BE Semiconductor Industries NV (EURONEXT: BESI) — Precision Leadership Meets Memory-Timing Risk

Published: 2026-09-12 · Verdict: Watch · Research confidence: High (88%)

Executive conclusion

Analyst Take

Recommendation: WATCH. Besi is one of the most economically attractive franchises in semiconductor assembly equipment, but the security is not priced like a normal cyclical equipment supplier. At the September 11, 2026 close of €189.25, the company had an equity value of approximately €15.3 billion and an enterprise value near €15.1 billion. Against reconstructed last-twelve-month revenue of €733.8 million, EBITDA of approximately €301.6 million, and net income of €208.7 million, that represents about 20.6 times sales, 50.2 times EBITDA, and 73.3 times earnings. September consensus of €4.33 EPS for 2026 and €6.58 for 2027 lowers the calculated multiples to approximately 43.7 and 28.8 times, but those estimates already assume that revenue rises from €591 million in 2025 to €993 million in 2026 and €1.37 billion in 2027. The drawdown from June’s €328.40 intraday high removed an extreme valuation, not the need for unusually strong execution. [S1][S2][S20][S21]

The operating evidence is substantially better than it was a year ago. H1 2026 revenue increased 48.8% to €434.7 million, orders increased 116.5% to €562.6 million, and operating income increased 108.7% to €172.7 million. Q2 gross margin reached 65.7%, last-twelve-month orders reached €987.6 million, and management guided Q3 revenue 10–15% above Q2, implying roughly €275–287 million. Die attach represented 84.3% of H1 revenue, and management classified approximately 60% of H1 system orders as AI-related. These are powerful near-term facts, although the interim financial statements are explicitly unaudited and the AI classification is management-defined. [S2][S3]

The balance sheet also improved. Conversion of the €175 million 2029 convertible notes removed the nearer maturity, leaving €509.1 million of cash and deposits, €345.1 million of debt, and €164.0 million of net cash at June 30. The economic cost was dilution: shares outstanding excluding treasury increased from approximately 79.28 million at year-end to 80.82 million, or 1.9%, because treasury shares were delivered on conversion. This is why gross buyback spending is not the same as per-share accretion. [S2][S16]

The central thesis, however, is narrower than the order acceleration. Besi already has credible production adoption in logic and a defensible precision-placement franchise. It reports leading market share in high-accuracy die attach, sustains gross margins above 60% through cycles, and combines placement technology with Applied Materials’ surface preparation, cleaning, metrology, and traceability in the Kinex system. Applied’s 9% equity investment validates strategic relevance. It does not create exclusivity, minimum equipment purchases, a board seat, or a takeover floor. [S4][S8][S9]

The load-bearing uncertainty is high-bandwidth memory. On the Q1 call, management said all three major memory manufacturers had Besi bonders and described evaluation milestones that would determine 2027 adoption. On the Q2 call, it acknowledged that some industry participants expected delay while others expected an imminent decision, declined to disclose hybrid-bonder units, and said Q3–Q4 evaluation results would determine how much and which applications adopted hybrid bonding. That change is not proof of a lost program, but it reduces confidence in calendar precision. More importantly, ASMPT disclosed that its thermocompression bonders were already meeting high-volume requirements for HBM3E 12-high and supporting low-volume HBM4 manufacturing. Kulicke & Soffa is also expanding thermocompression capacity while developing hybrid systems. Hybrid bonding can eventually be technically superior without dominating the next production generation. [S6][S7][S10][S11]

My central valuation range is approximately €190–215, with a bear case around €115–130 and a bull case around €290–320. This is an estimate, not a price fact. The central range assumes most current orders convert, non-memory applications broaden, and 2027 earnings approach consensus; it also assumes the market continues to award a premium multiple. The bear case does not require technological failure—only delayed HBM adoption, a lower earnings estimate, and a more ordinary equipment multiple. An accumulation zone would become more attractive below roughly €140, or at a higher price if binding production-scale HBM orders and disclosed repeat-unit evidence raise the sustainable earnings base. No formal metadata entry or target is supplied because the listed share and analysis are euro-denominated while the metadata fields are restricted to US dollars.

Investment conviction is therefore lower than business-quality conviction. Confidence is high that Besi has valuable precision-placement technology, an asset-light cost structure, and strong near-term demand. Confidence is moderate that 2027 consensus is achievable, and lower that the present price provides an adequate return if qualifications move by even two or three quarters. The strongest counter-case is that logic, photonics, advanced flip chip, and TC Next collectively make the memory debate less important while the outsourced model delivers exceptional incremental margins. That case is credible, but the current valuation requires much of it to occur rather than paying investors to wait.

The near-term decision sequence is observable. Q3 revenue must reach the guided range without a material gross-margin shortfall; receivables must convert to cash; hybrid programs must progress from evaluation counts to production-linked repeat systems; and non-HBM platforms must demonstrate that annual growth is not controlled by one memory transition. The call would become more constructive following multiple production-scale memory orders, disclosed repeat units or installed-base service economics, and stable mid-60s gross margins. It would become more negative if HBM insertion moved toward HBM5 or decade-end, orders remained evaluation-heavy, gross margin fell below 60%, or 2027 estimates declined while the multiple stayed elevated.

Stock Price Action — Five-Year Event Map

Besi’s five-year price history resembles a long-duration technology option layered on a cyclical equipment company. Prices are facts; the drivers assigned to them are interpretations unless a dated disclosure establishes causality. Company Financials daily data, reconciled to the Euronext listing, show the shares rising from a 2021 opening near €50.28 to €189.25 on September 11, 2026, with extraordinary volatility between those endpoints. The September price sat approximately 42% below the June 2026 record high but remained well above the 52-week low. [S20][S22]

• 2021 expansion: The shares traded between approximately €48.93 and €87.76 and ended at €75.02, up roughly 49% from the first closing price. Besi subsequently reported €749.3 million of revenue and €939.1 million of orders. The plausible interpretation is that the market capitalized the pandemic-era semiconductor-equipment boom and record bookings, although the price move itself does not prove that explanation. [S1][S20]

• 2022 contraction: The stock ended at €56.56 and reached approximately €38.46 in October. Revenue remained high at €722.9 million, but orders fell 29.3% to €663.7 million. The sequence supports the interpretation that investors anticipated a back-end equipment downturn before reported revenue fully contracted. [S1][S20]

• 2023 rerating before recovery: The stock finished at €136.45, about 141% above year-end 2022, despite revenue falling 19.9% and operating income declining 27.4%. This was a valuation rerating around AI, chiplets, and advanced-packaging expectations, not confirmation from contemporaneous earnings. That distinction remains relevant because the shares can rise or fall well before order conversion appears in the income statement. [S1][S20]

• 2024 order-optionality volatility: The stock reached about €182.90, fell to roughly €91.20, and closed at €132.30. Besi announced material hybrid-bonding activity, while annual revenue recovered only 4.9% to €607.5 million and operating profit still declined. The approximately 50% high-to-low range indicates that investors were using individual qualification and order announcements as proxies for a much larger prospective market. [S1][S20]

• 2025 strategic validation without earnings acceleration: The shares ranged from approximately €79.62 to €152.75 and ended at €133.75. Applied Materials acquired 9% through market purchases in April. The investment validated the strategic relationship but expressly came without board representation or plans for additional purchases. FY2025 revenue then declined 2.7%, although orders rose 16.8%, setting up the subsequent recovery. [S8][S20][S24]

• June 2026 peak: The stock reached an intraday record of €328.40 on June 15 and closed near €318 on the June 18 Investor Day, when management raised its undated long-term revenue model to €1.7–2.2 billion and operating-margin target to 45–55%. The move occurred before the target had been converted into customer-specific production evidence. [S5][S20]

• Post-peak reset: The shares closed at €229.90 on the July 23 Q2 release date and at €189.25 on September 11, 42.4% below the June intraday high but still 41.5% above year-end 2025. The likely drivers include profit-taking, duration compression, debate over HBM qualification timing, and recognition that thermocompression remains competitive. These are informed interpretations; no single issuer release attributes the decline. [S6][S10][S20]

The stock was therefore neither fully washed out nor simply expensive because it had once traded higher. A 42% peak decline coexisted with a large positive year-to-date return. The factor model was not supplied, so numerical betas, style scores, factor alpha, and sector exposures cannot be responsibly stated. Observable price behavior suggests sensitivity to AI capital spending, semiconductor-equipment expectations, long-duration growth valuations, and program-specific hybrid-bonding news; those are qualitative inferences, not factor-model outputs.

Verdict. Price action confirms high expectation duration and event sensitivity, not intrinsic value. The June peak is an irrelevant anchor unless the associated long-term targets become production orders and cash earnings.

Business Overview

Besi designs and supplies equipment used after wafers have been fabricated. Its tools pick dies from wafers, align them with substrates or other dies, create electrical and mechanical connections, encapsulate packages, singulate units, and perform plating or related wet processing. It reports one IFRS operating segment because management aggregates die attach, packaging, and plating for resource allocation, but it discloses product-family revenue that makes the economics easier to understand. In H1 2026, die attach generated €366.6 million, packaging €59.7 million, and plating €8.4 million—84.3%, 13.7%, and 1.9% of revenue, respectively. [S2]

How customers obtain value

The customer does not buy nominal placement accuracy. It buys acceptable cost per good package. A system is economically valuable only if it can place expensive known-good dies repeatedly, maintain alignment, avoid contamination, preserve yield, achieve required throughput, integrate with factory automation, and remain available during a product ramp. The cost of one failed advanced package can include multiple logic, memory, and interface dies, so a relatively expensive bonder can create value if it prevents yield loss or enables a package architecture that was otherwise impossible.

