ticker: 2454.TW layer: L3-Fabless verdict: red verify_date: 2026-05-28 #

MediaTek Inc (2454.TW) — Graham Value Screen #

Sector: Semiconductors (Fabless) | Layer: L3-Fabless Screened: 2026-05-28 | Currency: TWD unless noted

PRICE TRIGGER — READ FIRST #

Abbreviated-report threshold MET (YTD >80% AND within 10% of 52w high). Nevertheless, full screen follows because the thesis question — whether AI ASIC re-rating is priced in — requires all 10 sections. Abbreviated verdict would have been: 🔴 at current price, watchlist at TWD 2,000 or below.

1. Price & Market Structure #

| | | |--|--| | Price (TWD) | 4,640 | | Price (USD equiv, 31.5 TWD/USD) | ~$147 | | OTC (MDTKF) | Effectively untradeable — delisted / no data on Yahoo | | Market Cap | TWD 7.04T (~USD 223B, ~EUR 207B) | | Shares Outstanding | 1,596M | | Float | 1,403M (88% of SO) | | 200-day MA | TWD 1,724 (stock is 169% above it) | | 50-day MA | TWD 2,469 (stock is 88% above it) | | Beta (5yr) | 1.63 |

The stock was at TWD 1,109 in late August 2025. It has rallied +318% in roughly 9 months on successive AI ASIC catalysts. Taiwan Exchange temporarily froze trading in May 2026 as daily circuit limits were breached. This is not a value trough screen — it is a post-parabolic-rally assessment.

2. Valuation #

Graham Implied Growth from current price:

• Using trailing EPS (TWD 65.89): implied perpetual growth = 31% (absurd for a $200B company)
• Using FY2027 EPS (TWD 118.20): implied perpetual growth = 15.4% (aggressive but less absurd)

The market is pricing MediaTek as a hyperscaler AI ASIC compounder at 15%+ sustained growth — not as a mobile semiconductor cyclical. This is a narrative re-rating, not a valuation re-rating.

3. Balance Sheet #

Balance sheet is pristine. Net cash of TWD 234B (USD ~7.4B, ~EUR 6.9B) = 3.3% of current market cap. D/E of 0.03x is negligible. The current ratio declining from 2.1x to 1.2x reflects growing current liabilities (payables, deferred revenue) as business scales — not distress. For a fabless model where TSMC carries the capex, this balance sheet quality is exceptional.

4. Five-Year Income Statement & FCF #

All figures in TWD billions.

*TTM FCF lower due to working capital timing. FY2025 full-year FCF TWD 137B is the right number.

Key observations:

• Revenue fell 21% 2022→2023 (smartphone downcycle). This is the cyclicality the bear case references.
• FCF has been remarkably stable: TWD 125-149B across all four years including the trough. Fabless model advantage — no capex commitment during downturns.
• Net income down 2022→2023 (TWD 118B → TWD 77B) but recovered to TWD 105-106B by 2024-2025.
• R&D intensity 25% of revenue — highest in the peer group outside NVDA, QCOM. This is the moat investment, not waste.
• SBC: fabless Taiwan companies use RSUs but typically smaller than US peers. Not separately broken out in yfinance financials; payout ratio of 81% with stable buybacks suggests SBC-adjusted FCF remains strong (see shr-001 note at bottom).

5. Dividend — Including Special TWD Pattern #

MediaTek pays TWICE per year — an annual ordinary dividend in July (based on prior year results) and a special/interim dividend in January. This is a Taiwan corporate practice, not a one-off.

• TTM yield at TWD 4,640: 1.15% (TWD 53.50 total)
• 5-year average yield: 6.3% — this was at pre-rally prices (TWD 900-1,200 range)
• Payout ratio: 81% of TTM EPS — very high, but FCF is 131% of net income so it's cash-generative
• Key pattern: In cycle peak years (2021-2022), MediaTek paid special "super dividends" — TWD 37 in 2021, TWD 73 in 2022, TWD 76 in 2023. If AI ASIC ramp delivers $2B+ revenue in 2027, there is a real possibility of super dividends returning.
• At entry price of TWD 4,640, the yield is poor (1.15%) and does not provide Graham yield protection.

6. Insider Activity (TWSE) #

TWSE insider transaction data requires direct access to Taiwan FSC MOPS system (not available via Finnhub or yfinance for .TW listed). Available qualitative information:

• Tsai family (founding family) retains significant ownership. CEO/Vice Chairman Dr. Lih Shyng Tsai (age 74) and institutional structure suggest stable ownership.
• Held by insiders: 8.1% per yfinance (heldPercentInsiders)
• Institutional ownership: 45.6% (large for a Taiwan stock — foreign institutional money)
• No specific Form 4 equivalent available. TWSE insider disclosures filed monthly with FSC MOPS but not aggregated in Western data sources.
• Risk flag: At +277% YTD, the founding family and management holding stock options or RSUs granted at prior price levels would have extreme unrealized gains — potential for insider selling. Cannot verify without MOPS access.
• shr-002 applies: systematic C-suite selling would be bearish. Cannot assess currently.

