Nokia’s 30% collapse in July looked like the market calling time on a hype cycle. It was not. Read the quarter and the opposite happened: AI & Cloud order intake nearly tripled to €2.8bn, Optical Networks grew 20% and IP Networks 16%. Demand accelerated. What broke the stock was a cost line — Ericsson warned that memory-chip inflation would compress equipment margins into 2027, and the entire sector de-rated in sympathy. That is the single most useful thing to understand about Nokia (NYSE: NOK) at $10.76. Its bull case and its bear case are the same event seen from opposite ends. The AI memory shortage that turned Micron into an 8.4x winner is the shortage now raising Nokia’s input costs. Nokia is short exactly what Micron is long. The street prices that tension between $8.50 and $21.00, and where you land depends entirely on which side of the shortage you think dominates.
The insight: one shortage, two directions
The mechanism is worth spelling out because almost nobody connects the two halves. Three manufacturers — SK hynix, Samsung and Micron — control more than 95% of global DRAM production. From 2025 they began systematically reallocating wafer capacity toward high-bandwidth memory to serve AI accelerators, and by mid-2026 HBM was consuming roughly 25% of total DRAM wafer output. That capacity used to supply conventional DRAM to everyone else, and “everyone else” includes the people who build mobile base stations.
So the AI boom reaches Nokia twice. It arrives as revenue, through optical and IP networking gear sold into data centres, and it arrives as cost, through the conventional DRAM in every radio unit Nokia ships. The second effect is nastier than it sounds because of contract structure: Ericsson noted that most telecom equipment contracts are long-term and lack automatic price-pass-through clauses, so a vendor facing a component spike cannot simply reprice. It has to absorb the hit or renegotiate. Ericsson guided Q3 Networks adjusted gross margin down to 48–50% and described the inflation as building “gradually” through the second half of 2026 and into 2027.
Set that against the other side of the trade. Our Micron bull and bear case and the CXMT listing that rattled Micron and SK Hynix describe the same scarcity from the seller’s chair, where it shows up as record margins. Nokia sits in the buyer’s chair. Any forecast for NOK is therefore a forecast about which moves faster: AI-driven revenue arriving, or AI-driven input costs arriving.
Key facts
- NOK last close $10.76, up 1.89%; 52-week closing range $4.13 to $16.85 — 14 August 2026 (StockAnalysis)
- Street targets: consensus $15.02, high $21.00, low $8.50 across 11 analysts, consensus rating Buy (StockAnalysis)
- Q2 2026 net sales €4.82bn, up 8%; comparable operating profit €434m, up 18%; comparable operating margin 9% versus 8.3% — Nokia, 23 July 2026
- AI & Cloud order intake €2.8bn in Q2, up from roughly €1.0bn in Q1
- Optical Networks +20%, IP Networks +16%, while Fixed Networks fell 2%
- Nvidia invested $1bn for a 2.9% stake and a joint AI-RAN platform for 6G
- HBM consumes ~25% of DRAM wafer output by mid-2026, squeezing the conventional DRAM that base stations need — Ericsson, July 2026
What actually happened: a quadruple and a giveback
Nokia closed at $4.13 in August 2025, a price that valued it as a structurally declining telecom equipment vendor with a licensing business attached. Ten months later it closed at $16.85 on 2 June 2026, having roughly quadrupled. The catalyst was a genuine change of identity rather than a sentiment swing. Under Justin Hotard, who arrived from Intel’s data centre and AI group in 2025, Nokia repositioned from selling radios to telcos toward selling optical and IP networking into AI data centres.
Nvidia then validated it with money. In October 2025 Nvidia took a $1bn equity stake, becoming a 2.9% shareholder, alongside a partnership to build an AI-RAN platform for 6G and to explore incorporating Nokia’s data centre switching and optical technology into Nvidia’s future architectures. “The next leap in telecom isn’t just from 5G to 6G – it’s a fundamental redesign of the network to deliver AI-powered connectivity, capable of processing intelligence from the data center all the way to the edge,” said Justin Hotard, President and CEO of Nokia. “Our partnership with Nvidia will accelerate AI-RAN innovation to put an AI data center into everyone’s pocket.”
Then July happened. Ericsson reported a weak quarter and flagged component inflation, the broad technology tape sold off, and a stock that had quadrupled met concentrated profit-taking. Nokia fell roughly 30% over the month, bottoming near $9.83 before recovering to $10.76. Crucially, none of that was Nokia-specific news about demand. The Q2 report that landed on 23 July was, on its own terms, good.
