Camtek
Capital-equipment OEM: sells inspection/metrology tools (Eagle, Hawk, Golden Eye, MicroProf families) to memory makers and OSATs, with recurring service, spare parts, and software/upgrades attached to a growing installed base. Lumpy, order-driven hardware revenue with a smaller annuity tail.
The thesis on this name
State of the Memory Supercycle
The dominant HBM inspection name — estimated >40% share of HBM bump/pillar/TSV inspection, serving all three memory makers (ainvest/Kavout, 2025). Its Eagle/Emerald platforms do the 2D/3D inspection of micro-bumps, pillars and TSVs that every HBM stack requires, so content scales with HBM die-stack height (8-hi → 12-hi → 16-hi) and unit volume simultaneously — a double-lever. FY25 revenue ~$497M with double-digit FY26 growth guided and Q2'26 at $129–131M on strong HBM/OSAT backlog (PRNewswire, Q1'26). A pure-play on the HBM advanced-packaging layer specifically, mispriced relative to the mega-cap WFE names because the cyclicality scares generalists — but stack-height roadmap gives it secular content growth on top of the cycle.
State of the Memory Supercycle
The pure-play HBM advanced-packaging inspection toll-booth — TSV/hybrid-bonding/interposer inspection that gets MORE intensive at HBM4/HBM4e; the most HBM-levered small-cap, hence the most volatile.
State of the Memory Supercycle
Pure-play TSV/hybrid-bonding inspection; HBM4 is more metrology-intensive. High torque, small size; widen on >25% drawdowns.
Earnings, margins, COGS & capex
Camtek rode the AI/HBM packaging boom from $429M (FY24) to $496M (FY25, +16%), but Q1 FY26 was nearly flat YoY (+2.5%) as the HBM3E build-out paused ahead of the HBM4 transition. Management frames FY26 as H1-soft / H2-strong: it has booked $260M+ of HBM orders from two manufacturers for 2026–2027 and guides H2 FY26 revenue >25% above H1. Margins are healthy (~51% non-GAAP gross) but operating margin compressed ~600bps YoY on R&D/sales investment, and GAAP earnings were distorted by an ~$89M one-time loss on a 2025 convertible-note repurchase.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~50¢ is cost of goods and ~28¢ operating expense, leaving ~22¢ of operating profit (~29¢ net).
Revenue trend
Margins
down slightly (was 52.1% Q1 FY25; ~51.6% FY25)
roughly flat (~50.5% FY25)
down (31.5% Q1 FY25; 30.0% FY25) on opex investment ahead of HBM4
down (32.6% Q1 FY25)
seasonally weak quarter; structurally strong FY
COGS structure
COGS is dominated by purchased materials and outsourced components — optics, cameras/sensors, precision stages, motion controllers, light sources, and computing/AI hardware — assembled in-house; plus manufacturing labor, system integration, warranty, and field-service cost. Gross margin sits ~50–52% because the value is in proprietary optics + image-processing/AI defect-classification software, not raw bill-of-materials; richer 3D-metrology and HBM-specific configurations carry above-average margin, while heavy China volume and competitive 2D-inspection deals pull mix the other way.
Capex
Low and discretionary — funds cleanroom/demo-lab capacity, R&D equipment for the HBM4/hybrid-bonding roadmap (Eagle, Hawk, Golden Eye, MicroProf), application labs near customers, and IT. No fab; Camtek is an asset-light tool integrator, so the bigger capital story is the ~$580M cash + securities war-chest funded by the 2030 converts, available for M&A/buybacks.
Latest earnings
Q1 FY2026 was a slight revenue beat (above the ~$120M guide midpoint) and roughly in line on non-GAAP EPS at $0.70, and the shares fell anyway on the flat YoY growth and ~600bps of non-GAAP operating-margin compression. Nothing has been reported since. The August 10 print will be judged on two things the quarter itself cannot settle: whether Q2 lands inside the $129-131M guide (consensus ~$130.2M sits at the midpoint, so this is close to pre-determined) and whether management reaffirms 'H2 more than 25% above H1' now that SK Hynix has said its HBM4 ramp is running slower than expected and pushing revenue into the second half.
