Telefonaktiebolaget LM Ericsson

From a fundamental and capital-markets perspective, Ericsson sits in an awkward but investable middle ground. The balance sheet is strong. The company ended Q2 2026 with roughly SEK 59.8 billion of net cash, after ending 2025 at SEK 61.2 billion and while running a buyback. Gross margin has improved materially from the 2023 trough. Free cash flow over the last five years has been volatile but mostly positive outside the 2023 downturn. Yet revenue growth isn’t structurally high. The enterprise expansion still hasn’t proven itself cleanly, and the core RAN market isn’t in a broad upcycle. Dell’Oro said the total RAN market stabilized in 2025 and is expected to remain stable in 2026, with only about 1% CAGR through 2030. That is the setting in which Ericsson must convince investors that software layers and licensing can outgrow the industry’s flattish capex base.

The right qualitative portrait is a company in transition. It isn’t distressed: the balance sheet is too strong and the core franchise too relevant for that. Nor is it a mature cash cow, since the earnings stream is still too cyclical and the business mix is still shifting. And it’s certainly not high-quality compounding growth, given how much still depends on carrier spending and how little of the new software narrative is separately measurable in the accounts. Ericsson has already completed the operational turnaround phase that Börje Ekholm and his team spent years on. What comes next is harder. The next phase is proof, not repair: proof that AI in RAN can become recurring software revenue, proof that licensing momentum can keep broadening beyond smartphones, and proof that Ericsson can absorb memory inflation and flat RAN spending without surrendering the margin gains it has only recently rebuilt.

Company vertical history

Origins and listing path

Ericsson began in Stockholm in 1876 as Lars Magnus Ericsson’s telephony workshop, which existed because the telephone itself was still a fragile, fast-improving invention and national telecom systems had not yet consolidated into a handful of giant suppliers. The original problem was simple and highly practical: build and repair better telephone equipment in a market where reliability and local adaptation mattered. That DNA still shows up in today’s company: a supplier that won by staying close to network architecture shifts and by living inside standards and operators’ real deployment problems, not a lab and not a brand.

Its capital-markets path reflects its age. Ericsson’s A-share was listed on the Stockholm Stock Exchange on 1919-05-15, the B-share was listed in January 1929, and the New York-listed ADR was added in 1985. The present-day dual-class structure remains economically ordinary but politically meaningful: A-shares carry one vote, B-shares one-tenth of a vote, while dividends are the same. For investors, that means Ericsson is clearly public and liquid, but not a one-share-one-vote governance case.

Stage division and the enduring capability

Ericsson’s first long stage was the national and international telephony era, when the company built fixed-line know-how and export habits. The lasting capability from that period wasn’t manufacturing scale so much as standards fluency and customer intimacy in critical infrastructure: the ability to sell mission-critical equipment into regulated networks that cannot fail casually.

The second decisive stage was the mobile era, especially GSM and then 3G, when Ericsson became one of the global companies that effectively built mobile telecom as modern infrastructure. In that phase, Ericsson’s growth was driven by carriers’ multiyear capex waves and by the migration from country-specific systems to global wireless standards. The company’s central advantage became cumulative R&D and standards participation, which later supported both its radio business and its patent portfolio. The current business still rests on that foundation: more than 60,000 granted patents and a very large installed base outside China.

The third stage was the telecom bust and restructuring period around the early 2000s. Ericsson’s own history describes 2002 as a turning point, and the Hellström period was shaped by the stock-market crash, operator retrenchment, and the one-year negotiation that led to the Sony Ericsson mobile-phone joint venture. That stage matters because it taught Ericsson, the hard way, that being present across the telecom stack does not mean every adjacency is worth owning. It survived, but the period embedded a permanent lesson in cyclicality.

The fourth stage was the years after Börje Ekholm took over as CEO in 2017. Ericsson’s 2025 annual report notes that Ekholm has served as President and CEO since 2017. This period brought cost repair, portfolio tightening, product focus in Networks, and a financial reset from years of weak profitability. It also brought the opposite of discipline in one conspicuous case: the USD 6.2 billion Vonage acquisition announced in 2021 and completed in 2022. That deal was meant to push Ericsson into a global network-and-communications platform model, but in October 2023 Ericsson had to take a SEK 32 billion non-cash impairment against goodwill tied to Vonage. So the Ekholm era produced two different legacies at once: a real repair of the RAN core and a reminder that strategic expansion can be overpaid.