Besi’s CHAMEO family is marketed around sub-micron placement, cleanliness, inline inspection, wafer handling, and high throughput. Those specifications are vendor claims and should not be confused with sustained customer production results. Independent technical work from imec shows why the distinction matters: die-to-wafer hybrid bonding at a 2-micron pitch is feasible, but surface particles, chemical-mechanical polishing, singulation, overlay, and throughput remain linked production challenges. Imec reported good but far-from-perfect electrical yields in a research demonstration, reinforcing the distance between a technically successful test vehicle and memory-stack production economics. The equipment sale is therefore part machine, part recipe, part applications engineering, and part service commitment. [S4][S12]

The business is readily understandable at the driver level: revenue is principally sales of high-value assembly systems plus service, spares, tooling, upgrades, and support; demand depends on semiconductor units, dies per package, attach steps per die, required accuracy, tool throughput, and customer capital spending. Forecasting is difficult because each driver is multiplied by qualification timing. A single large hybrid order can materially change a year, while a qualification delay can shift the same economic opportunity into another reporting period. [S1][S6]

Revenue model and stability

System revenue is predominantly transactional. Customers place orders for specific machines, often tied to capacity additions or product qualifications, and most orders do not have the cancellation protection of a long-duration software contract. Besi describes assembly-equipment demand as lagging the semiconductor cycle, and management emphasized on the Q2 call that equipment upcycles remain cyclical and can create excess capacity after six to eight quarters. Approximately 15% of 2025 revenue came from service and spares, which provides a useful installed-base tail but is not large enough to make total revenue recurring. [S1][S6]

Revenue stability is low quarter to quarter and moderate only when viewed across a cycle: annual revenue was €749.3 million in 2021, €722.9 million in 2022, €578.9 million in 2023, €607.5 million in 2024, and €591.3 million in 2025 before H1 2026 accelerated. The 2021–2025 compound growth rate was negative even though gross margin improved. [S1][S22]

The important quality distinction is between volume stability and unit economics. Gross margin increased from 59.6% in 2021 to 64.9% in 2023 and 65.2% in 2024 despite the revenue decline, then eased to 63.3% in 2025. That indicates mix, pricing, and manufacturing flexibility. Operating margin nevertheless fell from 42.4% in 2021 to 29.3% in 2025 because R&D and selling infrastructure did not fall with shipments. The moat can preserve gross economics while fixed engineering expenditure still produces rapid profit contraction.

Services are economically attractive because an installed tool requires spares, calibration, application support, upgrades, and rapid field response. Yet the disclosed percentage combines activities with different recurrence and margin characteristics. A multiyear service agreement is more durable than a one-off conversion kit; an emergency spare may carry high margin but no contractual repeat. Until Besi separates those streams, applying a software-like valuation to all service revenue would be unjustified.

Customers, concentration, and acceptance risk

Besi sells to leading foundries, integrated device manufacturers, outsourced semiconductor assembly and test companies, memory producers, photonics assemblers, and electronics manufacturers. The top ten customers accounted for approximately 44% of 2025 revenue, and no single customer exceeded 10% of consolidated revenue. Those accounting statistics understate program concentration. Incremental hybrid revenue can depend on a few logic, memory, foundry, or hyperscaler-linked platforms even when consolidated customer concentration looks diversified. [S1][S4]

The relevant unit of analysis is therefore not just named-customer revenue. It is technology by customer, evaluation versus production tool, accepted versus unaccepted shipment, repeat-unit count, process-of-record status, and ultimate factory utilization. Management reported that hybrid-bonding adoption increased from 15 customers at year-end 2025 to 21 at June 2026. Other public presentation material has used an 18-customer 2025 reference, and one published bridge does not reconcile cleanly. Different cutoffs or definitions may explain the discrepancy, but it means customer counts are indicators of interest, not precise revenue-model inputs. [S1][S2][S4]

Revenue recognition follows IFRS policies, but cash collection can lag shipment when customer acceptance or normal credit terms apply. H1 2026 trade receivables rose by €74.9 million to €254.1 million. That movement is consistent with rapid growth, yet it is also why bookings and shipments cannot be treated as cash until acceptance and collection occur. [S2]

The order book is a leading indicator, not a contractual-equivalent revenue stream. Customers can order ahead to reserve supply, qualify two vendors, change delivery schedules, or cancel. Besi’s record €987.6 million trailing order intake demonstrates access to attractive programs and strong near-term demand, but the economic proof is the sequence from order to shipment, acceptance, payment, repeat order, and service activity. [S2][S6]

Geography and supply architecture

The business is international and Asia-centered. In 2025, approximately three-quarters of revenue was generated in Asia, where most semiconductor assembly capacity resides. H1 2026 concentration increased: disclosed revenue from China, Taiwan, Thailand, Malaysia, Korea, and other Asian locations totaled approximately €378.6 million, or 87.1% of group revenue. China alone represented €182.6 million, or 42.0%. [S1][S2]

Besi maintains important design, intellectual-property, and corporate functions in Europe while relying on Malaysia, China, Vietnam, Singapore, and other Asian locations for sourcing, assembly, applications, and service. Its outsourced, modular manufacturing model reduces owned factory capital and helps capacity flex, but it does not eliminate operational dependency. Suppliers must still reserve components and labor, and Besi must integrate, calibrate, test, and support finished systems. Geographic diversification lowers single-site risk but does not remove dependence on Asian customers, suppliers, shipping lanes, or regional stability.

Foreign low-cost labor by itself is not the principal competitive threat because Besi already sources and manufactures extensively in lower-cost Asian locations. The more important threat is a foreign vendor combining lower engineering costs, domestic customer access, public financing, and sufficiently good yield. Mature die attach, molding, and plating are more exposed than the most demanding sub-micron bonding steps.

Physical and intangible capital

The manufacturing model is physically asset-light. H1 2026 purchases of property, plant, and equipment were only €4.1 million on €434.7 million of revenue. Engineering is the true reinvestment: H1 expensed R&D was €46.5 million and capitalized development additions were €11.8 million. In 2025, expensed R&D was €81.0 million and capitalized development was €26.0 million. [S1][S2]

The most important unrecognized assets are process recipes, motion-control software, contamination-management knowledge, application engineers, supplier integration, installed-base history, and customer qualification records; the balance sheet recognizes only part of their value through goodwill, acquired intangibles, and capitalized development. [S1]

Those assets have unusual depreciation characteristics. Knowledge may remain productive for many years, but it can lose value quickly if another interconnect method becomes the process of record. Conversely, accounting amortization may occur even while qualification and installed-base value are increasing. This is why book value is not a useful standalone valuation anchor and why R&D should be analyzed as investment without automatically adding every euro back to earnings.

Security form and tax status

BESI.AS represents ordinary shares of a Dutch public limited company listed on Euronext Amsterdam. It is not a partnership, master limited partnership, K-1 issuer, or ADR; the separately traded OTC receipt does not change the legal form of the Amsterdam share. [S1][S20]

Dutch dividend withholding and treaty rules may affect an investor’s realized return depending on domicile and account type. That is an investor-specific tax matter rather than a change in operating value.

Verdict. Besi combines cyclical, transactional equipment revenue with unusually resilient gross margins, valuable engineering assets, and a growing but still modest service tail. The business model is understandable; the timing of customer acceptance and production qualification is not predictably recurring.

Industry Dynamics

Market structure and addressable market

Besi operates in a narrow portion of semiconductor capital equipment: assembly, advanced packaging, and related back-end processes. Management and its cited market researcher estimated the 2025 assembly-equipment market at about $5.4 billion and Besi’s served portion near $1.5 billion. The August 2026 presentation projected the broader market at $6.5 billion in 2026, $8.0 billion in 2027, and $8.3 billion in 2028. An earlier company presentation used a different scope and forecast approximately $8.8 billion for 2030. These are management-sponsored estimates, not reported market facts, and the changing numbers illustrate sensitivity to category definitions and forecast date. [S4][S5]

The more defensible structural conclusion is that package complexity is increasing. Chiplets allow designers to combine compute, memory, input/output, and specialized functions built on different process nodes. HBM stacks memory dies vertically. Co-packaged optics moves optical engines closer to compute. These architectures increase the number of placements, bonding steps, and yield-critical interfaces per finished package.

Assembly demand can be represented as semiconductor packages multiplied by dies per package, attach or bonding steps per die, and equipment time per step. Besi can therefore grow faster than semiconductor units if complexity and placement intensity rise. That mechanism also explains why a delay in hybrid bonding does not erase all demand: thermocompression, fine-pitch flip chip, advanced placement, and optical assembly can benefit from the same architectural shift.

Demand is overwhelmingly international. Leading packaging facilities and customers are located across Taiwan, Korea, China, Malaysia, Singapore, Japan, Europe, India, and the United States. Regionalization and government incentives can increase duplicate capacity, but announced factory spending is not equivalent to addressable tool orders. Cleanrooms may be built in phases, projects may be delayed, and only a portion of packaging capex reaches Besi’s process steps.

