real but early and unmonetized. An investor should value the Nvidia relationship as improving Nokia’s odds in a future (6G-era) market, not as a present-day revenue or moat event. The gap between the $40B+ of market value the partnership helped create and the ~zero revenue it has produced is the cleanest single measure of how much of Nokia’s price is narrative.

Verdict: the changes strengthen the narrative and the strategic positioning materially, but the financial proof — order conversion, fab ramp, and margin gearing — is a 2027–28 event the market is already paying for. The NI guidance raise (6–8% → 12–14%) is real evidence the order book is beginning to convert; but the 2026 operating-profit guidance barely above 2025, the capped 2028 NI margin target, and the cash demands of restructuring and capex confirm that operating leverage is deferred, not imminent.

9. Risk Analysis

Catastrophic-loss risk is low. A business with net cash, a licensing annuity, a 0.5 beta and diversified revenue does not face a plausible path to a total or near-total loss. The dominant risk is valuation, not solvency: a 99th-percentile multiple on a 3%-ROE base mean-reverts, and the de-rating alone drives the bear case. This is a “permanent capital impairment via multiple compression” risk, not a “zero” risk.

10. Valuation Discussion (Embedded Expectations)

No price target and no recommendation appear in this section (or anywhere in the analysis below); the analysis is confined to what the market is underwriting.

The setup. At $13.40 Nokia trades at the ~99th percentile of its own ten-year valuation on P/E (98.5), P/B (99.1) and P/S (99.1) simultaneously — it has essentially never been more expensive relative to its own history — while earning 3.1% ROE / 4.3% ROCE and trading ~26% above the Wall Street consensus target of $10.63. Peak-decile multiple, sub-cost-of-capital returns, and a price above sell-side fair value form the central tension.

The “expensive vs. cheap” debate is entirely an earnings-definition argument:

The bull says “~30x comparable earnings for an AI-infrastructure inflection”; the bear says “~84–106x reported earnings for a 3%-ROE telecom-equipment vendor.” Both are arithmetically correct. The honest read: even on the most generous comparable basis, ~29–30x is a growth multiple on a business guiding to 6–8% NI revenue growth and flat mobile. Nokia’s own 2028 targets imply group operating profit of roughly €2.7–3.2B (a low-double-digit CAGR from ~€2.0B) — genuine improvement, but it justifies the multiple only if the multiple holds while profit compounds.

Embedded-expectations bridge. To merely hold $13.40 over three years (earn ~9%), the market needs group operating profit near the top of the €2.7–3.2B range and retention of a ~25–29x EV/EBITDA multiple. But if profit reaches ~€3.2B (~€3.8B EBITDA) and the multiple mean-reverts to even a still-premium 15x, EV ≈ €57B + ~€3.4B net cash ≈ €60B ≈ ~$65B — below today’s ~$74.8B. In other words, the current price already discounts near-flawless execution of the 2028 plan and no multiple compression.

Sum-of-the-parts (EUR; the cleaner lens — value the annuity separately from the iron):

The SOTP lands at roughly $38–52B against a ~$74.8B market cap. The ~$23–37B gap is the AI option value — the market is paying roughly 30–45% of the equity value for an inflection not yet in the financials. Fundamentals support roughly half to two-thirds of the price; the balance is the supercycle narrative.

Peer cross-check. Nokia (83.75x P/E, 28.6x EV/EBITDA, 3.74x P/S) trades above its closest twin Ericsson (14.5x P/E, ~8x EV/EBITDA) despite near-identical end markets and Ericsson’s larger US radio share. Nokia’s multiple is being set by the data-center comp set (Arista ~42x EV/EBITDA; Ciena ~79x) rather than the radio comp set. The variant question reduces to: which comp set does Nokia belong in? Today it is priced in the expensive one while ~85% of revenue still lives in the cheap one.

Three-year scenarios (assumption-heavy):

The asymmetry is unfavorable from $13.40: even the base case — management hitting its plan — implies a lower price, because the multiple does the damage; only the bull case (peak execution and a sustained data-center multiple) clears the current level, and the downside scenario is roughly twice the magnitude of the upside.

