2308 — Delta Electronics, Inc.

Thesis

Delta Electronics is the dominant supplier of power delivery systems for AI data centers, and the business case is genuinely excellent while the stock at current prices is priced for perfection. Delta holds an estimated ~60% share of the AI server PSU market, is the only company globally offering integrated power delivery from 20,000V grid to 0.8V chip, and co-developed the 800V HVDC rack power shelf with NVIDIA that defines the next-generation data-center architecture (>98% conversion efficiency, 1.1 MW per rack, unveiled at NVIDIA GTC 2025). The Infrastructure segment grew +82% revenue and +413% EBIT in FY2025. The stock has re-rated violently — up ~530% over 12 months — and now trades at 89.9x TTM P/E / 54.6x forward P/E. The thesis is real; the valuation is extreme.

Verdict / stance: WATCH — do not buy at TWD 2,075. Stage entries at TWD 1,700 and below. This is a growth-compounder buy thesis, not a value or yield play. Conviction is Medium: the structural thesis (AI power density driving Delta's Infrastructure segment) is HIGH conviction; the valuation at current prices, which requires flawless execution and sustained AI capex into 2028 with no slowdown, is a LOW conviction assumption. Net: no new money above TWD 1,800; scale-in thesis only below TWD 1,200 (implied ~40% discount from current). Expected holding period 18–36 months — a cycle play on the AI infrastructure buildout, not a forever-hold at current multiples.

What has to be true: the AI-server Infrastructure segment must maintain +30%+ revenue growth in FY2026, which requires sustained hyperscaler capex and no disruption to Delta's NVIDIA co-development relationship or market share; gross margin must hold near 34–35% as AI mix grows; and the NVIDIA Vera Rubin upgrade cycle (flagged for Q3 2026) must arrive on schedule. The single most important thing that must go right is sustained hyperscaler capex feeding +30%+ IFB growth.

Single most important risk: AI revenue concentration is simultaneously the thesis and the risk — if hyperscaler capex pauses or NVIDIA shifts away from HVDC, the Infrastructure segment decelerates sharply and the multiple compresses at the same time (a double hit). All five FundamentEdge gates pass; management grade is Green/B+; the only failing gate is valuation.

Snapshot

Delta Electronics, Inc. — 2308 / Taiwan Stock Exchange (TWSE). Power electronics and thermal management; dominant AI server power-delivery player. GICS: Information Technology / Electronic Equipment, Instruments & Components. Founded April 4, 1971 (Taipei, to produce TV deflection coils); ~83,000 employees. HQ: 186 Ruey Kuang Road, Neihu District, Taipei (business); 131 Xingbang Road, Guishan District, Taoyuan (operations). Website: deltaww.com. ~200 facilities globally.

Price / valuation snapshot (as of April 26, 2026):

• Current share price: TWD 2,075
• Market cap: TWD 5,390B (~US$168B)
• Enterprise value: ~TWD 5,313B (net cash ~TWD 77B)
• P/E (TTM): 89.9x | Forward P/E: 54.6x
• EV/EBITDA: ~45x (est., if EBITDA ~TWD 118B; not available from public screeners)
• P/FCF: ~102–103x (TWD 5,390B mcap / TWD 52.4B FCF)
• FCF yield: ~1.0% | Dividend yield: 0.56% (TWD 11.60/share annualized)
• 52-week range: TWD 324.50 – TWD 2,115.00 (the buy-checklist quotes the low as TWD 332 / range "332 – 2,115"; discrepancy flagged — both figures appear in source fragments)
• All-time high: TWD 2,115 (current price within ~2% of ATH)

Stock up ~530% (variously cited 529–540%) over the trailing 12 months; valuation reflects a significant AI growth premium. Became the third-largest company in Taiwan by market cap during the 2025 surge. USD conversions throughout at ~TWD 32/USD. Financial data primarily from StockAnalysis.com (TWSE filings); FY = calendar year ending December 31.

Credit: S&P Global 'BBB+' / Stable; Taiwan Ratings 'twAA/twA-1+' (affirmed 2025/2026).

Business

What Delta does: It converts electricity into the precise voltages that keep industrial machines, automobiles, buildings, and — increasingly — AI servers running. No power delivery, no inference. It is the only major Taiwan supplier offering fully integrated power and thermal management under one roof, spanning 20,000V grid access down to 0.8V at the chip.

Business segments (FY2025 revenue)

Manufacturing-heavy OEM/ODM model. Revenue is largely one-time capital-equipment purchases by data-center operators, though service/maintenance contracts are growing.

Segment	Revenue (TWD B)	% of Total	YoY Growth	What it does
Power Electronics (PEB)	280.0	50.5%	+25% (EBIT +37%)	PSUs, AC-DC converters, fans, thermal modules for servers, telecom, industrial, consumer
Infrastructure (IFB)	182.0	32.8%	+82% (EBIT +413%)	AI data-center power systems (HVDC racks, UPS, liquid cooling/CDUs), energy storage, building/DC management
Automation (AUB)	54.9	9.9%	+5% (EBIT −39%)	Industrial automation, CNC drives, motion control, building automation
Mobility / EV (EVB)	37.0	6.7%	−16%	EV chargers, on-board chargers, battery management — EV-market headwinds, negative EBIT
Other	0.9	0.2%	flat

Infrastructure is the breakout segment — the +413% EBIT growth in FY2025 reflects high-margin system-level contracts. PEB AI-PSU mix is expanding; Delta stated >20% of total company revenue came from AI server PSUs in FY2025 (implying ~TWD 110B+, sitting across PEB and IFB combined).