Hybrid bonding’s economic proposition

Conventional microbump interconnects use solder-based structures between dies. Hybrid bonding directly joins copper and dielectric surfaces, enabling finer pitch and potentially reducing resistance, capacitance, package height, and thermal constraints. These attributes become more valuable as interconnect density and stack height increase. TSMC states that its SoIC process starts at a sub-10-micron bond pitch and that its 3-nanometre chip-stacking technology entered volume production in 2025, establishing that direct-bonding architectures are commercially relevant in logic. [S13]

Independent technical evidence supports feasibility but also explains slow qualification. Imec demonstrated die-to-wafer hybrid bonding at a 2-micron pad pitch with overlay below 350 nanometres, Kelvin electrical yield above 85%, and daisy-chain yield above 70%. Its work depended on exceptionally clean surfaces, controlled planarization, plasma dicing, overlay measurement, and throughput improvements. A laboratory or pilot result therefore proves a physical pathway, not high-volume cost per good package across multiple memory layers. [S12]

Samsung’s shipment of 12-layer HBM4E samples in May 2026 confirms that next-generation memory roadmaps are advancing. Samsung reported that HBM4 had already entered mass production and said HBM4E mass production would follow customer schedules. Neither disclosure identifies Besi as a production supplier or establishes that HBM4E volume manufacturing will use Besi equipment. It is evidence for the market opportunity, not Besi-specific revenue. [S14]

Thermocompression as a bridge and competitor

The strongest contradiction to a simple hybrid-bonding inevitability thesis comes from ASMPT. In its H1 2025 results, ASMPT reported an installed base exceeding 500 thermocompression bonders, said its bulk tools at a leading HBM customer met high-volume requirements for HBM3E 12-high, and said another customer began low-volume manufacturing of HBM4 12-high with its TCB equipment. It also reported first-half TCB orders up more than 50% year over year while continuing hybrid development. [S10]

This evidence changes the question. Investors should not ask only whether hybrid bonding is ultimately better at fine pitch. They should ask how long improved TCB can satisfy HBM yield, throughput, gap, and thermal requirements; which layers or generations require bumpless interconnects; and whether hybrid adoption occurs across a full stack or only selected interfaces. A credible bridge can defer equipment revenue even when the terminal technology thesis is correct.

Kulicke & Soffa reinforces that conclusion. Its APTURA platform is being extended from heterogeneous logic toward memory, the company expects substantial FY2026 TCB growth, and it has accelerated hybrid-bonding development. Its hybrid position appears earlier than Besi’s, but customers benefit from qualifying alternatives and process variants. [S11]

TCB and hybrid bonding are not perfect substitutes. TCB can be economically superior when pitch, thermal performance, and stack height remain adequate because customers can reuse knowledge and capacity. Hybrid becomes more valuable as bump geometry, heat, and interconnect density constrain the package. The crossover is application-specific, which makes a single industry-wide adoption date analytically misleading.

Profit pools and barriers to entry

Industry profitability is not uniform. Mature wire bonding, standard die attach, molding, and commodity assembly processes face greater price competition. Exceptional economics accrue to tools that control yield at a technically constrained process step. Besi’s sustained gross margin above 60%, despite falling revenue from 2021 through 2025, is evidence that its mix contains differentiated processes. ASMPT and K&S have broader portfolios and generally lower consolidated margins, but their installed bases, customer access, and willingness to fund adjacent platforms make them credible competitors.

The main barriers are precision mechatronics, surface science, contamination control, software, throughput, applications engineering, customer qualification, service coverage, and access to leading process-development programs. A machine must repeat a placement thousands of times, track each die, integrate wafers and carriers, maintain cleanliness, and generate acceptable yield over a production campaign. The entry barrier is an operating system around the machine, not the machine frame alone. [S9][S12]

Qualification creates both switching cost and time-to-market risk. A customer must validate recipes, materials, reliability, automation, traceability, downstream compatibility, and economics. The process can take quarters or years. However, the same sophisticated customers deliberately qualify multiple sources to protect supply, increase bargaining power, and avoid technology dependency. Qualification slows competitive entry; it does not create permanent exclusivity.

A useful test of industry profitability is whether returns persist after competitors invest. Besi’s historical margins establish past differentiation, but future returns depend on keeping enough performance distance that customer productivity outweighs price. If multiple vendors achieve acceptable yield and throughput, bargaining power can shift rapidly even while the end market grows.

Supply-side capital cycle

The demand outlook has attracted a supply response. Besi is expanding hybrid capacity and service support; ASMPT is shipping and qualifying TCB and hybrid systems; K&S is increasing memory investment; Applied is integrating preparation and metrology; EV Group is improving hybrid-bonding overlay measurement; and Asian equipment ecosystems continue to mature. EV Group’s 2025 announcement of full-die overlay measurement and materially higher metrology throughput illustrates that competitive investment is occurring across the process flow, not just at the placement step. [S9][S10][S11][S25]

A larger industry profit pool can coexist with worse future supplier economics if too many vendors and customers build ahead of sustainable utilization. On the Q2 call, Besi’s CEO noted that semiconductor-equipment upcycles often run six to eight quarters before excess capacity emerges. That warning should carry more weight than an indefinite secular-growth extrapolation because it comes from management’s own operating history. [S6]

Orders can exceed final production need when customers secure scarce capacity, qualify redundant vendors, or accelerate spending before policy changes. Most Besi orders are not protected by software-like cancellation economics. Monitoring book-to-bill, lead times, deposits, customer acceptance, installed utilization, and cancellations is therefore more useful than treating every order as permanent demand.

The capital cycle has two layers. Customers invest in packaging capacity, while equipment suppliers invest in engineering, outsourced supply, service staffing, and demonstration tools. Customer overcapacity weakens orders; supplier overcapacity weakens pricing and absorption. Besi’s outsourced model limits owned-factory downside, but it cannot eliminate engineering commitments or supplier reservation costs.

Policy, China, and geographic risk

Export controls create opposing effects. Restricted access to leading-edge front-end nodes can encourage Chinese customers to use chiplets and advanced packaging, increasing attach intensity. The same rules can restrict equipment sales, components, technical support, or collaboration. Besi’s H1 report said direct tariff effects were limited and that Ex Works terms generally place tariff payment on customers; it also described China-for-China sourcing. Contractual pass-through does not eliminate economic incidence if tariffs or controls reduce customer investment. [S2]

The low-cost threat is therefore capability-based rather than wage-based. A domestic Chinese supplier does not need to match every CHAMEO specification globally to pressure Besi; it needs adequate performance for a sufficiently large protected domestic market. No authoritative public evidence available at the cutoff showed a mainland vendor matching Besi’s qualified 100-nanometre die-to-wafer production capability. Absence of disclosure is not evidence that the gap will persist.

China concentration also complicates the apparent hedge. Advanced packaging may help Chinese customers extend mature front-end nodes, but that strategic importance may make the tools more likely—not less likely—to become controlled. Meanwhile, localization can transfer learning toward domestic suppliers. The same policy force can raise near-term demand and reduce the long-term addressable market.

Verdict. Advanced packaging is a genuine structural growth market, but the timing and supplier economics are less certain than the TAM charts imply. Hybrid bonding has demonstrated technical and logic-production relevance; TCB’s documented HBM4 progress and the expanding supplier capital cycle are the central disconfirming evidence.

Competitive Position

The correct peer set

ASMPT is the closest broad equipment competitor because it participates in TCB, hybrid bonding, wire bonding, and multiple assembly processes with deep Asian customer access. Kulicke & Soffa is relevant through mainstream bonding, TCB, and an emerging hybrid platform. Applied Materials is a strategic partner and potential platform-level coopetitor: it contributes surface preparation, cleaning, metrology, and traceability to Kinex, while Besi contributes precision placement. EV Group is important in wafer-to-wafer and metrology workflows. ASML and ASM International are not direct competitors, but they are useful valuation benchmarks for scarce, high-margin European semiconductor-equipment franchises. [S8][S9][S10][S11][S25]

Foundries, memory producers, and OSATs are customers or qualification gatekeepers, not equipment peers. Comparing their margins or multiples directly with Besi would mix manufacturing economics with equipment economics.

Market share evidence and limitations

Besi reports approximately 49% share in die attach and 83% in advanced placement below 7-micron accuracy. Those figures come from company-sponsored market estimates and category choices matter: some TCB, wafer-to-wafer, or adjacent processes may fall outside the denominator. They should be treated as management claims. The financial evidence is more independent of classification: gross margin remained between 59.6% and 65.2% during 2021–2025 and reached 65.7% in Q2 2026. A commodity assembler would be unlikely to sustain that pricing and mix through a revenue contraction. [S1][S3][S4]

The moat mechanism has five linked components. First, precision, cleanliness, yield, and throughput must be delivered simultaneously. Second, co-development with leading customers creates tacit process knowledge. Third, qualification raises the time and failure cost of switching. Fourth, an installed base supports service response and recipe continuity. Fifth, outsourced manufacturing allows Besi to concentrate capital on engineering and applications rather than owned subassembly capacity.

Each component has an observable failure signal. If precision leadership erodes, Besi should lose processes of record or accept lower pricing. If customer knowledge is not proprietary, qualification cycles should shorten for competitors. If service does not reinforce the installed base, service mix and repeat orders should fail to grow. If outsourcing is only financial engineering, shortages or warranty problems should offset the low physical capital. The moat is therefore testable through share, margin, repeat orders, customer acceptance, and service economics.

Nature of competition and brand

Competition is based on production yield, accuracy, throughput, total cost of ownership, contamination control, software integration, service response, and qualification against a customer’s roadmap. [S9][S10][S12]

Sticker price matters less at the leading edge than the value of good dies and time to market. At mature process nodes, price and local support matter more. Besi can consequently have strong economics in hybrid placement while facing much harsher competition in conventional attach, molding, or plating.

Brand matters only as compressed evidence of production reliability, process knowledge, and service. A semiconductor manufacturer does not pay for the Besi name as a consumer pays for a luxury label. It pays because prior qualifications and installed experience reduce the probability of yield loss or a missed launch. Brand value could deteriorate quickly after repeated installation delays, weak field support, contamination incidents, or loss of process-of-record status.

Performance specifications must be compared on the complete process rather than one advertised dimension. A tool with better placement accuracy can lose if surface preparation, queue time, service, or throughput makes the total line less economic. Conversely, an integrated line can improve contamination control but reduce flexibility when customers frequently change products. Applied and Besi describe Kinex’s integrated controls as advantageous, but management acknowledged that stand-alone configurations can remain useful where flexibility matters. [S6][S9]

Switching costs and customer power

Switching costs are meaningful but application-specific: replacing a qualified bonder can require new recipes, materials validation, yield learning, traceability integration, operator training, spares, and reliability testing, while customers can still dual-source before high-volume production. [S10][S12]

The strongest switching cost arises after a production process has achieved acceptable yield and utilization. The weakest exists during evaluation, when the customer’s purpose is often to compare vendors. This is why 21 evaluation or adoption customers cannot be valued like 21 entrenched production customers.