The fifth stage is the one investors are now trying to price. The business enters it with stronger margins, net cash, an ongoing buyback, and more explicit software monetization language. But it also enters with a CEO handoff already announced. Per Narvinger, currently head of Business Area Networks, will become CEO on 2026-10-01, while Ekholm steps down on 2026-09-30. That makes the current moment more consequential than a normal product launch cycle: Ericsson is moving into a new management chapter at the same time it launches AI in RAN, with the next CEO coming directly from the Networks core and inheriting both the repaired hardware franchise and the unresolved enterprise/software questions.

Key nodes that still matter today

The Sony Ericsson joint venture in 2001 was important because it marked Ericsson’s retreat from being a standalone handset maker. In hindsight, that was strategically correct. The modern Ericsson is a network company with licensing upside, not a consumer-device company. The same logic explains why Ericsson now benefits economically from smartphones through patents rather than by trying to out-design Apple or Samsung.

The 2019–2023 U.S. Department of Justice compliance sequence still matters because it imposed a governance cost. Ericsson disclosed in 2022 that the DOJ considered the company’s pre-DPA Iraq disclosure insufficient and said Ericsson had breached the 2019 Deferred Prosecution Agreement through later disclosure failures. In March 2023 Ericsson announced a resolution under which it would plead guilty to previously deferred charges relating to conduct prior to 2017 and pay a fine of USD 206.7 million. This did not destroy the company’s financial position, but it put a credibility discount on management and remains relevant whenever investors judge governance quality.

The Vonage acquisition and later impairment matter because they exposed the big strategic temptation hanging over every telecom-equipment vendor: escape the carrier capex cycle by buying a software story. Ericsson’s purchase rationale was clear enough. Vonage brought CPaaS, UCaaS, CCaaS, and about one million registered developers to a thesis about network APIs. But the later SEK 32 billion impairment showed that capital-market timing, rates, and enterprise-software comparables can make even a strategically plausible deal financially painful. That history is one reason investors are cautious about assigning a large multiple premium to Ericsson’s software ambitions today.

The 2025 iconectiv divestment and the 2026 capital-return decision matter for the opposite reason: they improved confidence that Ericsson’s board and management are again willing to prune, simplify, and return cash. The Q4 2025 results highlighted strong free cash flow, net cash of SEK 61.2 billion, a higher dividend of SEK 3.00, and the proposed SEK 15 billion buyback. Those are the actions of a company whose core franchise is throwing off cash again, even if that cash stream remains cyclical.

Financial vertical review and price history

The financial arc over 2021–2025 is easier to read than the strategic arc. Net sales went from SEK 232.3 billion in 2021 to SEK 271.5 billion in 2022, then fell to SEK 263.4 billion in 2023, SEK 247.9 billion in 2024, and SEK 236.7 billion in 2025. That is not a straight growth story. It is a cycle story. The improvement worth paying attention to is not top-line momentum; it is margin repair. Gross margin moved from 38.6% in 2023 to 44.1% in 2024 and 47.6% in 2025. Free cash flow before M&A swung from 13.8% of sales in 2021 to 8.2% in 2022, turned negative at -0.4% in 2023, then rebounded to 16.2% in 2024 and 11.3% in 2025. Return on capital employed followed the same path, reaching 24.1% in 2025 after only 2.6% in 2024 and a negative 10.8% in 2023.

The business reason behind those numbers is straightforward. When operator capex slowed and the enterprise strategy was still digesting Vonage, Ericsson’s revenue line weakened and the goodwill-heavy accounting of the enterprise push damaged reported profits. What repaired the picture was a combination of geographic mix, product competitiveness in Networks, cost discipline, and a less distorted cash-flow line than the income statement. Ericsson’s own five-year free-cash-flow data show that cash generation recovered much faster than confidence did, and the shares rerated into early July 2026 before the second-quarter reminder that inventories and component costs can still break the rhythm.

The balance sheet is one of Ericsson’s strongest arguments. The company ended 2025 with SEK 61.2 billion of net cash and Q2 2026 with about SEK 59.8 billion, even after the dividend and the start of the buyback. Equity ratio was 39.5% at year-end 2025. That does not make Ericsson immune to cyclical disappointment, but it sharply lowers permanent-capital-loss risk from leverage. The main balance-sheet watch items are working capital, inventories, and the residue of acquisition accounting rather than debt solvency.

The stock’s own recent history mirrors that operational story. By 2025 the market was rewarding Ericsson for recovering margin and cash flow in a flat RAN market. Then July 2026 produced a classic equipment-vendor drawdown: one revenue miss, one weak free-cash-flow quarter, one input-cost warning, and a fast multiple reset. The share-price decline from SEK 112.75 on 2026-07-13 to SEK 92.80 on 2026-07-21 tells you where the valuation center now sits. Investors will pay Ericsson for hard cash, licensing, and visible software monetization. They will not pay up for slogans about AI unless the financial path is visible.