Segment deep-dive economics

• PEB (50%): Switched-mode power supply at industrial scale; vertically integrated magnetics, PCB assembly, in-house test. Standard server PSU ASP US$50–300; high-density AI server PSUs (3kW–10kW) US$500–2,000+. The AI-specific PSUs carry the margin.
• IFB (33%): Not components but full rack-level power systems — the HVDC rack power shelf (1U–2U, 90–180kW at >98% efficiency, hot-swappable) bolts into NVIDIA-certified racks. Liquid cooling line: liquid-to-air CDUs, direct-liquid cooling (DLC) to cold plates, liquid-to-liquid CDUs. HVDC rack-shelf systems run US$50,000–500,000+ per rack; liquid-cooling CDUs US$10,000–100,000 per rack. Liquid cooling was ~9% of FY2025 total revenue (~TWD 50B). Once designed into a hyperscaler rack spec, Delta is locked in for that rack generation (2–4 years).
• AUB (10%): Industrial drives, CNC motion controllers, building automation, EV test systems. Mature; +5% revenue but EBIT −39% in FY2025 on pricing pressure. Optionality if industrial capex recovers.
• EVB (7%, declining): EV on-board chargers, EVSE fast chargers, BMS. −16% in FY2025; a ~TWD 37B revenue drag at negative EBIT. EV OEM inventory correction is structural, not purely cyclical. Not the thesis driver but a potential positive surprise if EV recovers to 2022 growth (could add 5–8% to total-company EBIT).

Customers & partners

#	Customer	Ticker	Est. Revenue Share	Relationship
1	Hyperscalers (Microsoft, Google, Meta, Amazon)	MSFT/GOOG/META/AMZN	~30%+ combined (undisclosed individually)	AI server OEM/ODM; HVDC rack qualified for GB200
2	NVIDIA	NVDA	Indirect (via OEM)	Co-development / ecosystem; co-developed 800V HVDC shelf; design authority for rack spec
3	Apple	AAPL	Undisclosed	PSU / power-component OEM (MacBook/desktop adapters); multi-decade
4	Tesla	TSLA	Undisclosed	EV on-board charger / EVSE; FY2025 headwinds
5	Server OEMs (Quanta, Foxconn/Hon Hai, Inventec, Wistron, Supermicro)	SMCI / private	Undisclosed	PSU ODM

Hyperscaler customer quality is high: Microsoft guided ~$80B FY2025 capex, Google ~$75B, Meta ~$65B, Amazon ~$105B (~$325B combined) — non-discretionary spend. No single customer is formally disclosed >10–20% of revenue, but top-3 hyperscalers likely collectively ~25–35%. Switching cost is high: hyperscalers qualify AI-rack power systems over a 6–18 month design-in and certification cycle and do not swap mid-generation; the NVIDIA co-development relationship is the strongest protection against substitution.

Moat & competitive position

1. Scale and manufacturing depth — ~200 facilities, 83,000 employees; no PSU competitor matches the vertical integration.
2. Technology lead — >98% conversion efficiency (Titanium-certified) from 50 years of proprietary magnetics, control algorithms, and process engineering, amplified by WBG (GaN via Texas Instruments, SiC via Infineon) partnerships giving preferred device allocation; competitors cannot close this in 3–5 years.
3. Full-stack power delivery — the only company offering an integrated solution from 20kV grid to 0.8V chip. Vertiv does UPS and cooling but not the server-PSU layer; Advanced Energy does precision power for semicon; Lite-On does PSUs but not liquid cooling or HVDC distribution.
4. NVIDIA co-development IP — being embedded in NVIDIA's reference HVDC architecture is possibly the most durable advantage; a competitor would need NVIDIA to certify a new power shelf mid-generation.
5. Geographic manufacturing resilience — Thai, Indian, and US (Plano) capacity built over 3+ years gives a tariff-resilient advantage with US hyperscalers.

Competitors: Lite-On Technology (2301.TW, direct PSU/server power), Vertiv Holdings (VRT, data-center power/cooling, ~80% DC exposure, more software/service-oriented), Advanced Energy Industries (AEIS, precision/semicon power), FSP Group (3015.TW, PSU mfg), Chicony Power (6412.TW, consumer/laptop chargers), Eaton (ETN, DC power/UPS/PDU). For all server PSUs (incl. non-AI), Delta + Lite-On hold ~68% combined global share; for AI-specific HVDC systems Delta alone holds ~60% (near-monopoly). Emerging-threat watch: Chinese PSU makers (Shenzhen Honor Electronic, Great Wall Technology, Xin Lian) — strong on cost, weak on AI efficiency and US hyperscaler access (currently tariff-blocked); hyperscaler in-house power design (Meta/Google OCP specs); Vertiv encroachment downmarket.

Geographic & operations footprint

~70% international, ~30% Taiwan/China domestic (inferred from customer base; not formally disclosed). Americas (US hyperscalers) is the fastest-growing region. Manufacturing: Taiwan (Taoyuan, Chungli, Pingjhen, Tainan, Taichung — 10+ plants), China (Dongguan Plants 1–8 + Wujiang Plants 1–5 + Chenzhou = 13–15+ plants, the largest single base), Thailand (Bangpoo Automotive, Bangpoo Power Supply, Wellgrow; 3 new factories online end-2025; targeted as second manufacturing HQ), India (Gurgaon, Bangalore, Rudrapur, Krishnagiri, Chennai), US (Plano, Texas: 435,000 sqft, expanding to ~1.5 msf by 2031), plus Japan/Korea/Singapore/Vietnam/Europe. Delta cut China headcount by >50% while expanding Thailand — a proactive supply-chain diversification ahead of tariffs.