7. Analyst Coverage #

Critical observation: The consensus mean target of TWD 3,680 and median of TWD 3,500 are 20-25% BELOW the current price of TWD 4,640. Even the Strong Buy consensus is forecasting downside from here.

• Goldman Sachs raised PT to TWD 2,454 citing AI ASIC momentum (this is now 47% below current price)
• A US brokerage issued street-high of TWD 5,000 (still only 8% upside)
• The median analyst (TWD 3,500) would be a seller at current prices

This is the clearest data point of the screen: the stock has rallied through and above where analysts expected it to go. This is narrative momentum, not fundamental consensus.

8. Last Two Earnings Results #

Q1 2026 (Reported 2026-04-30) #

• Revenue: TWD 149.2B (-2.7% YoY, beat consensus TWD 146.6B by 1.8%)
• EPS: TWD 15.52 (down from TWD 18.74 a year ago, -17% YoY)
• Operating margin: 15.8% (down from 20.0% YoY)
• Net margin: 16.7% (down from 19.6%)
• Revenue mix shift: Mobile 49% (-15% YoY), Smart Edge 46% (+13% YoY)
• AI ASIC: First hyperscale ASIC on track for production

Q2 2026 Guidance:

• Revenue: TWD 140.2-149.2B (flat to -6% QoQ; -1% to -7% YoY)
• Gross margin: 46% ±1.5%

Full Year 2026: Mid-to-high single-digit % USD revenue growth. AI ASIC expected to hit ~$2B in Q4 2026 run rate. Second ASIC project targeting mass production end of FY2027.

Q4 2025 (Reported early 2026) #

• Revenue: TWD 160.5B (+25% YoY, beat) — strong smartphone restocking + early AI ASIC orders
• Gross margin: 46.9% (above guidance)
• EPS: TWD 22.9 (strong beat on higher revenue)
• AI ASIC: Initial customer design win confirmed (Google TPU v7e/v8e partnership)

Pattern: Q4 2025 was the blowout beat that triggered the stock's parabolic move. Q1 2026 was a softer-but-OK quarter. Q2 guidance is for sequential decline — the AI ASIC $2B target is heavily back-half 2026 loaded.

9. Dimensity + AI Edge + Cycle News #

Dimensity Mobile (Bear Case)

• Mobile revenue -15% YoY in Q1 2026. Seasonal + competitive.
• Qualcomm Snapdragon 8 Elite remains premium leader. MediaTek's Dimensity 9400 is competitive at mid-premium but loses design wins at the very top (Apple is TSMC direct, Samsung has Exynos/QCOM split).
• Management explicitly deprioritizing Dimensity R&D relative to AI ASIC. Long-term risk to mobile franchise.
• China smartphone recovery is real (Huawei-related), providing some support.

AI ASIC (Bull Case)

• Google TPU v7e/v8e: MediaTek secured production contracts (not just design wins). Multi-year visibility.
• Revenue doubled target: initial $1B 2026 target raised to $2B in Q4 2026 run-rate.
• Second hyperscaler ASIC in late-stage discussions (Amazon/Meta/Microsoft speculation).
• Goldman Sachs models $12.3B AI ASIC revenue by 2027 (39% of total) — if accurate, this is a transformative earnings inflection.
• TSMC CoWoS packaging allocation requested at 7x current levels — supply side confirming demand.

AI Edge (Medium-term)

• Dimensity Auto for ADAS growing (Bosch, Continental partnerships)
• Kompanio AI PC chips — small share vs Qualcomm X Elite but early foothold
• Genio IoT/edge chips — subscale but high-margin

Tariff/Geopolitical Risk

• Management running tariff impact simulations. Direct exposure limited (MediaTek sells to OEMs, not direct US consumer).
• US chip tariff on Chinese-made chips delayed to June 2027 — gives MediaTek buffer on China revenue stream.
• Taiwan geopolitical risk is a permanent discount factor — cannot be resolved analytically; VWCE holds some MediaTek exposure which provides partial hedge.

10. shr-020 Red Flag Check (Pre-Buy Diligence) #

Graham 7 Filters (PASS/FAIL) #

*Earnings stability asterisk: 2022→2023 saw NI fall from TWD 118B to TWD 77B (-35%) — a one-year deficit in profitability trend. Fails the spirit of Graham filter 3 even if no year was a net loss.

Graham Defensive Score: 2-3 PASS / 7 — FAIL by any Graham interpretation.

Cyclicality vs Value Trap (shr-017): MediaTek is NOT a value trap. Its fundamentals are excellent — net cash balance sheet, consistent FCF, strong R&D moat. The valuation failure is purely a function of a stock that has 3x'd off a legitimate trough. The trough (August 2025 at TWD 1,109, ~12x trailing P/E) would have been the Graham buy. At TWD 4,640 the stock is priced as a hypergrowth AI company, which may prove correct but is not a Graham defensive screen candidate.

