BESI — Research Report (2026-07-19)

TL;DR

• What it is: Dutch maker of chip assembly machines — the tools that stack and connect finished chips — and the world leader in hybrid bonding, the key technique for building AI chips in 3D.
• My take: one of the highest-quality monopoly-like assets in semiconductors, at a price (~134× trailing earnings) that already prepays most of the 2030 dream — study now, buy only in tranches on weakness.
• Q1-26 orders €269.7M, +104.5% YoY (the AI-packaging demand inflection is real, not a story).
• FY-25 FCF €162M on €591M revenue (~27% free-cash-flow margin — a rare cash machine).
• Trailing P/E ~134× vs European semis ~58× (you pay today for earnings expected years out).
• 2030 revenue target raised to €1.7–2.2bn vs €591M in 2025 (management wants ~3× the business).
• Strongest bull point: near-monopoly in hybrid bonding just as logic, memory (HBM) and optics all start adopting it — plus real takeover interest from Lam Research.
• Strongest bear point: memory makers may delay hybrid bonding for HBM (KBC downgraded to Reduce on exactly this), while ASMPT, Hanmi and Samsung-Hanwha race to catch up.
• Watch next: Q2-26 results (late July) — does the +30–40% sequential revenue guide hold, and do HBM makers commit to hybrid bonding for next-gen memory?

1. Business overview

So what: BESI sells the picks and shovels for the "stack chips instead of shrinking them" era — few machines, very high margins, sold to the world's biggest chipmakers.

BE Semiconductor Industries (Euronext Amsterdam: BESI) designs assembly equipment: machines that pick up finished silicon dies, attach them to substrates or to each other, and package them into usable processors. Its product families are die attach (the bulk of revenue), packaging, and plating. Customers are chipmakers (TSMC, Intel, Samsung, SK Hynix — not all disclosed by name), OSATs (outsourced assembly factories, mostly in Asia), and automotive/industrial electronics makers.

The crown jewel is hybrid bonding: joining two chips copper-to-copper with no solder bumps, enabling far denser 3D stacks. This is the technique behind next-generation AI processors and, eventually, high-bandwidth memory (HBM — the stacked memory feeding AI chips). BESI ships these systems in co-development with Applied Materials, and per its Q1-26 release saw "particular strength in hybrid bonding, mobile and photonics applications" (Besi Q1-26 release).

Geographically most revenue ships to Asia (subcontractors and foundry-adjacent assembly), which was the growth driver in 2025 as "Asian subcontractors for AI-related datacenter and photonics applications" offset weak mobile/auto/industrial demand (FY-25 release). Customer concentration is high in any given year — advanced-packaging capex is decided by a handful of giants.

2. Financials (5 years)

So what: revenue has gone sideways through a brutal chip-cycle trough, but margins stayed extraordinary and cash kept flowing — and the order book just exploded.

2021–22 figures from the FY-22 release (2021 derived from its stated −3.5% / −14.8% declines — labeled estimate); 2023 from the FY-23 release; 2024–25 incl. cash flow from the FY-25 release. Blanks = not in the sources fetched; update in place next quarter.

Revenue CAGR 2021→2025 is about −5.7%/yr — that is the semiconductor downcycle, not decay: mobile/auto demand collapsed while AI-related demand grew underneath. One flag: FY-25 net income fell 27.7% while EBIT fell only 11.5% — the gap sits below the operating line (tax/financial items); cause not verified here, check the annual report.

The inflection is in orders, not yet revenue. FY-25 orders +16.8%; then Q1-26 orders €269.7M, +104.5% YoY, revenue €184.9M (+28.3%), net income €51.6M (+63.8%), and guidance for Q2-26 revenue up another 30–40% sequentially with gross margin rising to 64–66% (Q1-26 release).

Balance sheet: €543M total cash at end-2025 against roughly €507M of debt (mostly convertible notes — bonds that can turn into shares; derived: cash minus stated net cash), leaving a small net cash position of €36M, which grew to €103.3M by Q1-26. BESI pays out nearly everything: proposed dividend €1.58/share (95% payout ratio) plus €82M of buybacks in 2025.

3. Key metrics — and why each matters here

4. Unit economics

So what: the business runs on selling a small number of very expensive, very high-margin machines — and the mix shift to hybrid bonding raises both price and margin per unit.

BESI's one controlling number is gross margin per system, and the structural driver is mix: hybrid bonders are far more complex (and expensive) than mainstream die bonders. Management's own guidance shows the mechanism — Q2-26 gross margin guided up to 64–66% because hybrid bonding, AI and photonics shipments rise in the mix (Q1-26 release).