Why it matters / TAM

AI servers consume 5–10x more power per rack than conventional servers: an NVIDIA GB200 NVL72 rack draws ~120 kW vs ~10–20 kW for a standard rack; Vera Rubin ~200kW+. Electricity is the binding constraint for AI. TAM: AI server PSU market ~$2B (2024) → $8–12B by 2032 at ~25%+ CAGR (one set of estimates); the deep-dive also cites ~$5B in 2025 → ~$15B by 2033 at ~15% CAGR (more conservative) — estimates differ across fragments, both retained. Data-center power and cooling combined: $50B+ by 2030, with a high-end $173B (30% CAGR) scenario in the Vertiv×Delta comparison. NVIDIA aligned on 800V HVDC (Delta's architecture) rather than ±400V bipolar (OCP/Meta/Google preferred), bifurcating the market toward Delta for NVIDIA-affiliated deployments.

Technology first-principles

The 800V HVDC power chain eliminates multiple AC-DC conversion stages, reduces cable weight ~9× (~450 lbs Cu → ~50 lbs Cu for 1MW), and enables >98% conversion efficiency vs ~87% for legacy AC. Power chain: Utility AC (11–20kV) → Solid-State Transformer → 800VDC → 800V HVDC busway → AC-DC server power shelf (180kW/2U or 72kW/1U, >98%) → in-row HVDC/DC shelf (800VDC→50VDC, 90kW/1U, >98%) → GPU power rail (~50V/12V) → chip-level VRMs (0.8–1.8V). Total system efficiency 92%+ vs ~87% conventional. WBG semiconductors (GaN 3.4 eV, SiC 3.2 eV bandgap) switch in the MHz range, run hotter, handle higher voltages — enabling smaller magnetics, higher efficiency/density. Value-chain bottleneck: GaN power-device supply (tightest as TSMC exits GaN), but TI and Infineon co-development agreements give Delta preferred allocation. Key suppliers: Infineon (IFX.DE, SiC/GaN), Texas Instruments (TXN, GaN ICs/controllers), TSMC (2330.TW, GaN-on-SiC foundry, exiting), in-house captive magnetics, commodity copper.

Financials

All figures TWD millions unless noted; USD at ~32 TWD/USD. FY = calendar year ending December 31.

Income statement & margins

Metric	FY2022	FY2023	FY2024	FY2025	FY2026E
Revenue (TWD M)	384,443	401,227	421,148	554,885	~720,000+ (est.)
Revenue growth YoY	+22.2%	+4.4%	+5.0%	+31.8%	~30%+ (mgmt guide + Q1 +34%)
Gross profit (TWD M)	110,773	117,213	136,580	190,157	~250,000E
Gross margin %	28.8%	29.2%	32.4%	34.3%	~34–35%E
EBIT (TWD M)	41,439	40,942	47,734	84,042	~115,000E
EBIT margin %	10.8%	10.2%	11.3%	15.2%	~16%E
Net income (TWD M)	32,666	33,393	35,229	60,108	~80,000–85,000E
Net margin %	8.5%	8.3%	8.4%	10.8%	~11–12%E
EPS (basic, TWD)	12.58	12.86	13.56	23.14	~31–33E
EBITDA (TWD M)	N/A	N/A	N/A	117,900	~155,000E
EBITDA margin %	N/A	N/A	N/A	21.3%	~21–22%E

FY2026E is author estimate based on Q1 2026 +34% YoY run rate, management double-digit guidance, and continuation of margin structure — not verified consensus.

Margin expansion is real: gross margin expanded 5.5pp (FY2023 29.2% → FY2025 34.3%); EBIT margin expanded 5pp (10.2% → 15.2%). Operating leverage is genuine and accelerating as IFB mixes up. R&D at ~9% of sales (~TWD 50B FY2025) is the main gross-to-EBIT drag — and the moat investment.

Incremental margin (FY2025 vs FY2024)

• Gross profit +TWD 53.6B on +TWD 133.8B revenue = 40.1% incremental gross margin (vs reported 34.3%) — new revenue higher quality than base.
• EBIT +TWD 36.3B on +TWD 133.8B revenue = 27.1% incremental EBIT margin (vs reported 15.2%) — very strong operating leverage from IFB mix.

Quarterly revenue & second-derivative

Quarter	Revenue (TWD B)	YoY %	QoQ %	Gross margin	Op. margin	EPS (TWD)
Q1 2025	118.9	+30%	+4.5%	31.8%	~10.5%	~3.87*
Q2 2025	124.0	+20%	+4.3%	35.5%	15.1%	5.37
Q3 2025	150.3	+34%	+21%	34.9%	16.5%	7.16
Q4 2025	161.6	+42%	+7.5%	34.6%	16.3%	6.67
Q1 2026	159.4	+34%	−1.4%	—	—	—

*Q1 2025 net income/EPS estimated from FY total minus Q2–Q4 confirmed figures. YoY growth has held in the +30–42% band since Q1 2025 — a sustained step-change from 4–5% in FY2023–2024. Q2 2025 GM of 35.5% was a high-water mark; Q3/Q4 normalization to ~34.6–34.9% is not concerning. Q4 2025 EPS TWD 6.67 vs TWD 2.76 a year prior and vs TWD 7.16 prior quarter.