A note on the dividend and the “yield support” argument. Some bulls anchor on Nokia’s ~1.2–1.4% dividend yield and net-cash balance sheet as downside protection. This is weak support at $13.40: the yield is below the broad market, the payout (~€759M) consumes roughly half of free cash flow that is about to be pressured by a capex step-up, and a 1.2% yield provides negligible valuation floor against a multiple that could compress 40%+. The genuine downside protection is the licensing annuity plus net cash — together worth perhaps €16–20B, or ~$4–5 per share of “hard” value — not the dividend. That floor is real but sits far below the current price.

Why the multiple is the whole game. In all three scenarios, the dominant driver of the outcome is not revenue or even operating profit — it is the exit multiple. Holding the base-case ~€2.9B of 2028 operating profit constant, the implied equity value swings from ~$40B at a 9x EV/EBITDA (radio comp) to ~$95B at a 24x EV/EBITDA (data-center comp). That ~$55B swing — roughly three-quarters of today’s market cap — turns entirely on a classification question: does the market continue to believe Nokia is a data-center company, or does it revert to seeing a telecom-equipment company with an optical option? History, the 7-year financial arc, and Ericsson’s persistent ~8x multiple all argue for reversion; the order book and the Nvidia halo argue against it. An investor at $13.40 is, whether they realize it or not, primarily making a bet on multiple persistence, not on business fundamentals.

What the market is underwriting correctly vs. incorrectly. Correctly: the order book is real, the strategic pivot is credible, the balance sheet is strong, and the licensing annuity is high-quality. Incorrectly (or at least aggressively): that a structurally low-return business hitting its own targets justifies a data-center multiple ~2x its identical twin, with the price already above consensus fair value at the top of its historical valuation range.

11. Variant Perception

Consensus belief is bifurcated. The sell-side target ($10.63) sits below the market price — fundamental analysts think the stock has run ahead of fundamentals — yet the market has bid it ~26% above that target on momentum and the Nvidia/AI-RAN narrative. Fundamental analysts are cautious; thematic/momentum money is in control. The marginal buyer is underwriting an AI-infrastructure inflection, not a 3%-ROE legacy telco.

The strongest bull case: Nokia is a genuine AI-infrastructure inflection mispriced by a stale “telecom-equipment” frame. Nine of ten hyperscalers use its optics; €2.4B AI/cloud orders (3x), €1B in Q1-2026; Infinera brings 800G coherent optics and US indium-phosphide vertical integration; Nvidia’s $1B stake and AI-RAN partnership validate the platform; a credible new ex-Intel CEO is executing a cost and portfolio reset. If NI compounds at 6–8% into a growing optical market with margins reaching 13–17%, and the order book converts, group operating profit hits ~€3.2B and the data-center multiple is deserved.

The strongest bear case: this is a narrative-driven re-rating of a structurally low-return business and a textbook capital-cycle setup — a commoditizing optical-hardware market where high returns are pulling in a flood of indium-phosphide capacity that will mean-revert margins. ~70% of segment profit comes from a 7.5%-of-revenue licensing annuity facing renewal cliffs; the actual equipment business earns low-single-digit margins and 3% ROE. The ~7:1 order-to-revenue gap means the “growth” is a backlog promise, not realized economics, and management itself flags supply as the binding constraint. At the 99th valuation percentile, above consensus fair value, the asymmetry points down.

The 3–5 assumptions that matter most:

1. Does the AI-and-cloud order book convert to recognized, profitable revenue? (The bull’s whole case; only ~€350M recognized so far.)
2. Does optical avoid commoditization as the industry adds indium-phosphide capacity? (The Marathon test.)
3. Does the data-center multiple persist, or mean-revert toward the radio comp (Ericsson ~8x EV/EBITDA)? (The single biggest price driver.)
4. Does Nokia Technologies renew its post-2030 licensing cliffs at current economics? (Roughly half the profit.)
5. Can the new team lift group ROE/ROCE above cost of capital? (4.3% ROCE today — the thesis fails if this doesn’t move.)

Falsification tests. Bull falsified if: by mid-2027, AI/cloud quarterly revenue is still a small fraction of the order book (conversion stalls), OR NI operating margin fails to climb out of ~10% toward the 13–17% target, OR optical gross margin compresses as capacity floods in. Bear falsified if: AI/cloud revenue ramps to a €3–4B run-rate with stable/expanding gross margin by FY2027, group ROCE crosses ~10%, and NI margin hits the mid-teens — proving the inflection is real and the multiple earned.