Q2 FY26: revenue $129-131M (revenue only - no margin or EPS guidance was given). H2 FY26: 'revenues expected to grow by over 25% versus first half 2026 revenues', which on the Q1 actual plus the Q2 guide midpoint implies a full year of at least ~$566M, or +14% YoY. Consensus sits at $571.26M for FY2026 and $698.07M for FY2027 - i.e. the Street is taking the guide at face value and adding almost nothing to it. Order-side support: $260M+ of 2026-2027 HBM bookings from two manufacturers, plus a separate, incremental $105M+ announced 2 June 2026 ($55M from a tier-1 OSAT and $50M+ of Hawk systems from a leading HBM player) all for 2027 delivery.
- Q1 FY26 revenue
- $121,659K (+2.5% YoY) - the flattest quarter since the AI ramp began
- Non-GAAP operating margin (Q1 FY26)
- 25.5%, from 31.5% a year earlier; GAAP 22.4% from 27.6%
- Opex growth vs revenue growth (Q1 FY26)
- +20.8% opex (R&D +38.2%) against +2.5% revenue
- FY2025 China revenue
- $243,935K = 49% of total (filed 20-F); Asia Pacific region ~91% of total
- FY2025 Korea revenue
- $36,888K, down 68% from $117,135K - from 27.3% of revenue to 7.4%
- FY2025 top customer
- 11% of total revenues (2024: 15%/10%/10%) - concentration is lower than the entry implied
- New orders announced 2 June 2026
- $105M+ ($55M tier-1 OSAT + $50M+ Hawk to an HBM maker), all for 2027 delivery
- Net cash / dilution
- ~$361.9M net cash; 0.00% 2030 converts add 5,123K diluted shares (~11% of basic)
Growth drivers
- HBM unit + capacity growth (SK Hynix ~62% share, Micron, Samsung all expanding) and rising inspection/metrology steps per HBM stack
- HBM4/HBM4e transition — more die per stack (12-/16-Hi), tighter overlay, and eventual hybrid-bonding all add inspection/metrology insertions per wafer — content-per-stack tailwind
- AI advanced-packaging beyond HBM — CoWoS/interposers, chiplets, panel-level packaging, and 2.5D/3D logic driving OSAT + foundry tool demand
- New products Eagle and Hawk (3D metrology + 2D inspection for chiplets/HBM/hybrid bonding) expected to roughly double in revenue in FY26 (mgmt; fact)
- Installed-base service/spares/software annuity compounding as units ship
- Booked backlog: $260M+ of HBM orders for 2026–2027 from two manufacturers (mgmt; fact)
Bull & bear
The order book kept building while the multiple halved. Camtek has $260M+ of 2026-2027 HBM bookings plus a separate $105M+ of 2027-delivery orders announced in June, ~50% of revenue is AI-linked (filed), the balance sheet holds ~$362M of net cash funded by zero-coupon paper, and the stock now trades ~42x TTM non-GAAP, down from ~52x - with consensus targets sitting above spot rather than below it.
- The 2027 book is being built in public: $260M+ of HBM orders from two manufacturers for 2026-2027, plus an incremental $105M+ announced 2 June 2026 - $55M from a tier-1 OSAT and $50M+ of Hawk systems from a leading HBM player, all for 2027 delivery (filed 6-K). That June announcement is the first hard dollar validation of the Hawk platform the bull case rests on
- Content-per-stack, not just unit volume: inspection and metrology insertions rise with each HBM generation (12-/16-Hi, HBM4/4e, and eventually hybrid bonding), so Camtek can grow served steps even on flat HBM units. SK Hynix's reported ~$31B capex plan is the spend behind that
- Margins compressed by choice, not by price: the Q1 squeeze is opex-driven (R&D +38.2% YoY) against flat revenue, with gross margin down only ~90-110bps. If the guided H2 volume arrives, the operating leverage reverses mechanically
- Genuinely asset-light and unlevered: $22.5M of FY2025 capex (4.5% of revenue), $141.9M of FY2025 operating cash flow, ~$362M net cash and 0.00%-coupon converts due 2030 - the war chest costs nothing to carry and funds M&A on the company's own timetable
- Less customer-concentrated than the story suggests: the largest FY2025 customer was 11% of revenue, down from 15% in 2024, across an installed base of over 3,000 systems (filed 20-F). Buyer pricing leverage is a weaker argument than the concentration story suggests
- Valuation reset with no reported deterioration: ~$132 vs ~$154-168 in late June, ~42x TTM non-GAAP against ~52x, and a consensus target range of $160-$220 that now sits entirely ABOVE spot
The single claim the whole entry rests on - H2 more than 25% above H1 - is still unverified 8 days before it gets tested, and the largest HBM maker has just told the market its HBM4 ramp is running late. Underneath it, the filings show FY2025's growth came from China (+84%) while Korea collapsed 68%, so the 'HBM toll-booth' framing and the 49% China export-control exposure are the same risk wearing two labels.
