Ibiden
B2B advanced-manufacturing supplier; sells high-layer-count package substrates to CPU/GPU/ASIC makers and OSATs, increasingly under multi-year customer-financed capacity contracts (advance payments); secondary ceramics segment sells DPF/GPF filters and SiC to automakers.
Earnings, margins, COGS & capex
Two-engine company that has effectively become an AI-substrate pure-play at the margin. FY ended March 2026 net sales JPY 416.2B (+12.7%), company-reported operating profit JPY 62.0B (14.9% margin, +30.3% YoY, ~JPY 1B above the JPY 61B guidance), net income JPY 63.7B (net income slightly exceeds OP and jumped ~89% YoY, boosted by investment-related/one-off gains). Growth and virtually all incremental profit come from the Electronics segment (IC package substrates for AI CPUs/GPUs); the Ceramics segment (auto DPF/GPF) is shrinking. Capex is enormous and mostly customer-financed via advance payments, so free cash flow sits near zero during the build-out.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~68¢ is cost of goods and ~17¢ operating expense, leaving ~15¢ of operating profit (~15¢ net).
Revenue trend
Margins
rising from ~27.7% (FY-Mar-2024) as AI-substrate mix improves
recovering off a ~12-13% trough (FY-Mar-2024/25) but still below the ~17-18% peak of FY-Mar-2022/23
inflated by investment/one-off gains; not a clean run-rate
improving
up sharply (+65.9% OP YoY)
declining (-36.8% OP YoY)
COGS structure
COGS ~68% of sales (gross margin ~31.6%). Cost base is capital-and-materials heavy: cleanroom substrate fabs, copper, resin/prepreg, and Ajinomoto ABF build-up film (the defining input for FC-BGA). Rising layer counts and larger body sizes raise unit cost but also unit price; yield/warpage control on large AI-accelerator bodies is the key cost-and-differentiation lever. Japan-based cost base benefits from a weak yen on export pricing.
Capex
Structurally elevated. Feb 3, 2026 board-approved ~JPY 500B electronics capex over FY-Mar-2026 to FY-Mar-2028: ~JPY 220B (phase 1) at the Gama plant (Ogaki, Gifu) with mass production from FY-Mar-2027, and ~JPY 280B at the new Ono plant (Ono-cho, Gifu), which began AI-server substrate production in Oct 2025 at roughly half building utilization and ramps to full production by end of FY-Mar-2027. Goal: ~2.5x generative-AI-server substrate capacity by FY-Mar-2028 vs FY-Mar-2026. Policy is to sign customer contracts first and collect advance payments (~JPY 92.1B on B/S) sized to the investment, de-risking the build.
Latest earnings
Full FY-Mar-2026 beat guidance on profit: company-reported operating profit JPY 62.0B vs the JPY 61B forecast; net sales JPY 416.2B (just shy of the ~JPY 420B forecast); net income JPY 63.7B, well above the more conservative ~JPY 37B net-income forecast on investment gains
FY-Mar-2027 guidance (issued with FY26 results): net sales ~JPY 500B and operating profit ~JPY 90B (assumes ~JPY 150/USD), net income ~JPY 65B, electronics-led (Electronics guide ~JPY 330B sales / ~JPY 75B OP / ~22.7% margin) -- an upward revision to the mid-term plan. FY-Mar-2030 midterm target: net sales ~JPY 1.0T and operating profit ~JPY 300B (~30% OP margin) -- roughly 2.4x sales and ~5x OP vs FY-Mar-2026; electronics-led. This target is the pillar of the equity story.
- Electronics 9M revenue
- JPY 171.9B (+18.2% YoY)
- Electronics 9M OP
- JPY 33.0B (+65.9% YoY)
- Ceramics 9M revenue
- JPY 60.6B (-2.4% YoY)
- Customer advance payments on B/S
- ~JPY 92.1B
- Committed electronics capex (3yr)
- ~JPY 500B
Growth drivers
- AI accelerator content growth — Nvidia Blackwell needs >2x the substrate area of Hopper; Rubin adds a further ~75% over Blackwell; more layers + larger bodies per GPU/ASIC
- Tight ABF/FC-BGA supply returning to imbalance from H1 2026 -> pricing power for the top suppliers
- 2.5x AI-substrate capacity ramp (Ono + Gama plants) under customer-financed contracts
- Data-center CPU/GPU/custom-ASIC (hyperscaler silicon) unit and complexity growth
- Mix shift toward the high-margin Electronics segment
- Weak yen boosting export competitiveness vs Taiwan/Korea peers
Bull & bear
Ibiden is the picks-and-shovels chokepoint of the AI hardware build-out: every AI accelerator sits on an FC-BGA substrate, substrate area per chip is exploding, and only a handful of firms can make the biggest, highest-layer-count parts at yield. Customer-financed capacity plus a net-cash balance sheet lets Ibiden ramp 2.5x into a tight market with the FY-Mar-2030 target of ~JPY 300B OP (~5x current) as the prize.