Customer power remains substantial. The leading foundries, memory companies, and OSATs are large, technically capable, and economically motivated to qualify alternatives. Their own process knowledge can reduce supplier dependence, and a large order can materially affect Besi’s annual results. A vendor can hold a technical lead while still conceding price, support, or roadmap commitments to secure a strategic program.

The three memory manufacturers are particularly important because they control the qualification calendar and can compare hybrid bonding against improving TCB rather than merely comparing one hybrid tool with another. Besi’s access to all three is valuable. It is not equivalent to winning any of them at production scale.

Applied Materials relationship

Applied’s April 2025 market purchase of 9% of Besi is stronger evidence than an ordinary cooperation announcement. It aligns a major ecosystem supplier with Besi’s placement platform and supports the strategic relevance of integrated hybrid bonding. Kinex combines Applied’s surface preparation, cleaning, metrology, and die tracing with Besi’s placement, potentially reducing queue time and contamination between steps. Applied later said Kinex was being used by multiple logic, memory, and OSAT customers. [S8][S9]

The boundaries are equally important. Applied said it did not seek board representation and had no plans to purchase additional shares. No public agreement establishes minimum purchases, exclusive customer access, or an acquisition obligation. The stake validates technical alignment, not end-customer demand. Applied also participates broadly across advanced packaging and has incentives to preserve customer relationships beyond one partner.

The relationship can create both advantage and dependency. Integrated process learning may improve qualification and widen Besi’s customer access, but part of the system-level value belongs to Applied. Customers may also prefer modular solutions or insist on alternative preparation and metrology tools. Besi’s standalone economics must therefore remain competitive even if Kinex adoption is slower than expected.

Competitive contradiction: installed production versus claimed leadership

Besi’s strongest proof is logic production and its high-accuracy economics. ASMPT’s strongest rebuttal is installed production evidence in memory. ASMPT’s disclosure of more than 500 TCB systems, HBM3E high-volume performance, and HBM4 low-volume manufacturing is more commercially specific than a count of hybrid evaluations. [S10]

That does not establish that ASMPT will win hybrid bonding. It establishes that the transition path is contested and that TCB can generate customer learning, service relationships, and process-of-record status before hybrid is required. K&S provides another credible path through fine-pitch TCB and direct copper-to-copper development. [S11]

The likely industry outcome is coexistence rather than a single winner: different logic, memory, photonics, and chiplet applications may use different processes, with hybrid at the finest pitches and TCB surviving where its yield and economics remain sufficient. Besi can still win a large market in that outcome, but its valuation should not assume control of the whole bonding flow.

Customer concentration beneath the consolidated disclosure

No single customer exceeded 10% of 2025 revenue, yet a few leading programs can determine hybrid-system demand. The monitoring unit should be production customers by application, installed and accepted tools, repeat systems, and service activity—not only consolidated top-ten concentration. [S1][S2]

Inconsistent hybrid-customer baselines reinforce the point. A customer count can change because of cutoff, definition, joint platforms, or how one corporate group is counted. Units, revenue, utilization, and acceptance are harder economic evidence.

A durable competitive advantage would eventually show up in three reconciliations: hybrid and advanced-placement revenue should rise faster than the relevant market; gross margin should remain strong despite competitor qualification; and production tools should create recurring service and repeat-system demand. Without all three, market-share presentations alone cannot establish the size or durability of the moat.

Verdict. Besi has a real but application-specific moat in precision placement, evidenced by margins, reported share, and logic production. It does not have a monopoly over advanced packaging, and direct competitors possess stronger disclosed HBM production evidence than the headline hybrid-customer count suggests.

Growth History and Forward Opportunities

The historical record is cyclical rather than compound-like. Revenue peaked at €749.3 million in 2021, contracted to €578.9 million in 2023, remained below €610 million through 2025, and recovered to a reconstructed €733.8 million last-twelve-month level by June 2026. Orders led the recovery: €548.3 million in 2023, €586.7 million in 2024, €685.0 million in 2025, and €987.6 million over the twelve months through Q2 2026. [S1][S2]

The product outlook is strongest for hybrid bonding, advanced flip chip, TC Next, co-packaged optics, and other multi-die applications, with conventional mobile, automotive, and industrial demand providing a secondary cyclical recovery. [S4][S5][S6]

Hybrid bonding

Hybrid bonding is the largest upside platform because finer pitch can increase interconnect density and improve electrical and thermal performance. Besi’s roadmap moves from current 100-nanometre placement toward a 50-nanometre platform and longer-term sub-micron pitch. These are engineering targets, not customer yield guarantees. [S4]

Logic is the strongest current proof point. TSMC’s SoIC volume-production statement establishes commercial direct-bonding use, while Besi’s repeat logic orders and management commentary support a position in that ecosystem. Public evidence does not disclose Besi’s units, customer yields, or revenue by logic program, so concentration remains opaque. [S6][S13]

Memory is the larger swing factor. Higher HBM stacks increase vertical connections and make thermal management more difficult. Hybrid copper bonding is a credible solution, but memory yield is unforgiving because one defective layer can impair an expensive stack. Besi’s Q1 milestones contemplated 2026 evaluations informing 2027 adoption; the Q2 call preserved the opportunity but emphasized uncertainty. ASMPT’s HBM4 TCB production evidence means the first commercially meaningful hybrid insertion could be later, narrower, or layer-specific. [S6][S7][S10][S14]

Management’s own comments identify cost and yield as the principal hurdles. It said logic qualification took more than three years and acknowledged that detailed customer yields are not shared with Besi. The confidence that memory stacking should reach production yield is therefore a management hypothesis based partly on logic experience, not a disclosed memory result. [S6]

TC Next and advanced flip chip

TC Next is both an opportunity and a hedge. If memory customers retain microbumps longer, high-accuracy thermocompression can still participate. The hedge is incomplete because TCB faces stronger installed competitors and may have lower strategic scarcity than hybrid placement.

Advanced flip chip benefits from bridges, chiplets, power-management dies, and complex substrates. Management discussed further customer qualifications and potential repeat waves but did not give firm order size. Investors should distinguish qualified capability, evaluation shipments, and production capacity orders. [S6][S7]

A delayed hybrid transition could increase near-term TCB demand but reduce the valuation multiple investors assign to the revenue. The important variable is not simply whether Besi sells a different bonder; it is the margin, competitive intensity, unit requirement, and service attachment of that alternative.

Photonics and co-packaged optics

Co-packaged optics can create high-accuracy placement demand as optical engines move closer to processors and switches. Photonics also diversifies Besi away from one memory roadmap. The present revenue base is smaller and customer programs can be concentrated, so it should be treated as a credible option with current orders—not as a substitute for disclosed HBM economics. [S3][S4]

Management distinguished current pluggable-photonics demand from future co-packaged optics on the Q1 call. Existing pluggables can require more bonding steps, while co-packaged optics uses a different process and remains earlier. That separation prevents present photonics strength from being automatically capitalized as mature CPO revenue. [S7]

Services

Service and spares represented approximately 15% of 2025 revenue. Management believes the mix could approach 18–20% as hybrid tools expand because leading-edge systems require calibration, software, preventive maintenance, and field support. That aspiration is plausible but unproven. The company does not separately disclose contractual recurring service, transactional spares, upgrades, or service margin. [S1][S7]

A larger installed base should produce measurable evidence: service revenue should rise relative to systems across a downturn, field headcount should scale more slowly than installed units, and repeat customers should show higher lifetime contribution. Without those disclosures, service is a stabilizer, not an annuity valuation.

Long-term model requirements

The Investor Day model calls for €1.7–2.2 billion of revenue and a 45–55% operating margin. Management did not attach a specific year. From €733.8 million of reconstructed last-twelve-month revenue, the low end still requires more than doubling. It requires several applications to qualify, suppliers to support capacity, pricing to remain strong, and customer facilities to reach utilization. [S5]

The margin target requires gross margin to stay around the mid-60s while operating expenses grow much more slowly than revenue. That can occur under an outsourced model, as H1’s operating leverage demonstrates. It can also reverse if capacity is built ahead of demand or R&D remains high through a pause.

The long-term model should be decomposed rather than accepted as one forecast. Conventional systems must recover; logic hybrid must expand; at least part of memory must adopt; photonics and CPO must become material; and service must rise with the installed base. Failure in one component can be offset by another, but the low end is too large to be reached through evaluation tools alone.

Verdict. Besi has multiple credible growth vectors and does not require immediate HBM dominance merely to grow. The long-term target, however, requires production conversion across several platforms; evaluation breadth and roadmap relevance are leading indicators, not substitutes for revenue, yields, units, and cash.

Financial Quality

Five-year income statement

The figures reconcile the 2025 annual report’s five-year record with Company Financials statements. Dividend figures are presented by the earnings year for which they were declared, which differs from a cash-paid-in-year series. [S1][S22]

The table shows unusually resilient gross margins and substantial operating cyclicality. Revenue fell 21.1% from 2021 to 2025, while R&D more than doubled. Operating income fell 45.5%. This was not necessarily poor cost control: the company funded hybrid, TCB, and advanced-placement programs during a downturn. It does mean the next peak must support a permanently larger engineering base.

Earnings were near a cyclical high in 2021–2022, reached a recent cyclical low in 2025, and were recovering rapidly in H1 2026; whether the recovery becomes a structural multi-application step-up remains unproved. [S1][S2][S6]

H1 2026 acceleration

These are reported facts from unaudited interim statements. Q2 alone generated €249.9 million of revenue, €108.8 million of operating income, and €89.0 million of net income. The 43.5% Q2 operating margin shows how quickly profit scales when high-margin systems pass through an expanded R&D and commercial base. [S2][S3]

The Q3 guide of 10–15% sequential revenue growth implies approximately €275–287 million. At a 63–65% guided gross margin and broadly stable expenses, another sequential operating-profit increase is plausible. That is an analyst inference, not company profit guidance.