- The filed geography does not match the narrative: FY2025 growth was China +$111.4M (+84%) and Asia Pacific +$35.2M against Korea -$80.2M (-68%, from 27.3% of revenue to 7.4%). Attribution is by shipment destination so HBM tools can land outside Korea - but you cannot look at this table and cleanly separate 'AI/HBM toll-booth' revenue from 'China' revenue, and 49% of the company is China
- The H2 guide now has a named, dated headwind: SK Hynix reported slower-than-expected HBM4 shipment ramps pushing revenue into H2 2026. Camtek's 'over 25% above H1' depends on those same tools landing on schedule; a quarter of slippage moves the step-up into 2027 and breaks the year
- Q1 was the warning shot and it has not been answered: +2.5% revenue growth, non-GAAP operating margin down ~600bps to 25.5%, operating income down 17% in absolute terms, and operating cash flow of $3.1M against a $40.9M receivables build. Three months later there is still no newer data point
- Sector de-rating on AI-capex sustainability, not on Camtek: over $1 trillion of semiconductor and AI-hardware market cap erased between 24 and 29 July, with CAMT down 7.5% and 9.4% on consecutive sessions. If the market is right that AI infrastructure spend is peaking, the most HBM-levered small-cap in the equipment group is the wrong place to be
- Export-control exposure is unchanged and unhedged: 49% of FY2025 revenue shipped to China and ~91% to Asia Pacific (filed 20-F). One rule change is a step-down in half the business, and there is no geographic diversification in the FY2025 numbers to cushion it
- Quiet dilution and a standing supply path: the 0.00% converts add 5,123K shares to the diluted count (~11% of basic), and in June the registration-rights agreement with Priortech and Chroma ATE was extended to 2031 with the demand threshold cut to 1% of share capital. Neither is urgent
- Competitive overhang is unresolved: KLA, Applied Materials, Onto Innovation and Nova are all larger and all pushing into advanced-packaging inspection and metrology, and if HBM4 stays on microbumps the hybrid-bonding content step-up slips right
What it is worth
Peer-relative multiple (non-GAAP forward P/E vs Onto/KLA/Nova) cross-checked with a reverse-DCF sanity test on the implied HBM-content growth.
~$110–$130
HBM digestion / margin compression persists, multiple de-rates toward ~30–35x on flat-to-low-teens growth.
~$165–$185
H2 FY26 delivers >25% H1 and the HBM4 order book converts; ~45–50x on re-accelerating mid-teens-plus growth (near consensus).
~$220–$260
HBM4/hybrid-bonding content step-up plus Eagle/Hawk ramp drive 25%+ growth and margin recovery, sustaining a ~55–60x premium.
At ~$170 the stock trades ~52x non-GAAP TTM EPS (~$3.26 FY25) — a premium to the broad equipment group that requires HBM4 to re-accelerate growth back to the 20%+ range; the GAAP ~170x P/E is distorted by the one-time $89M convert-repurchase loss and should be ignored. Consensus 12-mo target ~$166 (range $120–$190) sits at/below spot, i.e. priced for execution.