Resolution. The tension is not whether Nokia is improving — it demonstrably is. It is whether a demonstrable operational turnaround at a 3% ROE base justifies a 99th-percentile, above-consensus, data-center multiple. The fundamentals support roughly half to two-thirds of the price; the balance is AI option value riding on a backlog not yet converted. This sits closer to a narrative-driven re-rating of a low-return business (mean-reversion candidate) than to a fully de-risked inflection — the inflection is possible and partly underway, but the price already pays for its success.

12. Fact vs. Interpretation

13. Open Questions

1. Order-to-revenue conversion: What is the true incremental gross margin of AI/cloud optical and IP revenue, and how fast does the €3.4B+ order book convert? (Only ~€350M recognized in Q1-2026.)
2. Licensing renewal calendar: What are the post-2030 re-licensing economics, and what is the litigation/arbitration calendar for major licensees? (Roughly half of profit.)
3. Mobile Networks isolated economics: What does the radio business earn excluding any licensing allocation — does it ever clear cost of capital, or is it a permanent drag / divestiture candidate?
4. Fab execution: Can Nokia ramp indium-phosphide capacity (San Jose Fab 2) at acceptable yield — a semiconductor competency it has never demonstrated at scale?
5. AI-RAN monetization: When does AI-RAN generate >€100M of revenue, and at what margin, given it relies on Nvidia’s non-proprietary CUDA stack?
6. Tariffs: What is the quantified US/Finland tariff exposure, net of CHIPS-Act-supported onshoring?
7. DTA durability: Will Finnish taxable income support the €2.5B deferred tax asset, or is an impairment (and equity hit) likely?

14. What Must Be True

For the bull case to be right (the inflection is real and the multiple earned):

• Network Infrastructure must convert its order book into recognized revenue at a €3–4B AI/cloud run-rate by FY2027 with stable-to-rising gross margin, and NI operating margin must climb from ~10% toward the 13–17% target.
• Group ROCE must cross ~10% — proving the new team has lifted returns above cost of capital for the first time in a decade.
• The optical market must avoid margin compression despite the industry-wide indium-phosphide capacity build.
• Falsification test: if, by mid-2027, AI/cloud quarterly revenue remains a small fraction of the order book, OR NI margin is stuck near 10%, OR optical gross margin is compressing, the bull thesis is broken.

For the bear case to be right (a narrative re-rating that mean-reverts):

• The multiple must compress from ~29x EV/EBITDA toward the radio-comp range (Ericsson ~8x) as the market re-classifies Nokia back toward “telecom equipment,” OR optical margins must compress as capacity floods in, OR a licensing renewal cliff must hit reported revenue.
• Group returns must remain sub-cost-of-capital (ROCE near ~4–5%), confirming the equipment business cannot scale economically.
• Falsification test: if AI/cloud revenue ramps to a €3–4B run-rate at stable/expanding margin by FY2027, group ROCE crosses ~10%, and the NI margin hits the mid-teens, the bear thesis is broken and the re-rating is validated.

The two falsification tests are mirror images, which is the point: the entire debate resolves on a single observable over the next 18–24 months — whether the AI/cloud order book converts into profitable, margin-accretive revenue, or remains a backlog promise on a peak multiple.

15. Source Appendix

See the Source Appendix below for the full source list with URLs and access dates. Primary sources: Nokia FY2025 Form 20-F (filed 2026-03-05, SEC CIK 0000924613); Nokia Q4-2025/FY2025 and Q1-2026 earnings releases and transcripts; Nokia Capital Markets Day (November 19, 2025); Nokia Special Call (October 28, 2025); J.P. Morgan Global Technology Conference presentation (May 19, 2026). Quantitative cross-checks: SEC EDGAR XBRL; aggregated market-data feeds; Yahoo Finance (2026-06-11). Peer data: Ericsson, Cisco, Arista, Ciena. All management commentary is treated as hypothesis and validated against filings and external evidence.

This is independent fundamental research for general information only. The analysis contains no buy/sell recommendation and no price target; valuation is discussed only as embedded expectations and scenarios. The sole exception is the labeled “Claude’s Take” block at the top, which is the author’s own subjective opinion. This document is not investment advice. The author holds no position in the security and nothing herein is a solicitation to buy or sell.

APPENDIX A — Standard Diligence Questionnaire — Nokia Oyj (NYSE: NOK)

Supplemental to the analysis above. Answers are grounded in the FY2025 20-F, FY2025/Q1-2026 earnings, the November 2025 Capital Markets Day, and quantitative data. Fact / Interpretation / Assumption labels applied where material.