Key data tables

Sources: Ericsson Annual Report 2025.

This table looks messy if you expect a secular growth company. It looks much healthier if you expect a cyclical infrastructure vendor with a repaired margin structure. Revenue has not compounded. Returns and cash conversion have recovered sharply. That is why Ericsson can support dividends and buybacks without depending on leverage, but also why the equity cannot yet earn a growth-stock multiple.

Business model, moat, and industry cycle

How the business machine runs

Ericsson’s revenue structure is dominated by Networks. In 2025, Networks contributed 64% of sales, Cloud Software and Services 26%, Enterprise 9%, and Other 1%. Networks generated SEK 151.0 billion of sales and SEK 29.8 billion of EBIT in 2025, implying a 19.7% EBIT margin. That is the subgroup that pays for most of the company. The licensing stream is booked differently, but it lifts the quality of the broader mix: 2025 IPR licensing revenue was SEK 14.5 billion, up from SEK 14.0 billion in 2024 and SEK 11.1 billion in 2023.

That mix explains Ericsson’s operating leverage. A large share of costs are fixed or quasi-fixed: R&D, standards work, software development, field support, and the commercial infrastructure required to serve global operators. When revenue falls, profit can move violently if mix worsens. When software, licensing, and high-value upgrades rise inside the total, margins recover quickly. Ericsson’s recent gross-margin improvement came from exactly that kind of mix and execution repair. The 2025 annual report also notes capex of only SEK 2.6 billion, about 1.1% of sales. This isn’t a capital-light business in the software sense, nor a heavy fabs-and-plant one; the real sustaining investment is R&D and product development.

AI in RAN and the software-mix question

Ericsson’s June 2026 AI in RAN launch matters because it is not simply an R&D paper or a roadmap slide. The company described it as a software subscription, commercially scalable, available first in Q2 2026, and designed to run in real time on Ericsson basebands and radios without additional hardware. Ericsson says the product works with Ericsson 5G Advanced across both purpose-built and Cloud RAN platforms, and the public trial work with T-Mobile showed a neural-network-driven scheduler improving spectral efficiency and downlink performance directly on deployed Ericsson hardware. In telecom terms, that is real. It lives at the radio timescale, not in a distant analytics layer.

The monetization case is real too, but still mostly potential. Ericsson has not disclosed a separate AI in RAN revenue line, has not disclosed pricing, and has not disclosed attach rates. There is also no disclosed figure for the subset of the installed base that is immediately eligible. The addressable footprint is large in principle. Ericsson says its equipment is installed at nearly a quarter-million U.S. sites and carries 60% of U.S. wireless traffic. It also says roughly half of mobile traffic outside China runs on Ericsson networks. If AI in RAN can be sold as a recurring software layer across even a modest slice of that base, the revenue quality could improve meaningfully. But the financial statements do not yet show that happening in a separately visible way.

The right judgment is that AI in RAN is strategically important and financially early: not vaporware, but not yet a proven second growth engine either. For 2026, the most reasonable expectation is that it helps defend Ericsson’s radio franchise, supports software mix inside Networks, and perhaps begins contributing at the margin. It is too early to underwrite a full valuation rerating on the product alone.

Real moats and weaker moats

Ericsson’s first real moat is scale in standards-based mobile infrastructure. Global RAN is an industry with very few true full-stack vendors, and Dell’Oro’s 2024 and 2025 market updates still place Huawei, Ericsson, Nokia, ZTE, and Samsung as the top suppliers, with Ericsson second globally and first outside China in 2024. That matters because telecom operators do not casually swap out nationwide radio vendors. The switching costs are technical, operational, and political at the same time.

Its second real moat is cumulative radio and systems know-how. The installed base, interoperability work, optimization data, and product-in-field experience are difficult for a smaller entrant to recreate. This is the practical reason Ericsson can credibly launch software-only upgrades like AI in RAN: the company already owns the hardware platform, the scheduler logic, the silicon path, and the operator relationship.

Its third real moat is patents. Ericsson reports more than 60,000 granted patents and generated SEK 14.5 billion of IPR licensing revenue in 2025. That portfolio is not a side hobby. It is the monetized residue of decades of R&D. The relevance of the patent moat is reinforced by 2026 events: the Transsion settlement, the Verifone payment-terminal agreement, and the expansion of Avanci vehicle licenses. Those are all signs that Ericsson’s IP remains commercially enforceable across more than one device category.

The weaker moat is enterprise adjacency. Ericsson’s ambition to build a network-API and enterprise communications platform is understandable, but Vonage showed the limits of buying your way into a new profit pool. This business line may still work over time, especially around network APIs and enterprise wireless, but it has not yet earned the right to be treated as a proven moat on the same footing as the RAN and patent franchises.