- Substrate area per GPU is compounding (Blackwell >2x Hopper; Rubin ~+1.75x Blackwell) -- content growth on top of unit growth
- Consolidated market (top-5 ~74%) with returning supply-demand imbalance from H1 2026 = pricing power for a leader
- JPY 500B capex is largely pre-sold: contracts + ~JPY 92.1B advance payments de-risk utilization and protect ROIC
- Electronics OP grew ~66% YoY (9M) with ~19% margins -- operating leverage is already showing, and FY-Mar-2027 OP guidance of ~JPY 90B (+45%) extends it
- Net-cash balance sheet and weak yen mean the ramp is fundable and export-competitive vs Taiwan/Korea peers
- FY-Mar-2030 midterm target (JPY 1T sales / JPY 300B OP) frames a credible ~5x-OP compounding path if AI demand holds
The stock discounts near-flawless execution of an AI supercycle. At ~JPY 5.97T market cap on ~JPY 62B operating profit and near-zero free cash flow, Ibiden trades at a rich multiple where any of AI-capex digestion, Intel share erosion, a glass-substrate transition, or a ramp overbuild compresses both earnings and the multiple hard.
- Rich valuation (~94-100x trailing earnings, and net income is inflated by one-off gains) leaves no room for disappointment
- AI accelerator demand is cyclical -- a digestion year could strand the 2.5x, JPY 500B capacity build and crush margins on underutilization
- Legacy reliance on Intel, whose data-center and foundry roadmap is contested, is a concentration risk mid-transition to Nvidia/AMD/ASICs
- Glass-core and panel-level substrates (pushed by Intel, TSMC, Samsung) threaten to leapfrog the ABF franchise later this decade
- Free cash flow is ~zero during the build; a demand air-pocket turns heavy fixed costs into a margin problem fast
- Concentrated single-region (Gifu) manufacturing and single key input (Ajinomoto ABF) are tail risks; ceramics segment is a persistent drag
What it is worth
AI-supercycle growth premium anchored to the FY-Mar-2030 midterm target (JPY 1.0T sales / JPY 300B OP). Cross-checked on trailing multiples and a reverse-DCF of that target. At ~JPY 5.97T (~$37B) on FY-Mar-2026 company-reported OP JPY 62.0B and net income JPY 63.7B, the stock trades at a very high trailing multiple (~94-100x reported earnings, though net income is inflated by one-off gains; ~14x sales), i.e. it prices in successful execution of the ~5x-OP ramp.
An AI-capex digestion year, Intel share erosion, or an early glass-substrate transition leaves the JPY 500B build under-utilized; OP disappoints and the ~94-100x-earnings starting multiple compresses hard -> meaningful drawdown. Near-zero FCF offers no valuation floor during the build.
Electronics grows strongly but ramp timing, ceramics drag, and periodic capex digestion mean OP scales toward roughly JPY 150-220B by FY-Mar-2030 (below the JPY 300B stretch target); the stock compounds roughly with earnings as the multiple gently de-rates from today's premium.
AI demand holds, ABF stays tight, and the 2.5x capacity fills at strong price/mix -> OP marches from ~JPY 90B (FY-Mar-2027 guide) toward the ~JPY 300B FY-Mar-2030 target (~5x FY26); the multiple is sustained or re-rates on visible compounding. Upside from here even at a rich starting point.
The equity is priced as a call option on the AI-substrate supercycle. Valuation is 'supported' only if the FY-Mar-2030 target is broadly achieved; on trailing fundamentals alone it is expensive. Near-term FY-Mar-2027 guidance (~JPY 90B OP) provides an interim bridge but still leaves a large gap to the JPY 300B stretch target. Earnings-quality caveat: FY-Mar-2026 net income was lifted by investment/one-off gains, so headline P/E understates operating richness.