The bull case: the order book is not a legacy order book
Nokia’s Q2 produced €4.82bn of net sales, up 8%, with comparable operating profit of €434m, up 18%, lifting the comparable operating margin to 9% from 8.3% and the comparable gross margin to 46% from 45.3%. Comparable EPS came in at €0.07 against €0.04 a year earlier. Those are respectable numbers for a company long assumed to be structurally stuck.
The composition is what matters. Network Infrastructure reached €2.04bn from €1.83bn, with Optical Networks up 20% year on year — particularly strong in the Americas — and IP Networks up 16% on a constant-currency basis, both explicitly attributed to AI and cloud demand. Technology and licensing grew 15%. Mobile Networks, the traditional core, grew about 7%. Fixed Networks shrank 2%. There are visibly two companies inside Nokia, and the AI-exposed one is growing roughly three times as fast as the legacy one.
Management is guiding accordingly: Network Infrastructure net sales growth of 12–14% on a constant-currency portfolio basis for 2026, with IP Networks and Optical Networks combined at 18–20%, full-year comparable operating profit of €2.1–€2.6bn, free cash flow conversion of 55–75% of comparable operating profit, and capex of just €800–900m. That last figure deserves attention. Nokia is participating in the AI infrastructure build without the capital intensity that defines the neocloud operators or the merchant power developers. It sells picks and shovels and keeps its balance sheet.
The strongest single data point is the order intake. AI & Cloud orders of €2.8bn in one quarter, up from roughly €1.0bn in Q1, sit against consensus 2026 revenue growth of just 4.3% and 2027 growth of 6.7%. Orders convert to revenue with a lag, and a book building at that rate is difficult to reconcile with mid-single-digit revenue modelling. Either the order intake proves lumpy and non-repeating, or the estimates that anchor the $15.02 consensus are too low.
The bear case: a 3.5% net margin meeting a cost shock
The bear case does not require the AI story to be false. It requires only that the margin arithmetic stays punishing. Nokia generated $808.87m of net income on $23.30bn of trailing revenue — a net margin of about 3.5%. Trailing EPS is $0.14 and the trailing P/E is 74. This is a business with almost no cushion, which is precisely why a component cost shock is dangerous. A few points of gross margin is the difference between the guidance range’s top and bottom.
The DRAM squeeze is not speculative, and it is not close to resolving — SanDisk used its investor day to argue memory stays tight into 2028. It is already in a competitor’s guidance, and Nokia buys from the same constrained suppliers into the same long-term customer contracts without automatic pass-through. The mitigation available — raising prices to telco customers — is slow, contested, and lands in a market where operators have spent a decade forcing equipment prices down. Nokia’s comparable gross margin of 46% has roughly 40 percentage points less room than Micron’s 84.9%.
There is also a credibility discount that is entirely earned. Nokia has announced strategic transformations repeatedly since 2013 without producing durable margin expansion, and a 5.60bn-share count means dilution has done real work over the years. The $8.50 low target implies roughly 21 times the 2027 consensus EPS of $0.40, which is not obviously cheap for a company the street models growing revenue 6.7%. The bear does not have to believe Nokia fails. It only has to believe Nokia remains a mid-single-digit grower with thin margins that briefly got repriced as an AI stock.
The numbers: what the range actually assumes
At $10.76 the market is paying about 24 times forward earnings, against consensus EPS of $0.34 for 2026 and $0.40 for 2027 on revenue of roughly $20.75bn and $22.13bn. The bull and bear targets are best read as multiples on that 2027 figure.
J.P. Morgan’s Sandeep Deshpande sits at the $21.00 high, set on 12 June — before the drawdown — which implies roughly 52 times 2027 EPS. That only works if the AI & Cloud order intake converts into materially higher estimates than consensus carries today. The freshest bullish marks came after the fall and after Q2: Northland’s Tim Savageaux at $20 and Craig-Hallum’s Christian Schwab at $15 on 24 July, with Bank of America’s Oliver Wong at $18 on 23 July. Argus’s Jim Kelleher also carries $15. The $8.50 floor implies about 21 times 2027 EPS and assumes component inflation eats the operating leverage before it reaches shareholders.