Management, governance, and capital allocation

Management deserves real credit for the operational turnaround in the core business. Ericsson itself said in 2023, during the CFO transition, that under the outgoing CFO’s tenure the company had regained technology leadership and strengthened its financial position. The evidence is visible in gross margin, free cash flow, and net cash. Yet management also owns the Vonage overpayment and the compliance failures that culminated in the DOJ resolution, so management credibility is neither low nor unqualifiedly high. It is mixed, with strong marks in core execution and weaker marks in strategic M&A judgment.

Governance is stable but not pure. Ericsson’s A/B structure gives A-shares ten times the votes of B-shares, which supports long-term control stability but embeds a governance discount relative to one-share-one-vote structures. The approaching CEO change reduces one uncertainty and creates another. It is positive that succession looks orderly. It is still a fresh management transition at a moment when the business model is being tested by both AI opportunity and AI-driven input-cost pressure.

Horizontal competitor analysis

What each competitor became

Nokia is the most relevant public peer, walking a similar road from a different starting point: no longer the old handset Nokia in substance, but a network and IP company trying to use the AI era to rebalance away from slow carrier spending toward stronger optical, IP, and standards income. Its Q1 2026 report showed comparable sales growth, 45.5% gross margin, strong optical momentum, and 49% growth in AI-and-cloud-customer sales. The contrast with Ericsson is useful. Nokia’s current AI narrative is tilted more toward transport and datacenter-adjacent networking, while Ericsson’s is more tightly tied to mobile RAN software and radio intelligence. Customers pick Nokia when transport, optical scale, and Bell Labs-led breadth matter; they pick Ericsson when macro RAN execution and radio quality matter most.

Huawei is the hardest competitor to compare financially and the hardest to dismiss strategically. It is private, but it remains the largest RAN supplier globally and posted 2025 revenue of CNY 880.9 billion with CNY 192.3 billion of R&D spend, or 21.8% of revenue. Huawei’s commercial position rests on breadth, domestic scale, and the ability to spread telecom R&D across a much larger corporate base. Where politics allows it to compete, it remains a formidable benchmark. The practical consequence for Ericsson is that being “global number two” in RAN often means fighting for the non-China share of the market under geopolitical constraints, not holding a soft duopoly position.

Samsung Networks is smaller and more selective. Samsung Electronics does not give investors a clean standalone Networks income statement, which itself says something about the business’s role inside the group. Public Samsung disclosures emphasize vRAN, Open RAN, and North American wins, and Samsung said the Networks business improved both quarter on quarter and year on year in Q4 2025 on North America sales. Customers choose Samsung when they want an alternative vendor with strong silicon integration, Open RAN credibility, and the balance-sheet comfort of a giant parent. They usually do not choose Samsung for the same breadth or global installed-base depth that Ericsson and Nokia bring.

Cisco is not a direct RAN peer, but it matters around enterprise networking, cloud connectivity, security, and valuation context. Cisco’s current profile is that of a much more diversified networking and software company with a far larger market capitalization and a cash-return model the market treats as steadier. Its presence is a reminder of what Ericsson is not: a broad enterprise-networking franchise with a very large installed software base across IT teams. That comparison matters because investors sometimes overextend the “AI networking” trade across very different kinds of infrastructure companies. Ericsson’s AI exposure sits much closer to operator radio economics than Cisco’s does.

Competitive landscape and niche

Ericsson’s niche is the leading Western macro-RAN supplier outside China, with a strong patent book and a credible but still unproven path to more software-led monetization. That is a powerful niche in a market where there are only a few vendors that operators can trust for full-scale national deployments. At the same time, it is still a niche inside a mature market. Dell’Oro expects the total RAN market to be stable in 2026 and only about 1% CAGR through 2030. So Ericsson’s problem isn’t market access; it’s where incremental growth comes from when the industry’s volume base is almost flat.

Set against Nokia, the most direct comparison is margin trajectory: Ericsson’s gross-margin repair has been more visibly tied to Networks execution and North American exposure, while Nokia’s latest narrative has leaned more toward AI-and-cloud transport demand and portfolio reconfiguration. Against Huawei, the comparison that matters most is scale and political reach: Huawei has wider overall corporate scale, but Ericsson benefits where Huawei is excluded or politically constrained. Against Samsung, it comes down to breadth: Samsung can win important pockets, especially around virtualized and open architectures, but it does not yet match Ericsson’s installed RAN depth.