At the June 2026 investor day, management sized hybrid-bonding revenue at roughly €476M in 2026 alone — approaching the size of the entire FY-25 company — within a cumulative opportunity of 1,150–2,200 hybrid-bonding systems by 2030 (investor-day release, TradingView summary). Rough division of those targets implies a seven-figure-euro price per system (estimated — BESI does not disclose ASPs).

5. Value-chain position: bottleneck or commodity?

So what: today BESI is the bottleneck — the equity-research §0 playbook case (HBM → bonding equipment) in its purest form — but the moat is being assaulted from three directions.

Chip progress is shifting from shrinking transistors (lithography, ASML's domain) to stacking chips (advanced packaging). Whoever owns the critical stacking tool sits at a duopoly-or-better choke point. BESI is "widely regarded as the leading company in hybrid bonder technology" (36kr industry analysis), reinforced by co-development with Applied Materials, which bought a 9% stake to become BESI's largest shareholder (Yahoo Finance, Bits&Chips).

Pricing power over BESI's customers is real but conditional: TSMC/Intel/Samsung have no equivalent-precision alternative today — that is what 63%+ gross margins prove. The challengers: ASMPT plans to ship its second-generation hybrid bonder to HBM customers in Q3-26 (ASMPT, 36kr); Hanmi Semiconductor and a Samsung–Hanwha alliance are building Korean alternatives (36kr); ASML is reportedly eyeing hybrid bonding (TrendForce); and one research house asks whether partner Applied Materials is "betting on hybrid bonding, or hedging it" (HyperFRAME Research).

The niche itself is small but compounding fast — hybrid-bonding equipment estimated at ~$169M (2025) growing ~22%/yr toward ~$1.26bn by 2035 (InsightAce); TrendForce-cited work sees TCB + hybrid bonding leading a $1.3bn back-end equipment expansion by 2030 (Edge AI & Vision). BESI's whole thesis is that this niche becomes the mainstream assembly method.

6. Ownership, management, M&A angle

So what: a founder-CEO with no successor plus a strategic 9% shareholder makes BESI a live acquisition target — which supports the price but is not a thesis.

Richard Blickman has run BESI since 1995. No succession plan has ever been presented, which Il Sole 24 Ore flags as a driver of acquisition speculation (Il Sole 24 Ore). Reuters reported in March 2026 that BESI fielded takeover interest — talks with suitors including Lam Research began mid-2025, broke down early 2026 amid US–EU political tensions, but interest persists (Reuters via WKZO, TrendForce).

Applied Materials' 9% strategic stake (April 2025) doubles as partnership glue and a seat at any M&A table. Shareholder-friendliness is demonstrated, not promised: 95% payout ratio, and 2025 buybacks at an average €118.26 — less than half today's price, i.e. well-timed (FY-25 release). Insider-transaction detail (AFM filings) was not checked here — flagged as a gap, not assumed clean.

7. Valuation

So what: on every trailing measure BESI is priced for near-perfection; it's only "cheap" if the 2030 targets land near the top of the range — or someone pays a takeover premium.

Share price €251.20 (2026-07-15, Lightyear); market cap ~€20bn (≈€22bn on July 4, before a ~14% one-month pullback on hybrid-bonding doubts — De Belegger, StockAnalysis, MarketScreener NL).

P/E is the right lens here (profitable, net-cash balance sheet; EV/EBITDA would tell the same story, and EV/Sales is for pre-profit names). The honest reading: the trailing multiple is near-meaningless at a cyclical earnings trough, but even the forward ~56× assumes the Q2 guide lands and keeps compounding. Sanity check against management's own targets: €2.2bn revenue at a 45–55% EBIT margin (top of the investor-day range) implies roughly €1bn EBIT and €800M+ net income — today's ~€20bn cap is ~25× that best-case 2030 outcome. A strategic buyer (Lam/Applied) could justify a premium for the choke-point asset; a purely financial buyer at these multiples could not. Consensus (29 analysts: 19 buy/overweight, 7 hold, 3 sell) sits only ~17% above the price (ChartMill).

8. Bull / base / bear — named views

So what: the street agrees on the technology leadership and disagrees almost entirely on timing (memory adoption) and on what's already in the price.