The honest read is that the consensus $15.02 is not a forecast so much as an average of two incompatible views. Roughly 39% upside to consensus from here is unusually wide for a European incumbent, and it exists because the analysts genuinely disagree about whether the memory squeeze is a two-quarter irritation or a two-year margin regime.
What happens next
Prediction one: Q3 is a margin print, not a revenue print. Ericsson has already told the market that component inflation builds gradually through the second half. Nokia’s Q3 comparable gross margin — 46% in Q2 — is the number that decides the next leg. Hold it near 46% and the bear case loses its mechanism. Slip toward 43–44% and the full-year €2.1–2.6bn operating profit range resolves to its lower half, which is roughly where the $8.50 case lives.
Prediction two: the order-to-revenue conversion becomes the whole argument by early 2027. A €2.8bn AI & Cloud quarter has to start appearing in reported Network Infrastructure sales. If IP and Optical track toward the top of the 18–20% guided range and the order book keeps building, estimates move up and the gap between $15 consensus and $20–21 bull targets closes from below. If the €2.8bn proves to be one large lumpy award, the re-rating stalls.
Prediction three: Nokia becomes a relative trade against the memory makers. Because the same shortage drives both, the cleanest expression of a view is no longer NOK alone but NOK against Micron or SanDisk. If DRAM pricing keeps climbing, memory wins and equipment loses. If HBM capacity additions finally loosen conventional DRAM in 2027, the trade reverses and Nokia gets its margin back without selling a single extra router.
The stock at $10.76 sits almost exactly between a bear case built on a cost line and a bull case built on an order book, which is a reasonable place for it to be given nobody yet knows which one compounds faster. What has changed is that Nokia is no longer a bet on telecom capex cycles. It is a leveraged position on the spread between AI networking demand and AI memory costs — and that is a far more interesting, and far more volatile, thing to own than what this company was two years ago.
Frequently asked questions
What is the bull case price target for Nokia stock?
The highest live street target is $21.00 from J.P. Morgan’s Sandeep Deshpande, set on 12 June 2026, implying about 95% upside from the $10.76 close on 14 August 2026. That target predates the July drawdown. The most recent bullish marks are Northland Securities at $20 and Bank of America at $18, both set in late July after Q2 results. The 11-analyst consensus is $15.02.
What is the bear case price target for Nokia stock?
The lowest live street target is $8.50, implying roughly 21% downside. That case rests on memory-chip cost inflation compressing gross margins faster than AI-driven revenue arrives. It values Nokia at about 21 times 2027 consensus EPS of $0.40 — a normal multiple for a company growing revenue in the mid-single digits with a 3.5% net margin.
Why did Nokia stock fall about 30% in July 2026?
It was not company-specific bad news. Ericsson reported a weak quarter and warned that surging memory-chip prices would compress equipment margins into 2027, triggering contagion selling across telecom equipment stocks. That coincided with a broad technology selloff and heavy profit-taking after Nokia had roughly quadrupled from $4.13. Nokia’s own Q2, reported 23 July, beat on profit.
How does the AI memory shortage hurt Nokia?
SK hynix, Samsung and Micron control over 95% of DRAM production and have shifted capacity toward high-bandwidth memory for AI accelerators, with HBM taking around 25% of wafer output by mid-2026. That tightens the conventional DRAM used in base stations. Because most telecom equipment contracts are long-term without automatic price-pass-through, vendors absorb the cost increase rather than passing it on immediately.
What did Nvidia’s investment in Nokia actually buy?
Nvidia invested $1bn for a 2.9% equity stake in October 2025, alongside a partnership to build an AI-RAN platform for 6G. The two also agreed to explore incorporating Nokia’s data centre switching and optical technology into Nvidia’s future AI infrastructure architectures. It is a strategic validation and a potential channel, not a guaranteed revenue commitment.
Is Nokia an AI stock or a telecom stock?
Both, and that is the point. Optical Networks grew 20% and IP Networks 16% in Q2 on AI and cloud demand, while Fixed Networks shrank 2% and Mobile Networks grew about 7%. Nokia guides Network Infrastructure to 12–14% growth in 2026 with IP and Optical combined at 18–20%. The AI-exposed segments are growing roughly three times faster than the legacy business, but the legacy business is still the larger part of the company.
This article is for information only and is not investment advice. Prices, analyst targets and estimates are as of the close on 14 August 2026 and will have changed.