If the industry shifts toward more software-rich, operator-monetizable upgrades on existing infrastructure, Ericsson’s position strengthens because of its installed base and because AI in RAN does not require a hardware refresh. Shift instead toward pure price competition on commodity radio hardware while component costs rise, and Ericsson’s position weakens, because the very installed base that helps software monetization also ties the company to a large hardware footprint. That is the double edge in the current investment case.

Current fundamentals, valuation, and risk

The last four quarters and what the market is trading

The operating story through the last four reported quarters is not one of collapse. Q3 2025 showed sales of SEK 56.2 billion and adjusted gross margin of 48.1%. Q4 2025 then delivered organic growth in all three segments, strong free cash flow, a dividend increase, and the planned buyback. Q1 2026 showed SEK 49.3 billion of sales, adjusted gross margin of 48.1%, and free cash flow before M&A of SEK 5.9 billion. The break came in Q2 2026: sales fell to SEK 52.7 billion, adjusted gross margin held at 48.4%, but free cash flow before M&A dropped to SEK 0.4 billion as inventory rose ahead of Q3 deliveries and memory costs stayed elevated.

So the stock is currently trading two things at once. First, it is trading Ericsson as a repaired cash generator with net cash, buybacks, and a still-strong licensing stream. Second, it is trading fear that the margin and cash-flow restoration was too dependent on favorable mix and too exposed to hardware inputs. The July 2026 drawdown says the second fear dominated in the near term.

AI in RAN opportunity versus AI-driven cost pressure

On the opportunity side, AI in RAN is the most interesting new product development because it attacks the biggest strategic problem in telecom equipment: how to monetize an installed base without waiting for another capex wave. Ericsson’s launch language is unusually commercial for a telecom feature set. This is not “we are exploring.” It is “software subscription,” “commercially scalable,” and “available in Q2 2026.” That wording matters because the software business model, not just the technology, is what investors care about. If operators can activate these features on existing Ericsson 5G Advanced infrastructure, Ericsson can sell performance gains and energy savings with much less friction than a full new hardware cycle.

But the financial disclosure does not yet support a big 2026 earnings thesis. AI in RAN is not separately disclosed in the income statement. There is no disclosed site-based pricing, no disclosed attach rate, and no disclosed ARR target. The economic conclusion is therefore measured: AI in RAN is real as product, real as commercial packaging, and promising as margin mix, but still too early to call materially earnings-moving in 2026. The most probable near-term value is defensive. It helps Ericsson protect share, justify software upsell, and reinforce the quality of the Networks franchise before it changes the group revenue model.

On the risk side, the memory-cost issue looks real but probably not existential. Ericsson said in substance, through management commentary reported by Reuters, that AI hyperscaler demand for the same memory supply chain used in base-station hardware is pressuring gross margin into Q3. Omdia and related industry reporting expect the RAN market itself to remain stable in 2026, while smartphone and device reporting around mid-2026 also pointed to a memory shortage severe enough to alter vendor behavior and potentially ease only in early 2027. Ericsson does not disclose the BOM share of DRAM or HBM in its radio products, so any precise estimate would be false precision. The best inference from the disclosures is that this is a real but contained group-margin headwind, probably measured in tens of basis points rather than several percentage points of gross margin, because hardware represented 37% of 2025 sales and memory is only one part of hardware COGS. The danger would be a broader component squeeze that spreads beyond memory into a longer hardware-cost cycle.

Licensing momentum and present materiality

Licensing has the opposite profile from AI in RAN. It is already financially material but still often underappreciated because it is buried inside a larger industrial story. Ericsson’s 2025 IPR licensing revenue of SEK 14.5 billion equaled roughly 6% of group sales. On a low-single-digit operating margin business that figure matters a great deal. 2026 developments suggest the stream is broadening: IAM reported that all top 10 smartphone vendors are now licensed and that Ericsson’s annual licensing run-rate has reached about SEK 13.5 billion; Ericsson’s own Transsion announcement said the financial benefit would begin in Q3 2026; and Avanci’s reported Chinese-automaker momentum broadens the addressable auto base. The payment-terminal agreement with Verifone is small in absolute terms, but strategically useful because it proves Ericsson can take the cellular SEP model into another IoT vertical.

Licensing won’t replace Networks, but it can support normalized earnings even when RAN is flat, which makes Ericsson cheaper on owner earnings than a superficial “telecom equipment” label suggests.

Valuation analysis

Historically, Ericsson doesn’t deserve a premium multiple reserved for durable high-growth software, but it deserves more than a distressed hardware multiple when gross margin is near 48%, net cash is near SEK 60 billion, licensing revenue is recurring, and free cash flow is positive through the cycle. At SEK 92.80, the share trades below its July peak but not at a washout price. Using 2025 free cash flow before M&A of SEK 26.8 billion and 3.37 billion shares, trailing free cash flow per share is about SEK 7.9, implying an FCF yield of roughly 8.6%. That is not expensive. It is also not a screaming bargain once you remember that 2025 benefited from stronger operating conditions than Q2 2026 just showed.