Graham Intrinsic Value Sensitivity (shr-003 + shr-012) #

Using Trailing EPS (TWD 65.89) #

Using FY2027 EPS Estimate (TWD 118.20 — first full AI ASIC year) #

Implied growth from current price: 31% using trailing EPS, 15.4% using FY2027 EPS.

Key insight (shr-003): The gap between 31% implied (trailing) vs 15% implied (forward) is significant. The market is pricing in a massive earnings inflection — it's betting that TWD 118 FY2027 EPS is achievable AND that 15%+ perpetual growth follows. For Graham IV to support TWD 4,640 on a forward basis, MediaTek needs to sustain 15%+ growth indefinitely after a year when EPS more than doubles. That would require repeated hyperscaler ASIC wins at scale — possible, but priced with no margin of safety.

Graham's formula breaks down for companies with high-conviction AI growth narratives because the formula is designed for mature, stable compounders. The Growth Stock Addendum (shr-004) applies: the stock fails the addendum because implied growth (15-31%) significantly exceeds a conservative achievable estimate (8-12% for a $200B company without another Google-level win each year).

DEGIRO Access — Structural Flag #

Taiwan Stock Exchange (TWSE) is not available on DEGIRO. DEGIRO's Asian coverage: Japan (TSE), Hong Kong (HKEX), Singapore (SGX), Australia (ASX). Taiwan is explicitly not listed.

Options for EUR-based investors:

1. Interactive Brokers — offers direct TWSE access in TWD. Requires IBKR account.
2. Emerging market ETFs — VWCE (Vanguard FTSE All-World) includes MediaTek at ~0.4-0.5% weight; the user already holds VWCE exposure. An EM ETF would add more.
3. OTC (MDTKF) — effectively unavailable; no liquid market found.
4. Saxo Bank — offers some TWSE access but with currency conversion costs.

Even if access were available, at current prices this would be a 🔴 RED decision. The DEGIRO structural barrier is moot given the valuation.

Verdict #

🔴 RED — Do Not Buy at Current Price #

Summary rationale:

1. Already 3x'd from the Graham buy point. The thesis of "mobile-cycle skepticism keeping it depressed" was valid at TWD 1,109 (Aug 2025). The market has discovered and re-rated the AI ASIC story. Entry at TWD 4,640 requires you to believe in the narrative, not to buy value.

2. Analysts are forecasting 20-25% downside. The consensus mean PT of TWD 3,680 (Strong Buy consensus) is below the current price. A Strong Buy consensus forecasting downside is a reversal signal.

3. Graham IV requires 31% perpetual growth to justify current price on trailing EPS. Even using the AI-inflected FY2027 EPS, you need 15%+ sustained growth with zero margin of safety.

4. Q1 2026 and Q2 guidance are both negative YoY. The AI ASIC revenues are back-half 2026 weighted. There is a 6-month window of potentially disappointing results before the thesis is proven by actual delivery.

5. Earnings are declining in absolute terms in 2026. EPS Q1 2026 = TWD 15.52 vs TWD 18.74 a year ago (-17%). The stock is at all-time highs while near-term earnings are weaker.

6. DEGIRO cannot trade TWSE anyway. This is a moot screen for the current satellite account.

7. shr-017 valuation caution: This is not a cyclical trap (balance sheet is clean, FCF is real), but the Graham framework does not have a valid path to a "buy" at current multiples. The AI ASIC thesis is either priced in (base case) or has extreme upside if Goldman's $12B 2027 model is correct (bull case). Neither is a Graham margin-of-safety entry.

Watchlist Trigger #

Watchlist price: TWD 2,000 (~43% drawdown from today)

• At TWD 2,000: trailing P/E ~30x, fwd 2027 P/E ~17x, yield ~2.7%, Graham IV at g=15% ~TWD 2,537 (27% upside)
• This price would imply market skepticism about AI ASIC execution materializing — the right entry if a Q3 or Q4 2026 miss on the ASIC ramp creates a re-entry opportunity
• Kill condition for watchlist: if FY2027 EPS consensus drops below TWD 80 (AI ASIC execution failure), fundamental floor falls and watchlist price should move lower

SBC Adjustment Note (shr-001) #

Taiwan fabless companies have smaller SBC programs than US peers, but TSMC's stated SBC-adjusted FCF for peers suggests MediaTek's reported FCF of TWD 137B likely overstates true owner earnings by 5-12% depending on RSU grant pace in the AI-ramp buildout years. At current multiples, this adjustment does not change the verdict.

Sources: yfinance (price, financials, dividends, balance sheet — 2026-05-28), Futurum Q1 2026 analysis, TrendForce AI ASIC target, Bloomberg Google TPU tie-up, TrendForce Google TPU v7e/v8e, StockAnalysis 2454, DEGIRO markets

VERDICT LINE: 🔴 RED — TWD 4,640 is +277% YTD and at 52w high; trailing 70x P/E, analysts average 21% below current; Graham IV requires 31% sustained growth on trailing EPS; DEGIRO cannot trade TWSE. Watchlist: TWD 2,000 if AI ASIC execution disappoints in H2 2026.