General

What thoughtful questions have other investors asked about this company? The dominant investor debates: (1) Is the AI/data-center order book real revenue or a backlog mirage? — the ~7:1 order-to-recognized-revenue gap is the single most-asked question (Fact). (2) Which comp set does Nokia belong in — telecom radio (Ericsson ~8x EV/EBITDA) or data-center (Arista/Ciena 40–80x)? (3) Is Network Infrastructure margin gearing real, given the 2028 target tops out at only 13–17%? — analysts at the CMD pressed management that surging revenue implies “no dramatic gearing.” (4) Is the Nvidia partnership strategically meaningful or a non-exclusive endorsement? (5) Can Nokia, of all companies, execute a semiconductor (indium-phosphide) fab ramp? (Interpretation: these are the right questions; the bull/bear split turns on #1 and #3.)

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Mixed. Reported IFRS earnings are depressed (3.1% ROE, weighed by restructuring and amortization), but the valuation is at a cyclical/all-time high (~99th percentile). Optical is at a cyclical demand high; mobile/RAN is in a multi-year capex trough (Fact). (Interpretation: earnings are low, the multiple is peak — an unusual and unfavorable combination.)

Driven by external environment or internal actions? Both. Externally: AI/hyperscaler capex (tailwind) and the carrier-capex downcycle (headwind). Internally: the Infinera acquisition, restructuring (~14,000 headcount out), and the portfolio reorganization (Fact).

How stable are revenues? Moderately stable in aggregate (~€20B band for years) but lumpy within segments — equipment is project/acceptance-based; licensing is whipsawed by catch-up payments and renewals. The only genuinely recurring streams are licensing (contracted to 2030) and core software (Fact).

Outlook for products/services? How big will this market be? Network Infrastructure addressable market growing (management cites a 14% implied NI-market CAGR, 27% for AI/cloud specifically); mobile flat; licensing flat but durable, extending via 6G. International by nature (US ~30% of sales, then a long tail) (Fact/Assumption on the CAGRs — management estimates).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More, in equipment (price erosion, Samsung gaining, Open RAN commoditizing the radio, hyperscalers building own silicon). Licensing is structurally protected by standards-essential patents (Fact/Interpretation).

How profitable is the business (ROIC/ROE)? Poor: ROE 3.1%, ROCE 4.3% — below cost of capital (Fact). The exception is Nokia Technologies (~70% operating margin), which subsidizes the rest.

How profitable is the industry — competitors, barriers? Structurally low-return despite oligopoly, because of monopsony buyers and mandatory ~24% R&D. Five RAN players; barriers (standards, scale, trusted-vendor status) exist but do not translate to pricing power in hardware (Interpretation).

Can the business be easily understood? Only at a high level. The four→two segment reorganization, the comparable-vs-IFRS gap, the EUR/USD ADR split, and the hidden dominance of licensing make true earnings power hard to read (Interpretation).

Undermined by foreign low-cost labor? The competitive threat is Chinese vendors (Huawei/ZTE) on price/scale, blunted in the West by trusted-vendor exclusion — a policy moat, not a labor-cost one (Fact).

Do brands matter? Yes, in two ways: trusted-vendor reputation in carrier/defense procurement, and the Nokia brand itself is a licensed asset (brand licensing within Nokia Technologies) (Fact).

Nature of competition / switching costs? Hardware: low switching costs, multi-vendor by design. Core software: real switching costs (embedded in 65% of 5G SA networks). Licensing: not “switchable” — patents must be licensed (Fact/Interpretation).

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The patent portfolio (26,000+ families) is largely internally generated and carried far below economic value — the biggest hidden asset. The €2.5B re-recognized Finnish DTA is an on-balance-sheet asset of uncertain realizability (Fact/Interpretation).

Off-balance-sheet liabilities? Standard operating leases, pension obligations (defined-benefit plans referenced), and contingent litigation/contractual-claim provisions. Nothing flagged as unusually large (Fact, with an open question on contingent IP litigation).

How conservative is the accounting? Mixed. Conservative on cash. Aggressive in presentation: management guides/reports/is-paid on a “comparable” measure that excludes €1.2B (57%) of recurring-flavored costs, and reclassified venture-fund volatility out of operating profit in 2025 (Interpretation — the central QoE flag).