For cash-flow passthrough, Ericsson’s own definition of free cash flow before M&A is already close to an owner-earnings concept: operating cash flow less net capex, other investments excluding M&A, and lease repayments. With capex only SEK 2.6 billion in 2025, the delta between free cash flow before M&A and a rough owner-earnings view is not huge. The bigger issue is that reported net income is noisy because it has been distorted by the Vonage impairment in 2023, near-zero 2024 earnings, and the iconectiv-related strength in 2025. On that basis, owner earnings are the better anchor than accounting EPS.

The three valuation scenarios below therefore rest on normalized owner earnings, not on 2025 reported EPS.

This is valuation-scenario analysis within a research framework, not investment advice.

On expectation gap, the market is not demanding heroic growth. It is demanding proof that Ericsson can hold the repaired margin structure while the industry stays flat. The next few quarters matter less for raw revenue than for three line items: Networks gross margin, inventory normalization, and evidence that licensing and software mix can offset hardware cost pressure. Q3 2026 matters most of all: Ericsson itself said Q2 cash flow was hurt by inventory built for planned Q3 deliveries, and the Transsion settlement starts contributing in Q3. Bulls and bears will both get cleaner data at the same time.

On margin of safety, the answer is restrained. At SEK 92.80, the share is above the conservative scenario value band and inside the base-scenario hold zone. That means the margin of safety is not obvious. If earnings are flat for three years and Ericsson only returns cash through the current dividend rate, the return case remains positive but modest. This is a good company in a strategically useful niche, but not yet a bad enough price to create a large valuation cushion.

Risk analysis, catalysts, tracking indicators, and research uncertainties

The first real permanent-capital-loss risk is not debt. It is margin relapse. Probability medium, impact high. If AI-related memory inflation, operator procurement pressure, and weaker geographic mix drag group gross margin back toward the mid-40s for several quarters, the market will conclude that 2024–2025 was a mix-assisted peak rather than a structurally repaired level. The transmission path runs from gross margin to EBITA, then to cash flow, then to the multiple assigned to the whole group. What to watch is simple: group gross margin below 46% for two consecutive quarters, or Networks profitability rolling over while revenue is only flat.

The second is “software story without software numbers.” Probability medium, impact medium-to-high. AI in RAN can help strategically without helping valuation if Ericsson does not start showing it in attach rates, software mix, gross-margin lift, or customer references. Investors have already lived through one enterprise-adjacency disappointment with Vonage. The transmission path here is slower: expectation builds, financial visibility fails to follow, and the stock stays trapped on a hardware multiple.

The third is licensing normalization after a burst of settlement headlines. Probability medium, impact medium. The 2025 and 2026 licensing story is good, but smartphone licensing is maturing. IAM’s reporting that all top 10 smartphone makers are now licensed is good news, yet it also means future gains will depend more on renewals, compliance, new verticals like IoT and payment terminals, and the automotive pool. If those newer verticals scale more slowly than hoped, the licensing line may remain valuable but stop surprising.

The fourth is governance and capital-allocation relapse. Probability low-to-medium, impact medium. The balance sheet is strong, but the record is mixed. Ericsson’s core turnaround was real; the compliance case and the Vonage impairment were real too. The new CEO will inherit a business that no longer needs rescue but still needs strategic precision. The observable indicator is not a single ratio. It is whether management starts talking faster than the financials move, especially around enterprise or AI adjacencies.

Positive catalysts are concrete. The strongest would be a Q3 2026 report showing working-capital normalization, restored free cash flow, and a clean read-through from the Transsion settlement into IPR revenue. A second would be evidence that operators are paying for AI in RAN rather than just trialing it. A third would be further proof that Ericsson can widen licensing beyond smartphones and vehicles into more IoT categories.

Negative catalysts are equally concrete. Another quarter of weak free cash flow, persistent inventory growth, or explicit guidance that component inflation will persist past 2026 would hurt. So would any sign that AI in RAN is being treated by customers as a free feature rather than a separately monetizable subscription.

The next earnings date comes from Ericsson’s financial calendar: Q3 2026 is scheduled for 2026-10-15.

Research uncertainties are real here. The first is that Ericsson does not publicly disclose AI in RAN pricing or attach rates. The second is that memory-cost exposure is not quantified by BOM, so gross-margin estimates must remain inferential. The third is that Samsung Networks is not separately broken out cleanly enough for a perfect peer comparison. The fourth is that Huawei is private and geopolitically constrained, which makes its strategic weight easier to see than its segment economics.