SWOT
Strengths
- World-leading technology in high-layer-count, large-body FC-BGA/ABF package substrates -- the hardest substrates to yield
- Entrenched in a consolidated oligopoly (top-5 ABF suppliers ~74% of the market) supplying Intel, Nvidia and AMD
- Customer co-investment / advance-payment model de-risks a JPY 500B capex ramp
- Net-cash balance sheet funds the build-out without stress
- Improving mix and pricing lifting Electronics margins to ~19%
Weaknesses
- Historic customer concentration in Intel, whose data-center/foundry roadmap has stumbled
- Near-zero free cash flow during the multi-year capacity build
- Reported net income flattered by non-operating investment gains -- run-rate quality lower than headline
- Ceramics/auto (DPF) segment in structural decline as diesel/ICE fades
- Japan cost base and single-region manufacturing concentration (Gifu Prefecture)
Opportunities
- Secular AI-accelerator substrate content growth (Blackwell -> Rubin and beyond)
- Hyperscaler custom-ASIC (Google TPU, Amazon Trainium, Microsoft Maia, Meta) proliferation broadens the customer base beyond Intel
- ABF supply tightness through the 2026-2028 upcycle supports price/mix
- Adjacency into next-gen packaging (glass-core substrates, panel-level) if it leads rather than gets disrupted
- GPF (gasoline particulate filter) and SiC give the ceramics unit an EV/hybrid-era pivot
Threats
- AI capex is cyclical -- a digestion phase could leave the 2.5x capacity ramp underutilized
- Share loss to Taiwan/Korea rivals (Unimicron, Nan Ya PCB, Samsung Electro-Mechanics, LG Innotek) or to Shinko under new ownership
- Glass-core / panel-level substrate transition (Intel, TSMC, Samsung pushing glass) could disrupt the ABF franchise
- Dependence on Ajinomoto ABF film as a concentrated single key input
- FX swings and any yen strengthening erode the export cost advantage
Moats, dependencies & bottlenecks
Moats
high-layer-count FC-BGA substrates (yield, warpage, thermal control) High (multi-year to replicate at yield) The specific hard-to-make parts AI accelerators need are a narrow supplier set
Design-in and requalification cycles create switching friction; top-5 hold ~74% share
Moderate-Strong ~JPY 92.1B advances lock demand to Ibiden's new lines and de-risk ROIC
Balance-sheet strength (net cash) enabling counter-cyclical capacity Lets Ibiden out-invest weaker peers through the cycle
Real IP but in a structurally shrinking diesel/ICE end market
Dependencies
The entire growth case rides on sustained AI accelerator build-out
Customer concentration Historically the largest substrate customer; its roadmap/foundry issues cut both ways -- risk plus diversification incentive
Key input supplier ABF is the near-monopoly dielectric for FC-BGA; concentrated single-source input
Substrate demand tracks chip unit and complexity cycles
Japan cost base; yen weakness is a current tailwind that can reverse (FY-Mar-2027 guidance assumes ~JPY 150/USD)
Advantages
- Best-in-class capability at the largest, highest-layer-count substrates that AI GPUs/ASICs require
- Diversifying blue-chip customer base (Intel legacy + Nvidia/AMD/hyperscaler ASIC growth)
- Customer-financed, pre-sold capacity ramp protects utilization and returns
- Net-cash balance sheet funds counter-cyclical investment
- Weak-yen export cost advantage vs Taiwan/Korea competitors
Weaknesses
- Very rich valuation with earnings quality flattered by one-off investment gains
- Near-zero free cash flow during the capex supercycle
- Legacy Intel customer concentration amid Intel's own transition
- Structurally declining ceramics/auto (DPF) segment
- Single-region (Gifu) manufacturing concentration and single-source ABF input
- Exposure to a possible glass-substrate / panel-level packaging disruption later this decade
Bottlenecks
- ABF/FC-BGA substrate supply is itself the industry bottleneck for AI accelerators -- Ibiden's ramp pace gates customers
- Multi-year plant lead times (Ono/Gama) mean capacity cannot respond quickly to demand swings
- Ajinomoto ABF film supply as a concentrated upstream input
- Yield/warpage engineering on ever-larger body sizes and higher layer counts
- Skilled cleanroom labor and equipment availability during a simultaneous industry-wide expansion
Top signals & trends
Top signals
Management conviction + customer-backed demand; also raises overbuild risk if AI capex cools
Demand is contracted, not speculative -- de-risks the ramp
Operating leverage from AI mix already visible
Management raised targets on AI-substrate demand strength
Beat is partly non-operating -- discount for run-rate quality
Legacy segment is a persistent drag on the blended P&L
Tight supply supports price/mix for leaders
Improves liquidity/retail access; not fundamental
Trends
Blackwell >2x Hopper area; Rubin ~+75% over Blackwell -- structural per-unit tailwind
Supports pricing discipline and returns for scale leaders
Ceramics segment structurally shrinking; GPF/SiC only a partial offset
Mixed/Long-term risk · Could disrupt ABF next decade; opportunity if Ibiden leads the transition
Export cost advantage vs Taiwan/Korea peers; reverses if yen strengthens
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.
Near-monopoly supplier of ABF build-up film, the defining FC-BGA dielectric input
BT resin and electronic materials used in substrate build-up layers
Copper foil / laminate and specialty-chemical suppliers Copper, prepreg, plating and photoresist inputs for substrate fabrication
Semiconductor equipment / laser drilling and plating tool vendors Capacity-critical capital equipment for the Ono/Gama ramp
Historically the largest substrate customer (CPUs/data-center); roadmap in transition
AI GPU substrates (Hopper/Blackwell/Rubin) -- fastest-growing demand pool, often via OSATs
EPYC CPUs and Instinct AI accelerators drive high-end substrate demand
Custom-ASIC / networking silicon for hyperscalers needing large substrates
Ceramics segment DPF/GPF filter customers
Taiwan's largest ABF/FC-BGA substrate maker (~22% share, market leader); the primary global rival for AI-server substrates
Formosa-group Taiwanese substrate/PCB maker, a top-5 ABF player expanding in high-end FC-BGA
Japanese substrate peer, formerly TSE:6967; taken private by a JIC-led consortium (2024-2025). Close technical rival to Ibiden
Korean substrate/component maker scaling FC-BGA for servers/AI
Austrian high-end substrate/PCB maker, a top-5 ABF player, Intel-aligned, building AI-substrate capacity
Korean entrant investing in FC-BGA substrates
Taiwanese substrate maker (ASE-affiliated), mid-tier FC-BGA
Main rival in the ceramics/DPF-GPF segment, not substrates