The growth quality is favorable in mix but not fully transparent in concentration. Management attributed the quarter to hybrid bonding, photonics, data-center, and mobile demand, yet revenue by application and customer is not disclosed. The 65.7% gross margin supports a premium mix; it does not reveal how repeatable that exact mix is.

ROIC: correcting the denominator problem

ROIC is exceptional but less precise than a single headline estimate suggests. Company Financials’ standardized reported-capital calculation shows ROIC declining from approximately 33.8% in 2021 to 26.3% in 2022, 20.8% in 2023, 20.7% in 2024, and 14.7% in 2025. This series should anchor cross-period comparison because it applies one methodology. [S22]

A cash-netted analyst calculation produces a much higher figure. Using after-tax EBIT and ending equity plus interest-bearing debt less cash and deposits gives roughly 34–39% for 2025, depending on whether average or ending capital is used. Applying the same logic to last-twelve-month June 2026 results produces approximately the high-50% range because NOPAT recovered while net operating capital remained small. These are formula-sensitive estimates, not company KPIs.

The difference is not cosmetic. Deducting €543 million of year-end 2025 cash and deposits from invested capital removes more than half the gross capital base. An asset-light business with supplier financing can also report very small working capital relative to its economic dependence on engineering and customers. The correct conclusion is that returns are strong and rebounding—not that a single 57% figure is uniquely true.

Research adjustment would lower economic returns further. Besi expensed €81.0 million of R&D in 2025 and capitalized €26.0 million of development. A full research-capital approach would retain some historical expensed engineering in invested capital and amortize it over an assumed useful life. Because useful life and program attrition are not disclosed, reporting a precise adjusted ROIC would create false accuracy. The direction is clear: conventional accounting understates the accumulated capital supporting today’s technology.

Peer ROIC comparisons also require care. Broader competitors combine mature wire-bonding, SMT, display, or other activities with advanced packaging, while ASML and ASM operate at different process steps. Besi’s returns are evidence of a differentiated niche and an outsourced model, not proof that every advanced-packaging tool earns similar economics.

IFRS, normalization, and accounting conservatism

Besi reports under IFRS and emphasizes reported revenue, operating income, net income, and margins rather than presenting a large recurring adjusted-profit bridge. That is favorable for comparability. Share-based compensation, amortization, and development capitalization still require economic analysis.

Share-based compensation was €16.4 million in 2025 and €10.1 million in H1 2026. It is a real employee cost even when settled in equity. Capitalized development moves cash investment below operating cash flow, so physical-capex-only free cash flow overstates owner earnings. Amortization is noncash in the current period but relates to prior investment and should not automatically be added back when technology must be continually refreshed. [S1][S2]

There was no material H1 accounting-policy change, and new standards had no material effect. That supports period comparability but does not resolve judgment around when development meets IFRS capitalization criteria. [S2]

The accounting is reasonably clean but not maximally conservative. Expensing most research depresses current earnings, while capitalizing qualifying development increases current operating profit and creates later amortization. A conservative analytical approach keeps share compensation in cost, deducts capitalized development from owner cash, and tests the development asset for economic—not merely accounting—obsolescence.

Cash conversion and working capital

FY2025 cash from operations was €178.1 million versus €131.6 million of net income. After €15.8 million of physical capex, €26.0 million of capitalized development, and €5.2 million of investment-property spending, a conservative owner-cash estimate was approximately €131.1 million. That nearly matched net income. [S1]

H1 2026 cash from operations was €140.7 million versus €140.6 million of net income. Deducting €4.1 million of physical capex and €11.8 million of capitalized development gives approximately €124.8 million of owner cash. Full-period conversion was therefore sound.

Net income and cash from operations were broadly aligned for FY2025 and H1 2026, but Q2 alone produced only approximately €47.7 million of operating cash against €89.0 million of net income because receivables and inventory rose during the shipment ramp. [S2][S3]

Trade receivables used €74.9 million of H1 cash and inventory used €14.3 million, partly offset by a €25.7 million increase in payables. Growth working capital is the benign explanation. Acceptance delays, customer concentration, or inventory built for orders that slip are the adverse explanations. The falsifier for the benign view is continued weak collection after revenue growth moderates.

Cash conversion should be judged over the order-to-acceptance cycle, not one quarter. A production ramp naturally requires components and customer credit, but a strong franchise should ultimately collect without structurally longer receivable days, inventory write-downs, or cancellation losses. The next two quarters therefore matter more than the H1 aggregate alone.

Balance sheet and liquidity

At June 30, cash was €349.1 million and deposits were €160.0 million. Interest-bearing debt was €345.1 million, principally the €350 million 4.5% notes due 2031, yielding €164.0 million of net cash. Total assets were €1.163 billion, equity €573.3 million, and the current ratio remained above four times. [S2]

The conversion of the 2029 notes reduced refinancing risk but transferred value through equity issuance. Balance-sheet solvency risk is low under ordinary cyclicality. That does not make the stock low-risk because most of the market value consists of expected future economic rents rather than net tangible assets.

Liquidity supports R&D and supplier commitments through a downturn, but it is small relative to the €15.3 billion equity value. Net cash can protect operations; it cannot protect a premium valuation from earnings-estimate revisions.

Off-balance-sheet and contingent obligations

The filings do not identify a material unconsolidated financing structure; disclosed economic obligations consist principally of leases, supplier and purchase commitments, warranties, taxes, employee obligations, and the practical requirement to support installed equipment. [S1][S2]

The most material contingent exposures are operational. Orders can be delayed or cancelled, revenue can depend on acceptance, warranty failures can require field work, and supplier capacity reserved for an upcycle can become underutilized. These risks do not appear as a large balance-sheet liability before they occur.

Capital intensity

Capital intensity is low in physical assets but high in engineering: H1 physical capex was below 1% of revenue, while expensed and capitalized development together exceeded €58 million. [S2]

This model creates excellent incremental economics when revenue grows. It also concentrates risk in employee retention, software, intellectual property, supplier coordination, and process relevance. A physical factory can often be repurposed; obsolete process knowledge may have little salvage value.

A supply-chain ramp can also create economic capital not visible in property, plant, and equipment. Supplier reservations, deposits, inventory, field-service hiring, and customer demonstration machines absorb resources even when final assembly is outsourced. Physical-capex intensity therefore understates the total cost of supporting the long-term revenue target.

Data reconciliation

The primary-file bridge produces last-twelve-month revenue of €733.8 million and EBITDA of approximately €301.6 million. A standardized Company Financials TTM payload showed €749.2 million and €309.4 million, which does not reconcile to FY2025 minus H1 2025 plus H1 2026. The report therefore uses the directly reproducible primary-report bridge and treats the provider series as a cross-check rather than overriding the filings. [S1][S2][S22]

Verdict. Financial quality is high, but the correct evidence is resilient gross margin, sound full-period cash conversion, net cash, and strong standardized ROIC—not an unqualified 50%-plus return number. Cyclicality, development capitalization, working-capital timing, and a larger fixed R&D base remain material qualifications.

Capital Allocation

Besi’s allocation philosophy combines heavy internal technology spending with a high earnings payout and opportunistic repurchases. It has avoided a recent transformative acquisition, which limits integration and balance-sheet risk, but distributions have sometimes exceeded current-period owner cash generation. [S1][S2]

Reinvestment and M&A

The present portfolio reflects older acquisitions of Datacon, Esec, Fico, and Meco-related businesses. Those transactions created today’s die-attach, packaging, and plating breadth, but public disclosure does not provide a reliable asset-level return schedule decades later. No material operating acquisition appears in recent 2024–H1 2026 cash flows, so the current investment case turns on organic R&D, customer qualification, and the Applied partnership rather than acquisition synergies. [S1][S2]

Recent reinvestment is mainly engineering. Expensed plus capitalized development reached approximately €107 million in 2025 and €58 million in H1 2026. That is strategically coherent because hybrid bonding, TCB, and photonics require simultaneous platform development. The return must ultimately appear in incremental gross profit and service cash flow; product roadmaps alone do not establish it.

The old acquisitions appear strategically successful at the portfolio level because Besi retains leading positions and high margins across the acquired technology families. That is not the same as proving their original purchase returns. Without acquisition consideration, follow-on investment, divestiture proceeds, and asset-level cash flows, any precise historical IRR would be invented.

Dividends and coverage

The €1.58 dividend for 2025 represented approximately 95% of reported EPS and a cash outflow of €125.4 million in H1 2026. The dividend has varied with earnings: €3.33 for 2021, €2.85 for 2022, €2.15 for 2023, €2.18 for 2024, and €1.58 for 2025. It is a high-payout cyclical distribution, not a fixed-income substitute. [S1][S17]

Dividends plus repurchases were approximately €255 million in 2025, exceeding €178 million of operating cash flow and approximately €131 million of conservative owner cash. H1 2026 dividends and ordinary treasury purchases also exceeded owner cash. Management financed part of the gap with accumulated liquidity. That can be rational if technology investment is fully funded and shares are undervalued; it reduces resilience if the cycle reverses.

A high payout also exposes shareholders to procyclicality. Dividends can be largest when equipment profits and the stock valuation are near a peak, then fall in a downturn. Investors should value the distribution as a residual policy, not capitalize the latest payment as a perpetual yield.

Repurchases and net dilution

Besi completed a €100 million program in October 2025 and began a €60 million program. Recent disclosed purchases occurred around €191–198 per share, near the controlled-date price. [S15]

The company is repurchasing shares, but H1 repurchases did not reduce the net share count because conversion of the 2029 notes released approximately 1.54 million treasury shares and increased shares outstanding excluding treasury by about 1.9%. [S2][S16]

This distinction matters for valuation. Part of the buyback offsets convertibles and employee awards rather than shrinking the claim on the business. A program executed around 70 times trailing earnings also requires exceptional future cash flows to be accretive relative to retaining cash.