Cash flow & balance sheet

Metric	FY2022	FY2023	FY2024	FY2025
Operating CF (TWD M)	46,529	71,086	72,895	98,474
Capex (TWD M)	−21,824	−27,830	−33,485	−46,091
Free cash flow (TWD M)	24,705	43,257	39,410	52,383
FCF margin %	6.4%	10.8%	9.4%	9.4%
Net cash (TWD M)	13,293	31,078	54,006	77,091
Net debt / EBITDA	Net cash	Net cash	Net cash	Net cash
ROIC	~9%	~9%	~10%	~15%

Balance sheet is pristine: net cash TWD 77B (~US$2.4B) and growing; no leverage risk. ROIC: GuruFocus TTM 15.02% vs estimated WACC ~9–11% for a Taiwan large-cap exporter — value-creating and expanding. (Note: the prior canonical snapshot table listed ROIC as N/A; that was a screener gap later resolved to ~15% via GuruFocus — reconciled here to 15%.) Capex escalating to fund Thailand (3 factories online end-2025) and US Plano build-out: TWD 46B FY2025, guided ~TWD 40B FY2026 across four countries with greater automation focus (FY2025 actual ran ~15% above the ~TWD 40B guide). Capex efficiency (revenue per TWD of capex): 3.2× FY2022, 0.6× FY2023, 0.6× FY2024, 2.9× FY2025 — the lean-year FY2023–24 spend set up the FY2025 surge.

No dilution risk: share count stable at ~2.60B shares; FCF-generative, net-cash, no convertible/warrant overhang, no material SBC. FCF of TWD 52.4B comfortably covers capex (~TWD 40–46B) and dividends (~TWD 30B at TWD 11.60/share). Asset turns ~0.87× (TWD 555B revenue / ~TWD 640B assets) — normal for integrated power-electronics manufacturing.

Industry landscape

AI infrastructure: power delivery and thermal management are the binding constraint on AI compute scale-out. Industry structure is oligopolistic at the quality end — top-2 (Delta + Lite-On) ~68% of global server-PSU share; Delta alone ~60% of AI-specific HVDC. Barriers to entry are high: 6–18 month hyperscaler qualification cycles, patent-protected/trade-secret power-conversion efficiency IP, scale requirements, NVIDIA co-development gatekeeping, and GaN/SiC preferred-allocation agreements. The demand cycle is both secular (multi-year AI buildout; rack power density rising for 5–7 years) and cyclically overlaid (hyperscaler capex is not immune to rate cycles or AI-demand disappointment — the 2022–2023 cloud-capex deceleration is the cautionary precedent). Current position: early-to-mid cycle, GB200 ramp 2025–2026, Vera Rubin 2026–2027.

See sector page: ai-infrastructure

Management

Overall management grade: Green / B+ — trustworthy operators with real skin in the game; not financial engineers. No red flags. Founder-heritage Taiwan governance, not US ESG-premium standard.

Executive team

Name	Title	Tenure	Background
Ping Cheng (鄭平)	Chairman & CEO	CEO since Jan 2012 (14 yrs); Chairman since 2024	Son of founder Bruce Cheng; joined Delta 1988 as section chief; ran China/Thailand mfg expansion 1996–2008; SVP Power Supply Business 2008; first Chief Brand Officer 2010; architect of the "solutions-provider" transformation. BBA, California State University, East Bay. Awards: K.T. Li Award (2024), Order of Brilliant Star with Grand Cordon (2016), EY Entrepreneur of the Year (2010)
Bruce C.H. Cheng (鄭崇華)	Founder / Honorary Chairman	Founded 1971; stepped back from executive role 2012	Largest individual shareholder; founded Delta Foundation 1990; "Taiwan's Pioneer in Environmental Sustainability"
Simon Chang (張訓海)	President & COO; Director	Joined 1981 (45 yrs); COO ~20 yrs	Operational anchor; led smart-manufacturing initiative. MBA, Chung Yuan Christian University
Mark Ko / Tzu-shing Ko (柯子興)	Vice Chairman	Joined 1988 (38 yrs)	Ex-RCA Taiwan / Zenith Taiwan; Delta President & COO 2004 before Chang; strategic advisory
Victor Cheng	Director; CEO, Delta Electronics (Thailand) PCL (DELTA.BK)	—	Leads separately-listed Thai subsidiary; sits on parent board (disclosed)
Shan-Shan Guo (郭珊珊)	Director; Chief Brand Officer; VC, Delta Foundation	3.8 yrs on board	Brand/ESG; minimal ownership
Johnson Lee	EVP, Infrastructure (IFB)	—	Leads the +82% segment
Ted Shyy	EVP, Power Electronics (PEB)	—	Leads the core 50% segment
James Tang	EVP, Mobility (EVB)	—	EV/transport
Jimmy Yiin	EVP, Global Business Operations	—	Commercial/go-to-market
Rock Huang	EVP, Global Manufacturing	—	Manufacturing/supply chain

The Bruce → Ping Cheng succession (2012) and the Yancey Hai → Ping Cheng chairman transition (2024) were both orderly and pre-flagged; no abrupt departures. Average management tenure ~9.3 years.

Insider ownership (critical alignment signal)

Total identified insider ownership ~8.2% (Cheng family ~8.1% + professional management ~0.07%). Three Cheng family members sit on the 11-person board (27% family representation), a family-controlled company by influence with no dual-class shares.