• Bull — ING (Marc Hesselink): recent hybrid-bonding doubts are "temporary headwinds rather than a structural threat"; consensus bulls (14 buy + 5 strong buy, high PT ~€421) see hybrid bonding adopted across logic, HBM and co-packaged optics per the raised 2030 targets (AlphaSpread, investor-day release). Takeover interest from Lam Research is a free option (Reuters via WKZO).
• Base — Berenberg (Trion Reid), Hold, PT €240: raised from €140 after the order inflection but stayed at Hold — the growth is real and largely priced in (iex.nl). Morningstar-tracked fair-value estimates moved to €220 from €170 on similar logic (Morningstar).
• Bear — KBC Securities (Thibault Leneeuw), Reduce, PT €225: downgraded even while raising the target — positive on hybrid bonding in logic, "more cautious regarding the memory market" (De Aandeelhouder, iex.nl). Degroof Petercam (Michael Roeg) is also at Reduce with a €270 target despite noting BESI "stays ahead of competition" (iex.nl); Seeking Alpha published a valuation-driven rating downgrade (Seeking Alpha).

Catalysts: Q2-26 results, late July (the 30–40% QoQ guide is the nearest test) · HBM4 hybrid-bonding adoption decisions at SK Hynix/Samsung · ASMPT's Q3-26 hybrid-bonder shipments (first real competitive data point) · any revival of Lam/Applied takeover talks.

What would prove the bull thesis wrong: memory makers extending current bonding methods (TCB) another generation, pushing hybrid-bonding HBM volume past 2028; ASMPT or a Korean rival winning a marquee hybrid-bonding order at spec; book-to-bill breaking below 1× while the stock still trades above 50× forward earnings.

📚 What this company teaches

1. The bottleneck beats the brand: nobody outside the industry knows BESI, yet every AI-chip roadmap runs through its machines — the drill-down (macro trend → value chain → choke point with duopoly structure) finds these names before the crowd does, exactly like HBM → bonding equipment did.
2. Orders lead, revenue lags, price leads both: BESI's revenue fell for four years while orders and the share price inflected first — in equipment cyclicals, book-to-bill and backlog are the forward instruments; the income statement is the rear-view mirror.
3. A trough P/E of 134× and a peak P/E of 10× can both be traps: cyclical multiples must be read against where in the cycle earnings are — the same company looks absurdly expensive at the earnings bottom and deceptively cheap at the top.

Self-test: BESI trades at ~134× trailing but ~56× forward earnings while its sector averages ~58×. Which single operational metric from this report would you track each quarter to judge whether that forward multiple is being earned rather than merely hoped for — and why that one?

📖 Glossary

• P/E (price-to-earnings) — share price divided by profit per share; how many years of current profit you pay for the company.
• Forward P/E — the same ratio using expected next-year profit instead of last year's.
• EV/EBITDA — whole-company price (equity + debt − cash) versus raw operating profit; the standard deal multiple.
• EV/Sales — whole-company price versus revenue; used when there is no profit yet to measure.
• FCF (free cash flow) — cash from operations minus equipment spending; the money truly left over each year.
• OCF (operating cash flow) — cash generated by day-to-day business before investment spending.
• Gross margin — revenue minus direct product cost, as a % of revenue; a proxy for pricing power.
• EBIT / operating income — profit from operations before interest and tax.
• Net cash — cash minus all debt; positive means the company could repay every loan today.
• Payout ratio — the share of profit paid out as dividend.
• Convertible notes — bonds the holder can swap into shares; cheap debt with dilution risk attached.
• Book-to-bill — new orders divided by revenue in the same period; above 1 means demand is outrunning shipments.
• Backlog — orders received but not yet delivered or billed.
• CAGR — compound annual growth rate; smoothed per-year growth over a multi-year span.
• Advanced packaging — assembling multiple chips into one high-performance unit instead of shrinking a single chip further.
• Hybrid bonding — joining chips copper-to-copper with no solder, enabling ultra-dense 3D chip stacks.
• TCB (thermo-compression bonding) — today's mainstream chip-stacking method using heat, pressure and tiny solder bumps; hybrid bonding's incumbent rival.
• HBM (high-bandwidth memory) — stacks of memory chips wired vertically; essential to AI accelerators.
• 2.5D packaging — placing processor and memory side-by-side on a silicon bridge; today's dominant AI-chip package.
• Co-packaged optics / photonics — building light-based data links directly into the chip package for faster data movement.
• Die / die bonder — a "die" is one cut chip from a wafer; a die bonder is the machine that places and fixes it.
• OSAT — outsourced semiconductor assembly-and-test firms; the contract factories of chip packaging.
• Foundry — a factory that manufactures chips designed by others (e.g. TSMC).
• Lithography — printing circuit patterns onto silicon with light (ASML's domain); its scaling limits are what make packaging the new frontier.

Estimate-quality labels: 2021 figures derived from stated YoY changes in the FY-22 release; per-system prices estimated from investor-day totals; debt figure derived from total cash minus stated net cash; all other numbers from company releases or the cited data providers. Insider-transaction (AFM) check not performed — open item. Not financial advice; personal research-system output. Update in place at Q2-26 results.