Semiconductor exposure for Indian investors — Taiwan, Korea, and the US compared

The semiconductor industry is the closest thing the modern economy has to a single point of leverage. Everything — AI training clusters, phones, cars, weapons — runs on chips, and the supply chain that makes them is concentrated in a handful of companies across three geographies. For an Indian investor who wants exposure to this theme, the question is not really "which chip stock" but "which slice of the chip supply chain, owned through which market, with what tax and estate-tax consequences."

The three doors are Taiwan (the foundry — where chips are physically manufactured, dominated by TSMC), South Korea (memory — Samsung and SK Hynix, who make the high-bandwidth memory that AI accelerators depend on), and the US (chip design and equipment — NVIDIA, Broadcom, AMD, plus the broad listed-semiconductor universe). Each maps to a different ETF, a different risk profile, and a different tax treatment for an Indian resident. This guide compares all three so you can build deliberate chip exposure rather than accidentally tripling up on the same names.

The three layers of the chip supply chain

Before the funds, understand what each geography actually represents, because they are not substitutes — they are different links in one chain.

• Taiwan = the foundry layer. TSMC manufactures the leading-edge logic chips that almost every fabless designer (NVIDIA, Apple, AMD) relies on. If you own Taiwan, you own the bottleneck of advanced manufacturing.
• Korea = the memory layer. Samsung and SK Hynix dominate DRAM and NAND, and crucially the high-bandwidth memory (HBM) that sits next to every AI GPU. SK Hynix in particular has been a primary HBM supplier to NVIDIA's flagship accelerators.
• US = the design and equipment layer. NVIDIA, Broadcom, AMD design the chips; the US also hosts the broadest pool of listed semiconductor names, and US-listed chip ETFs capture global leaders including TSMC's own ADR.

Because these are different layers, a fully diversified chip allocation might touch all three. But the access and tax friction differs sharply by geography, and that is what shapes the practical decision for an Indian.

It helps to picture the dependency chain. NVIDIA designs an AI accelerator in California. TSMC in Taiwan fabricates the logic die on its leading-edge process. SK Hynix or Samsung in Korea supplies the high-bandwidth memory stacked alongside it. The whole thing is assembled — increasingly by Hon Hai/Foxconn in Taiwan — into a server that ends up in a data centre. Each step is a near-monopoly or duopoly held by a tiny number of firms, which is exactly why the chip supply chain is both extraordinarily profitable and extraordinarily fragile. Owning "semiconductors" without deciding which step you mean is how investors end up accidentally tripling their exposure to one company. The geography lens is really a lens on which choke-point you are buying.

The vehicles, compared

	Taiwan (EWT)	Korea (EWY)	US (SOXX / SMH)
What you get	Taiwan market, TSMC-led	Korea market, memory-led	Global listed chip names
Listing	NYSE (US-listed)	NYSE (US-listed)	NYSE / Nasdaq (US-listed)
Top names	TSMC (~21% capped)	Samsung + SK Hynix (~half the fund)	NVIDIA, Broadcom, AMD, TSMC ADR
Expense ratio	~0.59%	~0.59%	~0.34% (SOXX)
Concentration	TSMC-dominant	Memory-duopoly-dominant	Capped (SOXX) / top-heavy (SMH)
US estate-tax exposure	Yes	Yes	Yes

All three of these mainstream vehicles are US-listed, which is the single most important thing to register up front: the easy, retail-accessible way to own any of these geographies from India is through a US-listed wrapper, and every US-listed wrapper carries US estate-tax exposure. We return to that below — it is the great equaliser across all three doors.

Taiwan — EWT

EWT (iShares MSCI Taiwan) gives you the Taiwan market with TSMC capped near 21%, plus the wider Taiwanese tech supply chain. It is the diversified way to own the foundry layer. If you want pure TSMC, the TSM ADR is cleaner and cheaper; if you want diversified Taiwan, EWT is the accessible choice (the local Taiwan 50 ETF is cheaper but largely out of reach for retail Indians). Either way, you own the manufacturing bottleneck — high quality, but geographically concentrated in one island with the geopolitical risk that implies.