SWOT
Strengths
- Leading share in HBM/advanced-packaging optical inspection and 3D metrology — serves all three memory makers (SK Hynix, Micron, Samsung); estimated >40% HBM-inspection share (estimate)
- High, stable margins (~51% non-GAAP gross, ~30% FY op margin) for a hardware OEM, anchored by proprietary optics + AI defect-classification software
- Fortress balance sheet — ~$850M liquid assets, net cash ~$362M, funded cheaply by 0.00% converts due 2030 — optionality for M&A/buybacks
- Content-per-wafer tailwind — each HBM generation (12-/16-Hi, HBM4/4e, hybrid bonding) adds inspection/metrology insertions, structurally growing its served steps
- Fresh, well-received product cycle (Eagle, Hawk, Golden Eye, MicroProf) extending it into chiplets and hybrid bonding
Weaknesses
- Extreme HBM/memory-capex concentration — revenue tracks a single, notoriously cyclical end-market; a customer capex pause hits hard (Q1 FY26 +2.5% is the warning shot)
- ~49% of FY25 sales from China (fact) — heavy export-control and demand-policy exposure
- Customer concentration in a handful of memory makers + leading OSATs/foundries gives buyers pricing leverage
- Lumpy, order-driven revenue and back-half-loaded years make quarterly results volatile and hard to model
- Small-cap scale vs KLA/Applied/Onto — less R&D and balance-sheet firepower if a large incumbent pushes into HBM inspection/metrology
Opportunities
- HBM4/HBM4e ramp (2026–2027) and the eventual move to hybrid bonding — Camtek's highest content-per-stack inflection
- Expansion in CoWoS/interposer, chiplet, and panel-level packaging as AI packaging diversifies beyond HBM
- Deepening the software/AI defect-classification and metrology annuity to lift recurring revenue and margin
- M&A with the converts war-chest to add metrology adjacencies and reduce single-market dependence
- Share gains at OSATs (ASE, Amkor) and foundries as advanced packaging migrates from memory to logic
Threats
- HBM digestion/overbuild or an AI-capex pullback that delays memory-maker tool orders
- HBM4 staying on microbumps longer than expected — delays the hybrid-bonding content upgrade thesis
- KLA, Applied Materials, Onto Innovation, or Nova pushing harder into advanced-packaging inspection/metrology
- US/Netherlands export controls tightening on China shipments — a direct hit to ~49% of sales
- Memory-maker in-housing of inspection steps or pricing pressure from concentrated buyers
Moats, dependencies & bottlenecks
Moats
Proprietary optics + AI defect-classification software validated on live HBM lines High while it stays a generation ahead; erodes if an incumbent matches detection on HBM4/hybrid-bonding defects Detecting buried copper-to-copper voids and sub-micron overlay is the hard part; recipes are co-developed and tuned per customer line.
Qualification lock-in / switching cost at memory makers and OSATs re-qualifying inspection on a running yield-critical line is slow and risky, so incumbents stay Once a tool is in the process flow gating yield, swapping it risks the ramp; this is the stickiest part of the franchise.
grows with units but is a smaller share than hardware A compounding recurring tail that cushions hardware lumpiness.
KLA/Applied/Onto dwarf Camtek on R&D and balance sheet Camtek wins on focus and HBM-specific application depth, not scale; this is the soft flank.
High through the HBM4/4e/hybrid-bonding roadmap More inspection insertions per generation is a moat-adjacent growth engine, not a defensive barrier.
Dependencies
The dominant demand driver; a capex pause or HBM overbuild directly throttles orders — visible in Q1 FY26's flat YoY.
Concentrated geographic exposure subject to US/allied export controls and China demand policy.
Camtek is a derivative of the AI-compute capex cycle; an AI-spend slowdown flows straight through.
sensors/cameras, precision stages, compute components from suppliers Specialized optical and motion components; supply tightness or pricing hits COGS and lead times.
HBM4 adopting hybrid bonding / higher metrology intensity on schedule The content-upgrade thesis depends on the roadmap; HBM4 staying on microbumps longer delays the step-up.