Repurchases should be judged by net fully diluted shares and purchase price, not authorization headlines. A small reduction in ordinary dilution can still be useful, but it is different from retiring a durable percentage of the company below intrinsic value.

Insider issuance and transactions

Share-based compensation was €16.4 million in 2025 and €10.1 million in H1 2026. The April 2026 AFM filing increased CEO Richard Blickman’s holding by 42,223 shares at a reported value of zero, taking his position to approximately 1.53 million shares. That was an award or vesting event, not an open-market purchase and should not be presented as insider buying. [S1][S2][S19]

Equity compensation can align employees, but its economic cost survives even when cash flow statements add it back. The relevant shareholder test is whether per-share owner cash grows after all awards, convertible settlements, and repurchases.

Governance, incentives, and motivation

The Management Board consists only of CEO Richard Blickman, creating unusual key-person and succession concentration. His ownership of roughly 1.9% aligns a meaningful portion of his wealth with shareholders, but alignment does not remove incentives to emphasize a premium technology narrative. [S1][S19]

The short-term incentive weights financial measures including net margin, return on average equity, and operating-cash-flow conversion, with nonfinancial measures forming the remainder. The long-term program has used profitability and relative total shareholder return. These metrics encourage margin and cash discipline but can also reward cyclical peaks or relative share performance before long-duration projects generate cash. [S1]

Shareholders rejected the advisory 2024 remuneration report at the 2025 AGM, while all 2026 agenda items—including the revised remuneration report and dividend—received majority support. This demonstrates governance responsiveness but also confirms that pay design had become a material shareholder concern. [S17][S18]

Supervisory directors receive fixed compensation and the company reports them as independent under the Dutch governance framework. Applied’s 9% stake carries no board seat. These checks mitigate but do not remove dependence on a sole executive director.

Management’s behavior suggests confidence in the technology roadmap and a preference for returning excess liquidity rather than building a large cash hoard. The adverse interpretation is that buybacks at premium prices and a near-full earnings payout prioritize visible shareholder distributions over optionality. The evidence does not establish motive; it identifies the financial trade-off.

Verdict. Capital allocation has been disciplined on recent M&A and appropriately heavy in R&D, but shareholder distributions have exceeded conservative owner cash and buybacks have not prevented dilution. Price discipline, succession, and measurable returns on development are the issues to monitor.

Changes and Headwinds — Last Two Years

The business environment changed materially during 2025–2026: AI-related advanced packaging became a major order driver, Applied purchased 9%, hybrid evaluations broadened, TCB competitors demonstrated memory production progress, Asian localization increased, and export-control uncertainty intensified. [S2][S8][S10]

From downturn to order acceleration

In 2024, revenue recovered modestly but operating profit declined as R&D increased. In 2025, revenue fell another 2.7%, while orders rose 16.8% to €685.0 million and Q4 bookings accelerated. Those orders converted into H1 2026 revenue and earnings. The sequence indicates both an external AI-capex upturn and company-specific product wins. [S1][S24]

Results reflect both external demand and internal execution: AI and customer packaging investment expanded the opportunity, while Besi’s qualifications, R&D, product mix, supplier management, and outsourced manufacturing determined order capture and margin. [S2][S4]

The company should not receive all the credit for an industry capex wave, but neither is the recovery pure beta. Orders rose faster than the recent assembly-market estimates, and gross margin stayed in the mid-60s, supporting company-specific execution.

Applied relationship and platform integration

Applied’s April 2025 investment moved the relationship from ordinary co-development to co-development plus meaningful equity alignment. The October Kinex release then described an integrated flow across cleaning, preparation, metrology, traceability, and placement. The change improves system-level relevance but creates no disclosed demand guarantee. [S8][S9]

Applied’s statement that multiple logic, memory, and OSAT customers use Kinex strengthens the evidence beyond a laboratory announcement. It still does not disclose units, acceptance, production revenue, or yields. The commercial status remains more advanced than a concept but less transparent than a production-installed-base schedule.

Management commentary changed in precision

On the Q1 2026 call, management said all three leading memory manufacturers had Besi bonders and described evaluation milestones intended to inform 2027 adoption. On the Q2 call, management acknowledged divergent industry expectations, said Q3–Q4 outcomes would determine which applications and how much hybrid bonding was used, and declined to disclose system units. [S6][S7]

This is a narrative softening on timing, not evidence of a program loss. It raises the probability that memory revenue moves between periods. ASMPT’s disclosed HBM4 TCB progress makes that schedule risk economically credible rather than merely rhetorical. [S10]

The Q2 transcript also provided useful caution from management itself. Qualification hurdles are cost and yield; detailed memory yields are not disclosed to Besi; logic qualification took more than three years; and equipment cycles can create excess capacity. These statements are more decision-useful than broad claims about inevitable hybrid adoption. [S6]

Markets, facilities, and management

Markets shifted toward computing and AI, customers expanded global packaging capacity, and Besi added hybrid capacity and Asian service support while retaining an outsourced production footprint. [S2][S5]

Management discussed using Vietnam for additional mainstream capacity, preserving Malaysia for wafer-level equipment, and expanding support in Taiwan and Korea. China-for-China sourcing responds to localization and policy risk. Leadership did not change: Blickman remains the sole Management Board member.

The company increased theoretical hybrid-bonder capacity and prepared field support for a larger installed base. Capacity readiness is strategically necessary but not itself demand. The risk is a mismatch between supplier and service investment today and customer acceptance several quarters later.

Tariffs, controls, and currency

Besi reported limited direct H1 tariff impact, assembly equipment initially receiving favorable treatment, and Ex Works terms that generally place tariff payment with customers. Those contractual facts do not eliminate indirect demand effects. Export restrictions, customer profitability, component availability, retaliation, and regional duplication can all change capex. [S2]

A weaker US dollar can reduce euro-reported gross margin because sales and costs have different currency exposures. Currency is a translation and transaction driver, not evidence of weaker process competitiveness, unless it triggers price concessions or changes sourcing economics.

Accounting and financing

No material accounting-policy change was reported for H1 2026, and new standards had no material effect. [S2]

The major balance-sheet change was economic rather than accounting: conversion of the 2029 notes reduced debt, increased equity, released treasury shares, and diluted the outstanding share base. The governance change was shareholder rejection of the 2024 pay report followed by approval of the revised 2025 report.

Verdict. The last two years improved demand, strategic validation, and product breadth while increasing expectations and competitive investment. The biggest adverse change is not current revenue; it is that documented TCB progress and less precise memory timing have made the path to the long-term model more contestable.

Risk Analysis

The most plausible stock-decline path is an expectations reset: HBM qualification slips, 2027 earnings estimates fall, and the valuation moves toward a less scarce equipment multiple. [S6][S10][S21]

The interaction matters more than any single variable. A 20% reduction to €6.58 consensus EPS yields approximately €5.26. At 22 times earnings, that would imply about €116 per share—roughly 39% below the controlled price—without Besi becoming unprofitable. This is scenario arithmetic, not a forecast.

A second downside path does not require memory disappointment. Logic and photonics orders could remain healthy while an ordinary equipment digestion cycle reduces bookings, working capital reverses, and the market stops capitalizing undated long-term targets. The outsourced model would protect gross capital better than a vertically integrated factory model, but fixed R&D would still create reverse operating leverage.

A catastrophic investment loss could result from a Taiwan conflict or broader Asian production interruption, sustained exclusion from major customers, severe hybrid-process yield failure, or technological displacement combined with a premium starting valuation. [S1][S2][S12]

Financial distress is not the central risk. Net cash, profitable legacy products, multiple process families, and service revenue make a literal total loss remote. A near-total economic loss would require simultaneous technological obsolescence, exclusion from major Asian markets, collapse of service value, and failure to resize costs. [S1][S2]

A severe 60–75% equity drawdown is more plausible than bankruptcy. Such a case could leave a solvent, profitable company valued on mature assembly-equipment earnings. The balance sheet protects creditors and operating continuity more than it protects the equity multiple.

Additional risks should remain separate. Currency can lower reported margin without impairing customer yield. Tariffs can be paid by customers but still reduce demand. Applied’s stake may encourage strategic speculation but is not a bid. The €60 million repurchase is too small relative to a €15 billion market value to stabilize a fundamental rerating.

Risk monitoring should prioritize changes that connect directly to economics. A weaker euro-reported margin accompanied by stable units and pricing differs from a margin decline caused by competitive concessions. A receivable build during 50% growth differs from aging after shipments flatten. An evaluation delay differs from a competitor winning the process of record. Aggregating these into one generic “execution risk” would obscure the thesis.

Verdict. Ordinary insolvency risk is low, but security risk is high because qualification timing, estimate revisions, and valuation multiply one another. The bear case requires delay and normalization, not business failure.

Valuation Discussion

Current reconstructed valuation

At €189.25 and approximately 80.82 million shares outstanding excluding treasury, equity value was about €15.30 billion. Deducting June net cash of €164.0 million gives enterprise value near €15.13 billion. [S2][S20]

Last-twelve-month values are reconstructed as FY2025 minus H1 2025 plus H1 2026:

• Revenue: €591.3 million − €292.2 million + €434.7 million = €733.8 million.
• Net income: €131.6 million − €63.6 million + €140.6 million = €208.7 million.
• EBITDA: approximately €206.8 million − €97.5 million + €192.3 million = €301.6 million.