Name	Role	Stake	Shares	Est. Value
Bruce C.H. Cheng	Founder / Hon. Chairman	4.08%	~107M	~NT$219.8B (~US$6.9B)
Ping Cheng	Chairman & CEO	2.14%	~56M	~NT$115.5B (~US$3.6B)
An Cheng	Director (family)	1.98%	~52M	~NT$106.6B (~US$3.3B)
Simon Chang	President & COO	0.035%	~914K	~NT$1.9B (~US$60M)
Tzu-shing Ko	Vice Chairman	0.031%	~811K	~NT$1.7B (~US$53M)
Shan-Shan Guo	CBO / Director	0.00078%	~20K	~NT$42M (~US$1.3M)

The CEO's entire personal net worth (~US$3.6B) is in the company he runs; the founder is the largest individual holder at ~US$6.9B. Every 1% stock move = ~US$36M to Ping Cheng personally. This is the strongest possible alignment signal; no diversification away from Delta is evident, and no public insider selling was flagged during the 530% run (precise TWSE/MOPS insider-transaction data not accessible via US channels).

Board & governance

11 members; 5 independent (45%) — exceeds Taiwan's 1/3 minimum, below Western majority-independent best practice. Independents: Shyue-Ching Lu (4.8 yrs), Kai-Lien Tsou (4.8 yrs), J. T. Huang (4.8 yrs), Ji-Ren Lee (9.8 yrs, longest-serving), Audrey Tseng (3.8 yrs) — backgrounds not available in English. No dual-class shares, no known poison pill, no staggered board; the 8.1% family stake is the de facto takeover protection. ISS Governance QualityScore: 5/10 (Audit 3 | Board 4 | Shareholder Rights 5 | Compensation 8) — low Audit/Board scores partly reflect Taiwan's less-transparent disclosure framework, not necessarily malpractice.

Capital allocation

Grade B+. Conservative acquirer: Universal Instruments Corp (Dec 2021, ~$89M, automation bolt-on) and a small Japanese RF power company (Q3 2025, semiconductor-power capability) — two small deals in 10 years, no value-destructive mega-deals. No buyback history — the single most consistent criticism: at 8–20× P/E in 2015–2019 and 2022–2023, buybacks would have been accretive; instead all FCF went to dividends and capex (TECC timing verdict: Neutral — missed opportunity, not value destruction; reinvestment in Thailand/India/US capacity proved the correct call in hindsight). Dividend grew NT$3.50 (FY2020) → 6.00 → 6.00 → 6.43 → 7.00 → 11.60 (FY2025), payout ratio ~50%. No dilutive equity raises; share count stable ~2.60B. R&D held at ~9% of sales through cycles (not cut to juice earnings) — the highest single alignment indicator.

Credibility

~86% follow-through (6/7 tracked statements). Conservative / straight-shooter guidance — uses directional language ("double-digit," "stronger H2," "if things go smoothly"), under-promises and over-delivers. Examples: "double-digit revenue growth in 2026" → Q1 2026 +34%; "three new Thai factories by year-end" → confirmed; 2024 Chairman's letter "increase in sales volume" → actual +31.8%. The one blemish: Q3 2024 missed analyst consensus by 4.1% (revenue) and 16% (EPS) — primarily a consensus miss (analyst models too optimistic), not a management guidance break. "Full-year capex ~NT$40B" → actual NT$46.1B (ran ~15% above). ISS Compensation 8/10; no extraordinary SBC dilution.

Shell / cross-holdings scan

Clean. Straightforward parent-subsidiary structure: Delta Electronics, Inc. (2308.TW) → 100%-owned mfg subsidiaries (China/India/US) + separately-listed Delta Electronics (Thailand) PCL (DELTA.BK, Victor Cheng CEO) + Delta Networks + charitable Delta Foundation. No shell entities, offshore tax vehicles, IP-licensing-to-insider schemes, related-party extraction, or revenue circularity identified. Litigation: Delta was plaintiff (not defendant) in a patent claim against Appotronics (2022, lost/withdrew — routine IP litigation); one historical 2016 commercial dispute involving the Thai subsidiary; no SEC enforcement, sanctions, or personal bankruptcies.

Flags

• Green: family alignment (8%+ founder stake held 50+ years), CEO ~US$3.6B in stock, 14-yr CEO track record (revenue 4.4× under his tenure, ~NT$125B 2012 → NT$555B 2025), clean shell scan, ISS Compensation 8/10, no dilution, no litigation/enforcement red flags, conservative guidance.
• Yellow: board independence 45% (below Western standard); An Cheng (1.98%, ~US$3.3B) background opaque in English — almost certainly Cheng family, warrants Chinese-language TWSE filing review; family-dynasty succession optics; no buyback program; ISS Audit 3/10 and Board 4/10.
• Red: none identified.

Note: a separate pre-canonical scaffold (DELTA.BK, the SET-listed Thai subsidiary) carries its own management findings — including a B−/Yellow grade, a parent $525M exchangeable bond pledging DET shares (Jan 2025), and SET-disclosed related-party transactions. Those facts belong to DELTA.BK, NOT to 2308.TW, and are not merged here.

Catalysts & risks

Catalysts (bull)

Near-term (0–12 months):

• Q1 2026 full earnings (April 30, 2026): confirm +34% YoY revenue; watch IFB segment EBIT, gross margin, and Q2 guidance. (Buy-checklist flagged this as binary event risk 4 days out — do not initiate before the print.)
• GB200 NVL72 deployments ramping through H1 2026; power-shelf orders accelerating.
• Q2 2026: management guided "significant increases in AI power shipments."
• Q3 2026: generational product cycle aligned with NVIDIA Vera Rubin platform launch.
• Monthly TWSE revenue filings — watch for continued +30%+ YoY (TWD 50–60B+ monthly runs; March 2026 was a TWD 59.8B single-month record, +37.6% YoY).