How CapEx-hungry is the business? Historically capital-light (~3% of sales, manufacturing outsourced) — but rising to €900M–1B (~5%) in 2026 as Nokia vertically integrates into optical/InP fabs (Fact).

Capital Allocation & Management

How much FCF, and how is it used? ~€1,465M FCF (2025). Uses, in stated priority: R&D (€4.9B), bolt-on/minority M&A, a stable/growing dividend (€759M), then buybacks (€624M) (Fact).

Significant recent acquisitions? Infinera (~€2.5B, Feb 2025, the thesis-defining deal); Nokia Shanghai Bell 50% buy-in (€501M). And a divestiture: FWA to Inseego (Apr 2026) (Fact).

Buying back shares? Yes (€624M in 2025) — but simultaneously issuing shares (127.4M for Infinera, 166.4M to Nvidia), so net share count rose, and buying at a ~99th-percentile valuation (Interpretation: value-destructive timing / dilution-offset).

Issuing large amounts of stock to insiders? SBC €337M and rising (+41% YoY) — material and worth monitoring under the new Silicon-Valley-style leadership (Fact).

Compensation policy / incentives? CEO LTI tied to comparable operating profit and a cash-flow metric — i.e., the number that excludes restructuring and acquisition amortization, not IFRS profit/ROIC/TSR (Fact — alignment flag). CEO has a modest personal co-investment with 2:1 matching.

Motivations of management? New, externally hired team (ex-Intel/ex-Infinera) executing a credible AI/data-center repositioning; incentives lean toward the adjusted metric they are reshaping the company around (Interpretation).

Valuation & Market Data

ADR, MLP, or K-1 issuer? ADR (NYSE: NOK; 1 ADR ≈ 1 ordinary share). Primary listing Nasdaq Helsinki (NOKIA.HE). No MLP/K-1; foreign private issuer filing 20-F/6-K (Fact).

Dividend policy? “Recurring, stable, over time growing”; proposed €0.14/share for FY2025; yield ~1.2–1.4%; covered by FCF (Fact).

How profitable is the business? Low aggregate returns (3% ROE) with a high-margin licensing core — see above (Fact).

Net income vs. cash from operations diverging? Cash conversion is reasonable (FCF €1,465M vs €651M net income — FCF exceeds net income, helped by D&A and low cash taxes), which is a positive quality signal at the cash level even as reported profit is thin (Fact).

Risks & Downside

What would cause the stock to decline? Primarily multiple compression (99th-percentile valuation, 26% above consensus); secondarily order-to-revenue disappointment, optical margin commoditization, a licensing renewal cliff, or a mobile/RAN air-pocket (Interpretation — see the Risk Analysis section).

Risk of catastrophic loss? Low — net cash, licensing annuity, diversified revenue, 0.5 beta. The realistic downside is permanent capital impairment via de-rating (base case implies −17 to −26%; bear −55 to −62%), not insolvency (Interpretation).

Chance of a total loss? Negligible (Interpretation).

Recent News & Events

Has the business environment changed recently? Materially, yes — the Nvidia investment (Oct 2025), the AI/data-center order surge, the Infinera close, a new CEO/leadership team, and the two-segment reorganization have transformed the narrative and positioning over the past ~18 months (Fact). Whether the economics have changed is the open question (Interpretation).

Significant acquisitions / divestitures? Infinera (buy), Shanghai Bell buy-in, FWA/Inseego (sell) — see above.

Change in accounting policies? Yes — venture-fund fair-value gains/losses moved out of operating profit (2025); segment reporting changed to two segments (Jan 1, 2026) (Fact — both worth noting for comparability).

Recent operational changes? New optical fab capacity (San Jose), Nokia Defense unit, restructuring program (through 2026), and a steady cadence of carrier 5G wins (e.g., Indosat/Indonesia) and product launches (e.g., Deepfield security) (Fact — most are routine, not thesis-changing).