Sources

The core primary sources for this report are Ericsson’s 2025 annual report, Ericsson’s Q1 2026 and Q2 2026 earnings releases, Ericsson’s investor materials on foreign exchange, share information, governance, and the financial calendar, plus official AI in RAN and Transsion announcements. Secondary sources used for triangulation include Reuters, Dell’Oro Group, Omdia references, Nokia’s annual and Q1 2026 reports, Huawei’s 2025 annual report, Samsung Electronics disclosures, IAM, Avanci, and public market data pages used to verify dated closes.

Cross-synthesis summary

Looking across Ericsson’s full history, the capability it has most clearly proven isn’t device innovation, software-platform dominance, or financial engineering: it’s knowing how to survive and remain central in a market where the technology stack keeps changing but the customer problem remains the same, building, upgrading, and optimizing national communications infrastructure with very low tolerance for failure. That capability is what carried the company from fixed telephony through GSM, 3G, 4G, 5G, and now the first attempts to push AI directly into the radio layer. The success was never just luck, and never just cycle: it came from technical depth, standards participation, operator trust, and the patience to keep R&D relevant across decades. Those strengths are still present. What is less clear is whether Ericsson can turn them into a meaningfully higher-quality earnings stream than the market has historically assigned to telecom equipment.

Horizontally, Ericsson’s real advantage over competitors is not that it is the biggest. Huawei is bigger in RAN globally. Cisco is bigger in networking and software more broadly. Samsung can subsidize network ambitions with a giant parent balance sheet. Nokia has its own IP and transport strengths. Ericsson’s advantage is a narrower but still valuable combination: strong scale outside China, a deep macro-RAN installed base, high operator trust in the West, and a patent book that genuinely throws off cash. Its weakness is that the company still depends too much on a flat carrier capex market to earn a durable rerating. That weakness is temporary if AI in RAN becomes a visible software revenue layer and licensing keeps broadening. It is structural if those hopes stay strategically true but financially invisible.

The valuation isn’t rewarding fantasy, but it isn’t giving investors much of a cushion either. The current price largely pays for a repaired core business, a strong balance sheet, and some but not all of the licensing and software optionality. What the market is most likely still misjudging is the shape of the upside and downside. The upside here is a slower upgrading of revenue quality if AI in RAN, licensing, and software mix begin to show up in gross margin and cash conversion, not a sudden “AI company” rerating. The downside isn’t bankruptcy or balance-sheet stress: it’s that Ericsson remains a decent business valued as a decent business while the hoped-for second curve stays too small to matter.

For the next year, the critical variables are gross margin, inventory release, Q3 licensing contribution, and any hard proof of AI in RAN monetization. For the next three years, the decisive question is whether Ericsson can make the installed base behave more like a software asset and less like a recurring hardware reset. For the next five years, the deepest question is whether telecom networks become sufficiently programmable and monetizable that the winners capture more revenue after installation, not just at installation. Ericsson could become a better investment under three conditions: the shares move into a real margin-of-safety zone, AI in RAN starts showing up in measurable software economics, and licensing continues proving that it can expand beyond handsets without major legal or regulatory friction. An investor should revisit the whole judgment if gross margin slips below 46% for multiple quarters, if licensing weakens materially, or if new management repeats the style of capital allocation that produced the Vonage impairment.

Bull and bear reasons

The bull case begins with the balance sheet: Ericsson had about SEK 59.8 billion of net cash at Q2 2026 after ending 2025 at SEK 61.2 billion and while running a buyback, which gives the company room to absorb cyclical weakness without capital-structure stress.

The second bull reason is that gross-margin repair appears real: adjusted gross margin held around 48% in Q1 and Q2 2026 after a substantial rise from the 2023 trough, showing operational improvement rather than only topline luck.

The third is licensing quality: SEK 14.5 billion of IPR revenue in 2025, plus new 2026 settlements and automotive-platform expansion, gives Ericsson a higher-margin earnings stream that is not wholly tied to RAN shipment cycles.

The fourth is installed-base optionality: AI in RAN is already marketed as a software subscription that can run on existing Ericsson 5G Advanced infrastructure, which is exactly the kind of monetization path the company has long lacked.

The bear case starts with market structure: Dell’Oro expects the total RAN market to remain stable in 2026 and grow only about 1% CAGR through 2030, which limits how much organic growth Ericsson can get from the old engine.

The second bear reason is near-term execution risk: Q2 2026 showed how quickly weak cash flow, higher inventory, and component-cost inflation can knock confidence out of the shares.