Medium-term (1–3 years):

• Vera Rubin NVL144 rack ramp (~200kW/rack vs ~120kW GB200): revenue per rack nearly doubles.
• Liquid cooling penetration from ~9% → 15–20% of revenue; each +1pp = ~TWD 5–6B revenue.
• Solid-State Transformer (SST) commercial scale: converts medium-voltage AC directly to 800VDC, eliminating the conventional transformer + UPS — a step-change in ASP and margin.
• US Plano facility completion (phased to 2031); India data-center boom; Thailand additional capacity (2026–2027 as US customers require non-China sourcing).
• EV recovery optionality (EVB could add 5–8% to total-company EBIT if EV returns to 2022 growth).

Risks (bear)

Risk	Likelihood	Notes / mitigants
AI capex cycle slowdown / pause	Medium	IFB ~33% of revenue and ~50%+ of EBIT; a 30% IFB revenue decline cuts total EPS ~25–30% AND compresses the multiple — double hit. Mitigant: diversified Apple/Tesla/industrial base. Structural; managed via diversification. 2022–2023 cloud-capex pause is the precedent.
Valuation / earnings-miss risk	High	89.9× TTM / 54.6× fwd P/E; priced for perfection; consensus already behind the move. Q3 2024 precedent: 16% EPS miss → sharp drop. A 10% miss could produce a 20–30% decline. Closes only via earnings growth over time.
China manufacturing concentration	High	13–15+ plants in Dongguan/Wujiang; cross-strait + tariff exposure. Mitigant: Thailand (2nd HQ, 3 factories online), India, US Plano scaling; China headcount cut >50%. Partial — 3–5 year transition; cannot fully exit China near-term.
Hyperscaler customer concentration	Medium	AI PSU >20% revenue concentrated in 3–4 hyperscalers; none individually >20%. NVIDIA co-development is the substitution protection. Partial — inherent to market leadership.
GaN supply tightening (TSMC GaN exit)	Medium	Powerchip/GlobalFoundries alternatives unproven at scale. Mitigant: TI + Infineon preferred-allocation agreements + in-house magnetics. Partial — depends on GF/Powerchip ramp.
EV segment prolonged weakness	High (near-term)	EVB −16% FY2025, negative EBIT; only 6.7% of revenue, limited P&L impact. Cyclical — outside Delta's control.
NVIDIA architecture shift away from HVDC	Low	NVIDIA invested heavily in 800V HVDC standardization + Delta co-development. Low-probability re-rating event.
Hyperscaler in-house power design	Low–Medium	Meta/Google have OCP specs; could commoditize PSU. Mitigant: in-house design still needs Delta's manufacturing scale + efficiency lead.
Automation softness	—	AUB EBIT −39% FY2025 despite +5% revenue — margin compression.

Dilution risk: none (FCF-positive, net cash, no dilutive programs). Key-person risk: Moderate — Ping Cheng is the architect of the AI pivot; his departure would be a re-rating event; Simon Chang (45-yr COO) provides continuity; no public succession plan disclosed for Ping Cheng. Portfolio correlation: high with NVIDIA, Vertiv, and AI-infrastructure names broadly — not diversifying in an AI-heavy book.

What would make the thesis wrong

• IFB segment decelerates below +15–20% YoY for two consecutive quarters
• Hyperscaler capex guides down materially (any two of MSFT/Google/Meta/Amazon cut DC capex >15–20%)
• NVIDIA shifts to an alternative power architecture (at GB300 or Vera Rubin+) not requiring Delta's HVDC co-development
• A large competitor (Vertiv, a Chinese OEM) closes the AI-specific power-delivery technical gap at scale
• Gross margin compresses below 32% (AI-PSU pricing pressure or margin-dilutive mix)

Behavioral traps (active at current price)

FOMO (up 530%, buying near ATH), narrative seduction (AI-power dominance story may already be in the price), and recency bias (extrapolating +34–42% growth permanently; Vera Rubin is a one-time upgrade event, not a new steady state) are all flagged. Confirmation/authority/averaging-down traps not active.

Valuation / DCF

Stance: business quality exceptional, valuation stretched — entry price matters enormously. At TWD 2,075 with FY2025 EPS TWD 23.14: P/E 89.9× (TTM); forward P/E 54.6× (on consensus ~TWD 37–38 FY2026E EPS). The buy-checklist notes 89.9× is ~157% above the 10-year median (~35×) and well above the 5-yr median (~25×). P/FCF ~102×; FCF yield ~1.0%; dividend yield 0.56%.

What the market is pricing in: continued 30–40% revenue growth for 2–3 more years; gross margin sustained at 34%+; no AI capex deceleration; no geopolitical supply-chain disruption. At 54.6× forward P/E, justifying the multiple implies ~+65% EPS growth in FY2026 (achievable if IFB stays +80%+ and GM holds ~34–35%). To justify 56× forward with a terminal 25–30× P/E implies ~13–18% EPS CAGR over 5 years — achievable given market position but with essentially zero error tolerance on execution.