Korea — EWY

EWY (iShares MSCI South Korea) is heavily weighted to Samsung Electronics and SK Hynix, which together make up roughly half the fund — so EWY is, in large part, a memory-semiconductor bet wrapped in a country ETF. Korea's appeal in the current cycle is HBM: the memory that AI accelerators cannot function without, where SK Hynix and Samsung are the dominant suppliers. Note one quirk covered in the Korea market guide: Samsung Electronics is not available as a US ADR, so for individual-stock access you would need direct KRX access (easier since Korea abolished the Investor Registration Certificate at end-2023, but still paperwork-heavy) — which is why EWY is the practical retail route to Korean chips.

US — SOXX and SMH

The US chip ETFs capture the design-and-equipment layer plus global leaders. SOXX (iShares Semiconductor) tracks a US-listed semiconductor index with holdings capped near 10%, giving a relatively balanced spread across NVIDIA, Broadcom, AMD, Micron, and others — at a low ~0.34% expense ratio. SMH (VanEck Semiconductor) holds a more concentrated ~25-name basket with high single-stock caps, so it carries a heavy NVIDIA weight (around 17%) and a meaningful TSMC ADR slug (around 10%). SMH is the higher-conviction, more top-heavy bet; SOXX is the broader, more diversified one. Both, importantly, already give you indirect Taiwan exposure through the TSMC ADR — a reason to be careful about double-counting if you also hold EWT or TSM directly.

The double-counting trap

This is the mistake to avoid. Because TSMC is the foundry for the entire industry and appears (via its ADR) inside US chip ETFs, it is easy to end up massively overweight one company without realising it. Consider an investor who buys the TSM ADR for "TSMC exposure," EWT for "Taiwan exposure," and SMH for "chip exposure." All three contain TSMC — directly, at 21%, and at ~10% respectively. You have not diversified; you have built a leveraged TSMC position with extra expense ratios.

Build chip exposure by layer, not by ticker count. Decide how much foundry (Taiwan), how much memory (Korea), and how much design/equipment (US) you want, then pick the single most efficient vehicle for each layer rather than stacking overlapping funds. For most investors, one Taiwan vehicle plus optionally one US chip ETF is plenty; Korea is the add-on for those with a specific memory/HBM thesis.

The double-counting also runs through your other holdings, not just within your chip basket. If you own a broad US total-market fund or an S&P 500 fund, you already hold NVIDIA, Broadcom, AMD, and the other US chip designers at their market weights — and after the AI run-up, those weights are large. If you own a broad emerging-markets fund, you already hold TSMC and the Korean memory names. So a "dedicated" chip allocation is rarely additive in the way investors assume; it is concentrating exposure you partly already have. Before adding SOXX or SMH, look through your core funds and ask how much chip exposure you are already carrying. The honest answer for most globally diversified Indian portfolios is "quite a lot," which argues for a smaller, more deliberate dedicated tilt rather than a large one.

A note on volatility, too. Semiconductors are the most cyclical corner of the technology complex — capital-intensive, demand-swingy, and prone to violent boom-bust cycles as capacity over- and under-shoots. The AI cycle has masked this recently with extraordinary demand, but chip stocks have historically fallen 40-50% in downturns. A concentrated chip bet is therefore a high-beta position that can test your conviction badly in a bad year. Size it as the aggressive satellite it is, not as a sleep-easy core.

The tax stack across the three

For an Indian resident, the India-level tax is identical across all three because they are all foreign assets — but the source-country withholding differs.