The third is disclosure risk around the new software thesis: AI in RAN has strong product language but no separate revenue line, no public pricing, and no public attach-rate data.

The fourth is capital-allocation history: the SEK 32 billion Vonage impairment is too large to forget and limits the premium investors will pay for management’s narrative around new adjacencies.

Pre-mortem

One plausible 50% drawdown script runs like this. Through 2027, operator capex stays flat, AI in RAN remains mostly a feature rather than a paid software layer, and memory or broader component pressure keeps group gross margin near 45%–46%. Licensing stops surprising because the smartphone settlements are largely done and newer IoT verticals scale slowly. Free cash flow settles in the mid-teens of billions rather than the mid-20s, and the market re-rates Ericsson from roughly a low-teens owner-earnings multiple to 8–9x. In that case the B-share could trade down into the SEK 45–55 range. The stock would not be broken because the balance sheet is broken; it would be repriced because the transition case stalled.

A second script is management-specific. The October 2026 CEO handoff goes smoothly at first, but 2027 brings another push into enterprise or software adjacencies that lacks clear returns. Investors, already conditioned by Vonage, react harshly to any hint that cash from the repaired RAN core may be redeployed into low-visibility software ambitions. Even if operating results are mediocre rather than disastrous, the multiple can compress at the same time as earnings flatten. That is how a “good enough” business can still lose half its equity value without a solvency event.

Final research conclusion

Ericsson is worth owning only with the correct expectation. It isn’t a clean AI rerating candidate, nor merely a declining hardware name. It is a repaired telecom-infrastructure franchise with a strong balance sheet, a real patent annuity, and a promising but still mostly unproven path toward more software-like monetization inside the installed RAN base. The shares already discount part of that repair. They do not yet offer a large enough valuation cushion to pay investors for the execution risk still embedded in the next phase.

What worries me most is not cyclical volatility in itself. It is the possibility that Ericsson remains strategically directionally right but financially frustrating: AI in RAN proves useful but not separately material soon enough, licensing remains valuable but mature, and the enterprise/API story stays too small to shift the group multiple. What would change my mind is a combination of two things: a lower entry price and harder evidence that software and licensing are rising contributors to group cash generation and margin stability, not just narrative support beams.

【Company-profile scores】

• Fundamental quality: medium
• Growth: low
• Moat: medium
• Financial soundness: strong
• Management credibility: medium
• Valuation attractiveness: medium
• Risk level: medium
• Suitable investor type: cyclical

【Investment rating】

• Rating: Hold
• One-line thesis: Ericsson’s core RAN franchise is repaired and cash-rich, but AI-in-RAN monetization is still early and the stock is not yet cheap enough to ignore the cycle.
• Three price signals
  • Ideal buy price: see line below
  • Acceptable hold price: SEK 85–115
  • Clearly overvalued price: SEK 126 and above
• Current-price classification: acceptable hold
• Whether to wait for a better price: yes; a buy becomes interesting below roughly SEK 70, preferably with evidence that Q3/Q4 cash flow normalizes and AI in RAN shows real paid uptake. The opportunity cost of waiting is the dividend stream and the chance of a recovery rerating if Q3 surprises cleanly.
• Target holding horizon: 3–5 years
• Expected annualized return: conservative ≈ 0%–2%; base ≈ 5%–7%; optimistic ≈ 10%–12%
• Max-loss risk: roughly 40%–50% if gross margin relapses toward 45%–46%, software monetization stays immaterial, and the valuation compresses toward trough hardware multiples
• Reassessment-trigger signals: group adjusted gross margin below 46% for two consecutive quarters; trailing free cash flow before M&A below 6% of sales; annualized IPR revenue slipping below SEK 12 billion; no visible AI in RAN monetization markers by 2027; management pursuing another large, low-visibility acquisition

【Ideal Buy Price】SEK 64–70 Basis: at least a 20% margin of safety below the conservative scenario fair-value band of SEK 80–88.

【Valuation Range】

• current: 92.80 (close as of 2026-07-21)
• bear (conservative · ideal buy zone): [64, 70]
• base (fair · acceptable hold zone): [85, 115]
• bull (optimistic · above the clearly-overvalued line): [126, 136]

Other tickers mentioned

• NOKIA.HEL: direct listed peer in mobile networks, transport, and telecom IP
• 005930.KS: Samsung Electronics, whose Networks business is the most relevant Asian listed challenger outside China
• CSCO.US: enterprise-networking reference point for software mix, cash returns, and AI-networking valuation context
• QCOM.US: major SEP licensor used as a benchmark for the economics of cellular IP
• IDCC.US: smaller pure-play wireless licensing comparator for the value of recurring patent revenue

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.