The resulting multiples are approximately 20.6 times sales, 50.2 times EBITDA, and 73.3 times earnings. These are analyst calculations reconciled to primary reports, not issuer-presented valuation metrics. [S1][S2]

A standardized Company Financials TTM payload used different TTM revenue and EBITDA values that did not reconcile to the reported FY-minus-prior-H1-plus-current-H1 bridge. The primary-report reconstruction is used because its periods and arithmetic are explicit. This discrepancy is a reason not to copy a provider multiple without rebuilding its denominator. [S22]

Consensus and its embedded ramp

September consensus called for €993 million of 2026 revenue, €1.37 billion in 2027, €4.33 of 2026 EPS, and €6.58 in 2027. At €189.25, the calculated P/E ratios are approximately 43.7 and 28.8 times. The consensus page’s displayed multiples use a different price snapshot, so they should not be copied without recalculation. [S21]

Consensus requires 2026 revenue to rise about 68% from 2025 and 2027 revenue another 38%. H1 results and Q3 guidance make a large 2026 increase credible, but the 2027 forecast still requires conversion of current orders and new production demand.

The estimate path is more important than the point estimate. If 2027 revenue arrives in 2028 because qualification shifts, the eventual technology thesis can remain intact while the present value falls and the stock rerates. Investors are underwriting both magnitude and calendar.

Peer context

The peer set divides into direct competitors and quality benchmarks. ASMPT and K&S are direct process competitors; ASML and ASM International are valuation benchmarks rather than substitutes. Company Financials trailing data, adjusted to September 11 prices, indicate approximate trailing P/Es of 39 times for K&S, 54 times for ASML, and 46 times for ASM International, versus Besi near 73 times. Approximate EV/EBITDA ratios were around 28, 42, and 31 times, versus Besi near 50. Accounting, currencies, and business mix differ, so the spread is more useful than the last decimal. [S22]

Besi deserves a premium to K&S because its gross margins and hybrid positioning are stronger. It is harder to justify a premium to ASML without assuming faster earnings growth: ASML has a larger service base, deeper process monopolies, and more visible demand. Besi’s smaller addressable market and customer-program concentration raise forecast risk even if its incremental margin is higher.

ASMPT’s public valuation is less cleanly comparable because of Hong Kong reporting conventions and its broader portfolio. Its operating evidence is nevertheless essential: a lower-margin competitor can still delay Besi’s earnings by extending the life of TCB.

Own-history context also argues against calling the post-peak stock conventionally cheap. The share price has repeatedly capitalized advanced-packaging milestones before they appeared in revenue, and the 2026 peak produced extreme trailing multiples. The current price is below that extreme but remains far above a normal cyclical multiple on reported earnings. [S20][S22]

Scenario analysis

The scenarios use approximately 81 million shares. They explicitly separate operating assumptions from terminal valuation and are estimates, not targets.

The base case is not conservative in absolute terms. It assumes 2027 revenue more than doubles the 2025 level and that Besi retains a scarcity premium. The bear case still assumes a profitable and growing company. The bull case requires multiple production programs, not merely a larger evaluation count.

The terminal multiple is as important as earnings. A 30-times multiple can be justified only if investors view the earnings as structurally higher and less cyclical than the historical record. If the revenue mix remains transactional and concentrated, a lower multiple may be appropriate even after earnings grow.

What the current price embeds

A required-return method reaches a similar conclusion. To earn 10% annually before dividends over roughly four years, €189.25 would need to become about €277–284. At a 30 times terminal P/E, that requires roughly €9.3–9.5 of EPS, or approximately €755–770 million of net income on 81 million shares. At a 40–45% net margin, revenue would need to be approximately €1.7–1.9 billion.

The current price therefore embeds much of the low-to-middle portion of management’s €1.7–2.2 billion model, continued exceptional margins, and a durable premium multiple. [S5]

The market is right that hybrid bonding is technically relevant, logic adoption is real, and Besi can scale earnings rapidly. The fragile assumptions are calendar timing in memory, sustained mid-60s gross margin under competition, production conversion beyond a few programs, and the terminal multiple.

Reinvestment and dilution

All scenarios deduct capitalized development as reinvestment. The bear case assumes R&D does not fall in proportion to revenue; the base case assumes operating leverage; the bull case assumes supplier capacity and service scale without proportionate fixed-cost growth. Holding the share count near 81 million may be optimistic if employee issuance or future convertible settlement exceeds buybacks.

Dividends add cash return but do not fix an earnings shortfall. They reduce enterprise cash and should not be counted as value creation merely because the payout ratio is high.

The valuation should be updated with production evidence rather than price momentum. A higher price could become more attractive if accepted memory systems and service economics raise sustainable cash flow. A lower price could remain unattractive if orders collapse or process leadership is lost. Price and evidence must be evaluated together.

Verdict. The valuation is coherent only if a substantial part of the long-term model materializes. The business merits a premium, but the present price offers limited protection against a one-generation memory delay or a normal equipment multiple.

Variant Perception

Published consensus expects a rapid rise to approximately €1.37 billion of revenue and €6.58 EPS in 2027. The implied narrative is that current orders convert, hybrid and advanced placement broaden, and operating leverage remains powerful. [S21]

The most useful investor questions now concern which HBM generation adopts hybrid bonding, how many evaluation customers become high-volume production customers, what sustained yields and units are achieved, and how concentrated the order book is by program. [S6][S7][S10]

Other important questions concern the service mix, customer acceptance terms, supplier commitments, the inconsistent hybrid-customer baseline, and why management will not disclose system units. These are more decision-useful than asking whether AI packaging grows in the abstract.

Strongest bull case

The bull case is that attach intensity is structurally underestimated. Logic hybrid bonding is already in volume production in the ecosystem; HBM stacks become taller; co-packaged optics brings additional precision placements; and Besi combines the leading die-placement platform with Applied’s integrated preparation and metrology. The outsourced model can translate a broad adoption wave into operating margins approaching management’s 45–55% target. [S5][S9][S13][S14]

Waiting for disclosed HBM revenue may also miss the rerating because qualification and equipment orders precede production. H1 orders, repeat logic demand, photonics, and Q3 guidance are evidence that the opportunity is broader than one memory program.

The bull also argues that coexistence is enough. Besi need not displace every TCB tool if hybrid wins the finest-pitch logic interfaces, selected HBM generations, CPO, and consumer processors. Because the company’s physical capital needs are low, a smaller-than-monopoly share of several expanding niches can still generate exceptional incremental profit.

Strongest bear case

The bear case does not require hybrid bonding to fail. TCB can remain adequate through HBM4 and parts of HBM4E; memory qualification can take longer; customers can dual-source; and ASMPT or K&S can win enough share to reduce Besi’s economics. ASMPT’s disclosed HBM3E high-volume and HBM4 low-volume TCB performance is the strongest evidence for this view. [S10][S11]

Logic and photonics can grow while remaining too small to support a €15 billion equity value. The enlarged R&D base creates reverse leverage when orders normalize, and a 42% drawdown does not create cheapness when trailing multiples remain exceptional.

The bear further argues that the market conflates technological necessity with fiscal-year revenue. Hybrid bonding may become necessary eventually, yet customers can extend TCB, limit hybrid to selected interfaces, or insert it one generation later. A two-year delay can materially reduce present value without disproving any engineering claim.

Load-bearing assumptions

1. Several of the reported hybrid customers progress beyond evaluation and small development orders.
2. Material memory-equipment orders arrive for 2027–2028 production rather than shifting toward HBM5 or decade-end.
3. Gross margin remains above approximately 62% despite currency, competition, and bargaining power.
4. Besi retains production share as ASMPT and K&S qualify alternatives.
5. R&D produces scalable cash earnings faster than dilution and distributions consume value.

Each assumption has a different evidentiary threshold. Customer-count growth tests access; accepted repeat tools test production conversion; service revenue and utilization test installed economic value; gross margin tests bargaining power; and owner cash per share tests whether the technology creates shareholder value.

Positioning and factor context

The factor model snapshot was unavailable, so no numerical beta, alpha, momentum, quality, value, or sector exposure is reported. The price record qualitatively resembles a high-duration, high-volatility AI semiconductor-equipment exposure. That description is an inference from returns and catalysts, not a factor-model result.

Applied’s stake reduces free float and validates strategic alignment but does not create takeover support. The buyback is immaterial relative to market capitalization. Positioning cannot be reliably inferred from either fact.

Falsifying each side

The bull weakens if customer counts rise without accepted repeat systems, memory insertion moves past HBM4E, gross margin falls because of competition rather than mix, or service fails to scale with the installed base. The bear weakens if multiple memory manufacturers place production-scale hybrid orders, non-memory revenue sustains the consensus ramp, and cash conversion remains strong through the capacity build. [S2][S6][S10]

Verdict. The differentiated view is that technical relevance is more firmly established than production timing. The market may be treating eventual necessity as if it fixed the fiscal year in which revenue arrives.

Fact vs. Interpretation

The most serious prior analytical overreach was treating one cash-netted ROIC estimate as the definitive return measure. Standardized reported-capital ROIC is materially lower. The most serious narrative risk is treating evaluation breadth as production evidence.