DCF / intrinsic value

Rough DCF (FY2025 FCF base TWD 52.4B; 30% growth 3 years then decelerating to 10%; WACC 10%): Year 1 ~TWD 68B, Year 2 ~TWD 88B, Year 3 ~TWD 115B, terminal ~TWD 1,200B → total DCF value ≈ TWD 1,400–1,600B, vs current EV ~TWD 5,313B → implied ~3–4× overvaluation on this scenario. Even a bull case (FCF margin to 15% on TWD 900B FY2027 revenue ≈ TWD 135B FCF, multiple compressing to 30× FCF) yields mcap ~TWD 4,050B — below current. Alpha Spread base-case intrinsic value: ~NT$1,049–1,539 (the summary cites NT$1,049 base ≈ 49–50% overvalued vs ~TWD 2,040; DCF component NT$1,539). Simply Wall St: ~26% above two-stage DCF intrinsic (Jan-2025 estimate, likely stale, directionally consistent with overvaluation).

Peer comparison

Company	Ticker	Fwd P/E	EV/EBITDA	P/FCF	FCF yield	Mkt cap
Delta Electronics	2308.TW	54.6×	~45× (est.)	~103×	~1.0%	~US$168B
Vertiv Holdings	VRT	47.5×	52.5×	54.5×	1.8%	~US$124B
Advanced Energy	AEIS	~20×	—	—	~4%	~US$6B

Delta is slightly cheaper than Vertiv on forward P/E and EV/EBITDA but materially more expensive on FCF (higher capex relative to EBIT); Vertiv has the better FCF yield. Delta is larger by revenue (~US$17.9B TTM vs Vertiv $10.8B). The market appears to price Delta as "the NVIDIA of power" — but NVIDIA carries ~70% gross margins vs Delta's ~34%; the premium is emotional as much as fundamental.

Scenarios / price targets (12–18 month)

• Base case: TWD 1,800 (deep-dive author target; analysts average ~TWD 1,380 but have consistently lagged the move).
• Bull case: TWD 2,500 (deep-dive) to TWD 2,500–2,800 (buy-checklist, if Vera Rubin drives another IFB step-change).
• Bear case: TWD 900 (deep-dive exec summary, AI-capex correction + multiple compression) — note discrepancy: other fragments cite TWD 500–600 (deep-dive §15 / 2308.md addendum, IFB +82%→−20% + 30× P/E on ~TWD 17–19 EPS = −71–75% from current) and TWD 600–800 (buy-checklist final). All bear figures retained; the cluster spans ~TWD 500–900.

Analyst sentiment

Coverage 19 analysts; consensus Buy — 18 Buy / 0 Hold / 1 Sell. Average 12-month PT TWD 1,283–1,381 (range TWD 550–1,770). Current price TWD 2,075 — stock trades ~50–60% ABOVE the average PT. Analysts have been consistently behind the move (a common pattern for AI-infrastructure names where revenue acceleration exceeds model assumptions; analysts were similarly wrong on NVIDIA 2022–2025). This is not an automatic bearish signal, but the gap between a ~US$10B analyst-consensus valuation anchor and a ~US$168B market cap means significant downside if the AI narrative stumbles. Consensus EPS is stale relative to the recent AI revenue trajectory; PT-implied upside/downside is not a reliable signal here.

Conclusion: the stock has run ahead of even the most optimistic fundamental scenario. New money at TWD 2,075 is a momentum trade, not a value trade. The business is excellent — the price is the risk.

Decision log

2026-04-26 — Pre-buy checklist verdict: WATCH. Do not buy at TWD 2,075; build a staged buy-list entry at TWD 1,700 and below. All five FundamentEdge hard-rule gates PASS — (1) revenue-growth primacy ✅, (2) second derivative accelerating ✅, (3) valuation not the thesis ✅, (4) quality defensible ✅, (5) estimate revisions upward ✅. Pre-buy scorecard: thesis clear (Yes), business understood (Yes), gates pass (Yes 5/5), mgmt incentives aligned (Yes — 8.1% family, CEO US$3.6B in stock, ISS Comp 8/10), financials healthy (Yes — net cash, expanding margins, FCF+), valuation reasonable (No — 89.9× TTM / 54.6× fwd; Alpha Spread ~49% overvalued; near ATH), behavioral traps (Yes — FOMO, narrative seduction, recency bias), technicals support buying now (No — MACD negative −0.170, within 2% of ATH, RSI ~36.5 divergence, ~3.7× extended above 200-day MA ~TWD 559.85; 50-day MA ~TWD 923.66; earnings 4 days out), position sized (staged plan defined), exit plan (Yes).

2026-04-26 — Conviction set: Medium. Structural thesis HIGH conviction; valuation at current price LOW conviction. No new money above TWD 1,800 (deep-dive/checklist); scale-in thesis only below TWD 1,200. Full 3–5% position only below TWD 1,200.

2026-04-26 — Management DD: Green / B+. Trustworthy operators, real skin in the game, no red flags. ISS QualityScore 5/10 (Audit 3 / Board 4 / SR 5 / Comp 8). One yellow on board independence (45%) and An Cheng opacity; would trust these people with capital "with confidence."

Staged entry plan (as of 2026-04-26)

Tranche	Size	Entry level	Trigger
Starter	1–1.5%	TWD 1,700–1,800	Post-earnings consolidation or any 10%+ pullback
Core	1.5–2.0%	TWD 1,400–1,600	Sustained pullback to prior breakout zone
Full	1.0–1.5%	<TWD 1,200	Significant dislocation (macro / earnings miss)
Total target	3–5% of portfolio		Do not add above TWD 1,900 on a new position

Max loss tolerance 30% from entry on full position (at ~TWD 1,700 avg entry implies ~TWD 1,190 stop / bear-case valuation). Exit / re-evaluate triggers: IFB growth below +20–30% for two consecutive quarters; hyperscaler capex guide-downs (any two of MSFT/Google/Meta/Amazon −15–20%); NVIDIA shifts away from HVDC / Delta co-development; gross margin below 32%; Q1/Q2 2026 earnings miss >10–15%; price hits TWD 2,600 (DCF-implied max bull) without fundamental upside revision. What would cause an add: Vera Rubin launch confirmation with Delta named rack power supplier; IFB >100% YoY two quarters; pullback to TWD 1,200–1,500 on unrelated market selloff.