APPENDIX B — Source Appendix — Nokia Oyj (NYSE: NOK)

*All sources accessed 2026-06-10/11 unless noted. Primary sources prioritized; management commentary treated as hypothesis and validated against filings and external evidence. *

Primary — Company filings & disclosures

1. Nokia Oyj Form 20-F, FY2025 (filed 2026-03-05). SEC EDGAR CIK 0000924613. https://www.sec.gov/Archives/edgar/data/924613/000162828026015034/nok-20251231.htm — segment results, comparable-vs-IFRS reconciliation, balance sheet, Infinera purchase-price allocation, deferred-tax-asset disclosure, risk factors, executive compensation, patent-portfolio disclosures.
2. Nokia 6-K filings (2025–2026) via EDGAR — quarterly results, the Nvidia investment, restructuring updates, debt issuance, buyback/AGM authorizations. SEC EDGAR filings index: https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0000924613&type=6-K
3. Nokia FY2025 / Q4-2025 results & earnings call (January 29, 2026) — FY2025 financials, 2026 guidance (comparable OP €2.0–2.5B; FCF conversion 65–75%; capex €900M–1B).
4. Nokia Q1-2026 results & earnings call (April 23, 2026) — Q1 comparable operating profit €281M (6.2% margin), FCF €629M, NI growth raised to 12–14%, AI/cloud orders €1B.
5. Nokia Capital Markets Day (November 19, 2025) — two-segment structure; 2028 targets (group double-digit OP growth; NI revenue CAGR 6–8%, NI op margin 13–17%; MI op profit from €1.5B base; Optical double-digit op margin); AI/cloud TAM CAGR.
6. Nokia Special Call (October 28, 2025) — Nvidia partnership and AI-RAN strategy; CEO strategic framing.
7. J.P. Morgan 54th Annual Global Technology Conference — Nokia presentation (May 19, 2026) — order conversion, supply constraints, fab ramp, optical margin trajectory, Inseego/FWA divestiture reference.
8. Nokia investor relations — https://www.nokia.com/about-us/investors/ — segment definitions, capital-allocation framework, patent-portfolio statistics.

Primary — Quantitative cross-checks

9. SEC EDGAR XBRL financial facts (CIK 0000924613) — reconciliation of reported figures.
10. Aggregated market-data feeds (2026-06-10/11) — multi-period statements (EUR), own-history valuation percentiles (composite 98.9; P/E 98.5; P/B 99.1; P/S 99.1), snapshot (GICS, employees, short interest, ownership), consensus target. Third-party aggregated data; reconciled to the 20-F.
11. yfinance (2026-06-11) — price $13.40, market cap, EV, trailing/forward multiples, peer comps. Unofficial; reconciled to filings.

Secondary — News & market

12. Benzinga, “What’s Going On With Nokia Stock Today?” (2026-06-09, id 399941) — 12-month return (+168%), technical levels, Indosat/Indonesia 5G + AI-RAN, Deepfield launch.
13. Benzinga, “Why Is Nokia Stock Falling Friday?” (2026-06-05, id 397524) — €500M 3.625% 2032 senior-notes issuance (refinancing 2028 notes); 12-month return (+208%); profit-taking.
14. Benzinga, Nokia/Indosat Ooredoo Hutchison 5G collaboration (2026-06-09, id 399311); Nokia Deepfield Genome Shield launch (2026-06-09, id 399292).

Peer & industry reference

15. Ericsson (closest twin) — Q3-2023 earnings commentary framing mobile-infrastructure investment as structurally flat/cyclical (“market normalization, not incremental market growth”); valuation comps (14.5x P/E, ~8x EV/EBITDA).
16. Cisco, Arista, Ciena — data-center/optical comp-set multiples (yfinance, 2026-06-11): CSCO 39.5x P/E / 28.5x EV-EBITDA / 7.7x P/S; ANET 52x / 42x / 19.7x; CIEN 145x / 79x / 11x.
17. Optical/datacom peers (Coherent, Lumentum) — optical/AI-datacenter TAM and capital-cycle framing (cross-read).

Analytical frameworks

18. Greenwald & Kahn, Competition Demystified — barriers-to-entry taxonomy applied per segment (intangibles/IP, switching costs, scale, cost).
19. Chancellor (ed.), Capital Returns (Marathon Asset Management) — supply-side capital-cycle analysis applied to the optical/indium-phosphide capacity build.

Note on currency: Nokia reports in EUR; the NOK ADR trades in USD. EUR/USD ≈ 1.085 used for cross-checks. Note on the four→two segment change (effective Jan 1, 2026): the FY2025 20-F reports the legacy four-segment structure; CMD figures are recast to the new two-segment basis. Both are presented in the memo, flagged.