The retrieved company memory did not contain a prior BESI-specific thesis suitable for confirmation or falsification. Most retrieved learnings concerned biotechnology accounting, banks, utilities, software, or unrelated capital structures and were not transferred. The only mechanisms revalidated here were broader analytical disciplines: retain development spending in owner-cash analysis, reconcile gross repurchases with dilution, and separate strategic alignment from contractual demand. Their applicability is supported by BESI-specific evidence rather than assumed from analogy. [S1][S2][S8][S16]

Verdict. Reported financial momentum, cash, and competitor production disclosures deserve the greatest weight. TAM, customer counts, and long-term targets deserve less weight because they depend on definitions, forecasts, and undisclosed qualification economics. [S2][S5][S10]

Open Questions

1. How many of the 21 hybrid customers have accepted systems for high-volume production rather than development or evaluation? [S2][S6]
2. What are sustained customer throughput, uptime, overlay, and multi-die-stack yields rather than dry-cycle specifications? [S4][S12]
3. Which HBM generation and which stack interfaces first require hybrid bonding at economic scale? [S10][S14]
4. How much of the €987.6 million last-twelve-month order total is cancellable, concentrated, or scheduled beyond 2027? [S2]
5. How does management reconcile the 15-versus-18 year-end 2025 hybrid-customer references? [S1][S2][S4]
6. What portion of service is contractual recurring support versus spares, tooling, upgrades, and field work? [S1][S7]
7. What supplier capacity and purchase commitments support the long-term target, and how quickly can they be reduced? [S2][S5]
8. What portion of capitalized development relates to hybrid, TCB, or legacy platforms, and what triggers impairment? [S1][S2]
9. What is the succession plan for a one-person Management Board? [S1]
10. What valuation discipline governs repurchases after distributions exceeded current owner cash? [S1][S15]
11. What explains the standardized TTM data’s failure to reconcile to the disclosed FY/H1 bridge? [S1][S2][S22]
12. Can Applied or Besi disclose whether the Kinex users are evaluating, qualifying, or producing at high volume? [S9]

Verdict. Missing evidence is concentrated in production units, yield, acceptance, concentration, and capital discipline—the variables that connect technical relevance to per-share cash value.

What Must Be True

Bull test

The constructive thesis requires observable production conversion:

• Q3 revenue reaches the guided €275–287 million range while gross margin remains within 63–65%, and cash collection improves after the H1 receivables build. [S2][S3]
• At least one major memory program generates production-linked hybrid orders, repeat systems, or disclosed generation inclusion by early 2027. [S6][S7]
• Logic, co-packaged optics, advanced flip chip, and TC Next remain strong enough that one HBM decision does not determine annual growth. [S4][S6]
• Gross margin remains above approximately 62% as ASMPT and K&S expand competing platforms. [S3][S10][S11]
• Service revenue rises with the installed base and demonstrates recurring contribution rather than only transactional spares. [S1][S7]
• Expensed and capitalized development produces revenue and owner-cash growth exceeding dilution. [S1][S2]

Bull falsifiers: no production-scale memory order by mid-2027; repeated customer-count growth without unit or revenue conversion; gross margin below 60% for two quarters without a clearly temporary currency effect; falling orders despite continued customer packaging capex; or disclosed process-of-record losses to a competing platform.

Bear test

The cautious thesis requires deterioration in timing or economics, not merely price volatility:

• HBM hybrid insertion moves beyond HBM4E toward HBM5 or decade-end. [S6][S10][S14]
• 2027 revenue or EPS consensus declines because evaluations fail to become production orders. [S6][S21]
• TCB continues to meet pitch, yield, stack-height, and thermal requirements for additional memory generations. [S10][S11]
• Competitors win production acceptance or cause pricing and gross-margin concessions. [S10][S11]
• Receivables, inventory, cancellations, or acceptance delays show that orders overstated cash demand. [S2]
• Management continues premium-priced distributions while owner earnings and net cash weaken. [S1][S15]

Bear falsifiers: multiple memory customers place volume hybrid orders; repeat production systems broaden beyond the existing logic base; HBM and non-HBM revenue sustain consensus while gross margin stays in the mid-60s; or disclosed service economics establish a materially larger recurring profit stream.

Monitoring hierarchy

The monitoring hierarchy is production qualification, accepted repeat units, revenue and margin conversion, cash collection, competitive wins, and only then TAM. That order prevents eventual technical relevance from being mistaken for a dated earnings forecast. [S2][S6][S10][S12]

The highest-quality positive signal would be a named-generation production order accompanied by repeat units or accepted capacity. The highest-quality negative signal would be a process-of-record loss or a customer confirming that improved TCB remains adequate for another generation. Changes in presentation customer counts, stock-price volatility, or undated TAM estimates rank below those signals.

Verdict. Besi can validate the upside through a small number of measurable milestones. Until then, the evidence supports confidence in the technology franchise but not confidence that the current valuation compensates investors for timing risk.

Linked sources: Besi financial reports, Besi investor presentations, Besi webcast archive, Applied Materials strategic investment, Applied Materials Kinex announcement, ASMPT H1 2025 results, Kulicke & Soffa memory-platform update, imec hybrid-bonding demonstration, TSMC SoIC, Samsung HBM4E, and Euronext BESI listing.

Public source appendix

• S1: Besi Annual Report 2025 — Primary company annual report; published 2026-03-16; Five-year financial highlights; business, customer and geographic disclosures; financial statements and notes; governance and remuneration sections
• S2: Besi Half Year 2026 Report — Primary company unaudited interim report; published 2026-07-23; Operating review pp. 4–7; statements pp. 9–12; accounting, equity, development, and geographic notes pp. 13–16
• S3: Besi Q2 2026 Results and Webinar Materials — Primary company earnings release and presentation; published 2026-07-23; Q2 and H1 revenue, orders, margins, earnings, cash flow, and Q3 guidance
• S4: Besi Investor Presentation — August 2026 — Primary company investor presentation; publication date unavailable; Company mix, market-share estimates, addressable-market forecasts, customer adoption, product specifications, and roadmap
• S5: Besi Increases Long-Term Financial Targets at 2026 Investor Day — Primary company investor-day release; published 2026-06-18; Undated long-term revenue target of €1.7–2.2 billion and operating-margin target of 45–55%
• S6: Company Financials — Q2 2026 Earnings-Call Transcript — Third-party transcript reconciled to primary company materials; published 2026-07-23; July 23, 2026 prepared remarks and Q&A on hybrid units, HBM timing, applications, TCB, service, capacity, China, and cycle duration; reconciled to issuer webcast and release
• S7: Company Financials — Q1 2026 Earnings-Call Transcript — Third-party transcript reconciled to primary company materials; published 2026-04-24; April 24, 2026 prepared remarks and Q&A on three memory customers, evaluation milestones, 2027 adoption, hybrid orders, TC Next, Vietnam, and service capacity
• S8: Applied Materials Announces Strategic Investment in Besi — Primary strategic-partner release; published 2025-04-14; 9% market-purchased stake, collaboration history, no board representation, and no plans for additional purchases
• S9: Applied Materials Unveils Kinex Integrated Die-to-Wafer Hybrid Bonding System — Primary strategic-partner product release; published 2025-10-07; Kinex integration of preparation, cleaning, metrology, traceability, and Besi placement technology; use by multiple logic, memory, and OSAT customers
• S10: ASMPT 2025 Interim Results Announcement — Primary competitor unaudited interim report; published 2025-07-24; Advanced-packaging review pp. 2–3: 500-plus TCB installed base, HBM3E high-volume requirements, HBM4 low-volume manufacturing, order growth, and hybrid qualification
• S11: Kulicke & Soffa Expands Memory Solutions Portfolio — Primary competitor release; published 2026-03-24; APTURA TCB platform, expected FY2026 TCB growth, memory positioning, and accelerated hybrid development
• S12: Imec Demonstrates Die-to-Wafer Hybrid Bonding at 2-Micron Pitch — Independent semiconductor research institute release; published 2024-05-29; Pitch, overlay, Kelvin and daisy-chain electrical yields; particle, CMP, singulation, accuracy, and throughput requirements
• S13: TSMC SoIC System on Integrated Chips — Primary foundry technology page; publication date unavailable; SoIC process description, sub-10-micron pitch, and 3-nanometre stacking volume-production statement
• S14: Samsung Begins Shipment of HBM4E Samples — Primary memory-manufacturer release; published 2026-05-29; HBM4E 12-layer sample shipment, HBM4 mass-production statement, and customer-schedule qualification
• S15: Besi September 2026 Share-Repurchase Update — Primary company capital-allocation release; published 2026-09-03; August 27–September 3, 2026 purchases, prices, volumes, and €60 million program reference
• S16: Besi Announces Early Redemption of 2029 Convertible Bonds — Primary company financing release; published 2026-05-05; €175 million principal, €115.50 conversion price, conversion deadline, and June 4 redemption
• S17: Besi 2026 Annual General Meeting Results — Primary company governance release; published 2026-04-23; Approval of all agenda items and €1.58 cash dividend
• S18: Besi 2025 Annual General Meeting Results — Primary company governance release; published 2025-04-23; Rejection of the advisory 2024 remuneration report and approval of other resolutions
• S19: AFM Director Transaction Record — Richard Blickman — Primary regulator insider-transaction record; published 2026-04-23; April 23, 2026 zero-value acquisition of 42,223 shares and resulting holding
• S20: Company Financials — BESI Daily Prices and Euronext Listing — Exchange listing and third-party market dataset; published 2026-09-11; Daily OHLC history through September 11, 2026, including €189.25 close and June 15 intraday high; listing and ISIN reconciled to Euronext
• S21: BESI Analyst Consensus Estimates — Third-party consensus dataset; published 2026-09-12; September 12, 2026 revenue, EBITDA, EPS, net-cash estimates, analyst distribution, and target-price snapshot; valuation multiples recalculated at €189.25
• S22: Company Financials — BESI and Peer Fundamentals — Third-party standardized financial and market dataset; published 2026-09-12; Exchange-qualified EURONEXT:BESI profile; 2021–2025 statements and standardized profitability ratios; TTM payload; daily prices; comparable KLIC, ASML, and ASM fundamentals; Besi values reconciled to primary reports
• S23: Besi Financial Reports Archive — Primary company filing archive; publication date unavailable; Annual reports 2021–2025, interim reports, and XBRL packages available during the trailing 60-month review
• S24: Besi Q4 and FY2025 Results — Primary company earnings release; published 2026-02-19; FY2025 revenue, orders, margins, earnings, cash flow, dividend proposal, and AI-related order mix
• S25: EV Group Achieves Hybrid-Bonding Overlay-Control Breakthrough — Primary adjacent-equipment vendor product release; published 2025-09-08; Full-die overlay measurement and claimed metrology-throughput improvement for die-to-wafer hybrid bonding