Near-term event gate: Q1 2026 full results due April 30, 2026 — do not initiate before this print.

Next update: post-Q1 2026 earnings (April 30, 2026) and post-Q2 2026 results. As of consolidation (2026-05-30), the Q1 2026 print and any subsequent price action have not been folded into this page — figures and the WATCH stance are as of the 2026-04-26 research date with a TWD 2,075 reference price.

Sources

Internal research fragments folded into this canonical page (all dated 2026-04-26):

• 2308-profile.md — /profile (company profile, segments, footprint, ownership)
• 2308-deep-dive.md — /deep-dive (full write-up: technology first-principles, value chain, sector inflection, financial analysis, DCF, position sizing)
• 2308-buy-checklist.md — /checklist (FundamentEdge gates, valuation discipline, behavioral audit, technicals, staged entry)
• 2308-mgmt-dd.md — /mgmt-dd (leadership profiles, insider ownership, shell/cross-holdings scan, credibility scorecard, governance)
• 2308.md — prior canonical page (2026-04-26, updated 2026-05-15) reconciled into this consolidation

External sources cited across fragments:

1. StockAnalysis.com — 2308 financials (income statement, cash flow, balance sheet): https://stockanalysis.com/quote/tpe/2308/
2. Alpha Spread — earnings-call summaries, intrinsic value/DCF, analyst estimates: https://www.alphaspread.com/security/twse/2308/
3. WAWT.tech — Delta's competitive edge in NVIDIA AI power ecosystem (Apr 2026)
4. Archilabs — HVDC power for AI data centers: 400V and 800V design
5. PRNewswire — Delta at NVIDIA GTC 2025; Delta COMPUTEX 2025 HVDC
6. Delta Americas — 800V HVDC at OCP Global Summit 2025
7. Texas Instruments — Delta/TI GaN partnership case study
8. Heisener Electronics — Delta/Infineon WBG partnership
9. GuruFocus — Delta ROIC 15% / P/E TTM: https://www.gurufocus.com/term/roic/TPE:2308
10. StockAnalysis — Vertiv (VRT) statistics (peer comparison)
11. TVBS World Taiwan — Q1 2026 revenue record
12. Taipei Times — Q1 2026 optimism
13. Digitimes — GaN in 800V systems, TSMC GaN exit
14. Wikipedia — Delta Electronics
15. Simply Wall St — ownership & management; Q3 2024 earnings miss; bear narrative (overvaluation / US-China decoupling)
16. MarketScreener — consensus, governance (ISS QualityScore data)
17. Investing.com — technical analysis 2308
18. Meyka — DLEGF (Delta ADR) Q1 2026 earnings preview
19. Delta IR — Chairman's Statement 2024; Leadership page; 2012 management announcement
20. The Org — Simon Chang / executive management profiles
21. Appotronics v Delta litigation record (insidertracking.com)
22. IBMI — Ping Cheng profile

Briefings:

• 2026-04-26 · Delta Electronics (2308) · https://pink.sakdiarpa.com/reports/2308.html · vault

Data caveats: Geographic revenue split not formally disclosed (inferred from customer base). EV/EBITDA not available from screeners (estimated). Insider ownership precision requires TWSE shareholder registry / MOPS — US SEC 13F is incomplete for Taiwan stocks. Q1 2026 figures are monthly-revenue-confirmed; full Q1 2026 P&L was due April 30, 2026 and is not folded into this page.

---

Consolidation queue (merged 2026-05-30)

These fragment files were folded into this canonical page on 2026-05-30 and remain live pending Pink's archive confirmation.

• [ ] 2308-profile.md
• [ ] 2308-deep-dive.md
• [ ] 2308-buy-checklist.md
• [ ] 2308-mgmt-dd.md
• [ ] 2308.md

Source updates (auto-maintained)

Vik's Newsletter (May 6, 26) - Vik's Newsletter

Delta showed a "Prefabricated AI Modular Data Center" container at Computex 2026 that integrates 800VDC in-row power and GoCool liquid cooling (260 kW and 3 MW units), cutting deployment time up to 60%; Jensen toured the booth with Delta's COO and signed the unit, though when asked if Delta is the primary 800V supplier he declined to confirm exclusivity.

Relevant to your thesis: Confirms the NVIDIA co-development relationship and 800V HVDC architecture are advancing toward product form, but Jensen's non-answer on primary supplier status is a direct signal against assumed near-monopoly — the key bear case risk is live.

Source: https://www.viksnewsletter.com/p/computex-2026-debrief

Drop/Books (Feb 25, 25) - SKOREA_20240820_2100

Morgan Stanley's August 2024 global tech cycle report lists Delta Electronics among "Most Favored — AI structural growth" stocks alongside NVIDIA and Alchip, citing it as resilient through a potential semiconductor downcycle peak.

Relevant to your thesis: Supports the structural bull case while implicitly flagging the late-cycle valuation risk the wiki already identifies as the primary bear point.

Source: dropfile://Books/Macro/SKOREA_20240820_2100.pdf