Advantages
- Purest public-equity exposure to HBM/advanced-packaging inspection — no diluting legacy segments
- Design-ins at all three memory makers plus leading OSATs/foundries
- Software/AI defect-classification depth tuned on live HBM yield lines
- Net-cash fortress (~$362M) funded by 0.00% converts — cheap optionality for M&A/buybacks
- High, resilient margins (~51% non-GAAP gross) for a hardware OEM
- Fresh, adopted product cycle (Eagle, Hawk) extending into chiplets and hybrid bonding
Weaknesses
- Single-end-market (memory/HBM) concentration amplifies cyclicality
- ~49% China revenue exposure (fact) — high regulatory tail risk
- Operating-margin compression in H1 FY26 (non-GAAP op margin 31.5%→25.5% YoY)
- Lumpy, back-half-loaded revenue is volatile and hard to forecast quarter to quarter
- Small-cap scale vs KLA/Applied/Onto on R&D and balance sheet
- Premium valuation leaves no margin for an HBM digestion miss
Bottlenecks
- Customer order timing — back-half-loaded years and lumpy HBM capex make near-term revenue hostage to a few buyers' ramp schedules
- Pace of the HBM4/hybrid-bonding transition — the content-per-stack upgrade only pays off when customers actually move the process
- Export-control regime on China shipments — a policy change can gate ~half of revenue overnight
- R&D/scale ceiling vs KLA/Applied/Onto — staying a generation ahead on defect detection with a fraction of their budgets
- Specialized optical-component and skilled-integration supply for ramping system output
Top signals & trends
Top signals
$260M+ already booked for 2026–2027; pace of follow-on orders confirms or breaks the H2 re-accel.
The whole 'soft H1 is just digestion' thesis hinges on this delivering; a miss reprices the stock.
Opex outran flat revenue in Q1 FY26; watch whether the H2 volume restores leverage.
~49% of sales; a tightening headline is the single biggest downside catalyst.
Validates the chiplet/hybrid-bonding expansion beyond core HBM 2D inspection.
Hybrid bonding is the content-intensity upgrade; a delay pushes the step-up into 2027+.
Trends
Primary structural tailwind; SK Hynix ~62% HBM share, Micron/Samsung expanding.
More insertions per stack = content growth independent of unit volume — the core toll-booth thesis.
Buried copper-void/overlay inspection is harder and higher-value; timing is the variable.
Diversifies demand beyond HBM toward OSATs and foundries over time.
Direct risk to ~49% of FY25 sales; a structural overhang.
The same lever that drove +36% FY24 can flatten growth fast, as Q1 FY26 (+2.5%) showed.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Optical-component & precision-lens makers (e.g. ZEISS, Edmund Optics) High-resolution optics are the core of the inspection/metrology head.
Sony, Teledyne) High-speed line/area sensors that capture defect images at throughput.
Precision motion / stage & controller vendors Sub-micron wafer-handling stages and motion controllers for scan accuracy.
NVIDIA, Intel, AMD/Xilinx) AI/compute hardware runs the real-time defect-classification pipeline inside the tools.
Contract manufacturers / machined-part & subassembly suppliers Outsourced fabrication of mechanical subassemblies that Camtek integrates in-house.
HBM leader (~62% share); largest HBM inspection/metrology customer base.
US memory maker scaling HBM3E/HBM4 — a core HBM customer.
Memory + foundry; HBM and advanced-packaging tool buyer.
CoWoS/advanced-packaging foundry — interposer/2.5D inspection demand.
Largest OSAT — bumping, packaging, and test inspection insertions.
#2 OSAT; advanced-packaging and chiplet inspection customer.
Closest US-listed peer — advanced-packaging metrology/inspection (Dragonfly), also HBM/AI-packaging levered; the most direct head-to-head.
Inspection/metrology giant (~25% overall share); dominant in front-end, increasingly active in advanced packaging — the scale threat.
Broad equipment leader with packaging and metrology (incl. eBeam) initiatives; vast R&D and balance sheet.
Israeli metrology peer (dimensional/materials/chemical); adjacent rather than direct in HBM optical inspection but overlapping in advanced-packaging metrology.
Inspection specialist (mask/EUV-adjacent); a niche competitor in select inspection segments.
Japanese inspection/metrology suppliers competing in specific packaging and process-control niches.