Source-country dividend withholding

Geography	Non-resident dividend WHT	India treaty rate
Taiwan	21%	12.5%
South Korea	22%	15%
US (the ETF wrapper itself)	30% statutory	25% under India–US DTAA

For the Taiwan dividend mechanics see the dedicated guide; Korea's 22% rate and 15% treaty cap are covered in the Korea market guide. In all cases, because you typically hold these as US-listed ETFs, the withholding interacts at the fund level, and you recover what you can in India via the foreign tax credit.

India-level tax (the same for all three)

• LTCG (24+ months): 12.5% without indexation, no Rs 1.25 lakh exemption (foreign assets do not get it).
• STCG (under 24 months): added to income, taxed at slab (up to ~39%).
• Dividends: slab rate, with Form 67 (being renumbered Form 44 from TY2026-27) for the FTC.

Chip ETFs tend to be low-yield (these are growth companies that reinvest), so the dividend friction is modest and most of your return — and tax — comes through capital gains. Model it with the US capital gains calculator and the Form 67 FTC calculator.

Geopolitical risk is the elephant in the room

No honest comparison of these three geographies can skip concentration risk of a different kind: geopolitics. The single biggest tail risk in the entire chip supply chain is a disruption to Taiwan — whether through cross-Strait tension, a blockade scenario, or a natural disaster on an island that fabricates the world's most advanced logic. Because TSMC is irreplaceable on a multi-year horizon, that risk is not diversifiable within the chip theme; it sits underneath all three doors to varying degrees.

Taiwan exposure (EWT, the TSM ADR) is most directly exposed. Korea (EWY) is exposed both to its own peninsula tensions and, indirectly, because a Taiwan disruption would scramble the memory firms' biggest customers. Even US chip ETFs (SOXX, SMH) carry the risk through TSMC's weight and through US designers' dependence on Taiwanese fabrication. The US has been pushing to onshore advanced manufacturing (the CHIPS Act, new Arizona fabs), and that gradually reduces — but does not eliminate — the concentration. For an Indian investor, the practical implication is humility about position size: a chip allocation is a bet not just on demand for chips but on the continued stability of a specific, contested region. That is a perfectly investable view, but it should be a sized view, not a maximal one.

The estate-tax equaliser

Here is the point that flattens the differences between the three doors. EWT, EWY, SOXX, and SMH are all US-listed, which makes them US-situs assets for US estate-tax purposes. Whichever geography you choose, if you access it through the mainstream US-listed ETF, you are inside the US estate-tax net — where the non-resident exemption is $60,000 and the top rate is 40%, with no India–US estate treaty to soften it.

This means estate tax is not a reason to prefer one geography over another among these funds — they are all equally exposed. It is, however, a reason to manage your total US-situs portfolio (chip ETFs plus your VOO, your direct US stocks, everything) against that $60,000 line, and to consider whether large positions should sit in non-US-domiciled wrappers instead. The full mechanics and the workarounds (UCITS funds, Indian feeders, keeping direct US-situs holdings modest) are in the $60,000 trap guide. The cleaner-but-harder routes — direct TWSE Taiwan shares or direct KRX Korean shares — sit outside US estate tax, but they carry the foreign-custody paperwork that makes them impractical for most retail investors.

Putting it together — three sample stances

The simple foundry-plus-design stance: one Taiwan vehicle (TSM ADR for the pure bet, or EWT for diversified Taiwan) plus optionally a US chip ETF (SOXX for breadth). Skip Korea unless you specifically want memory. This is the default for most Indians who want chip exposure without overcomplicating it.

The full-supply-chain stance: EWT (foundry) + EWY (memory) + SOXX (design/equipment), sized deliberately, watching for the TSMC double-count inside SMH if you use it. This is for the investor who genuinely wants all three layers and will rebalance them.

The single-fund stance: if you want one ticker for the whole theme, a broad US chip ETF (SOXX) already spans US designers, the TSMC ADR, and equipment makers. It misses pure Korean memory exposure, but it is the lowest-effort, lowest-cost way to own the theme — at the cost of US-situs estate exposure on the whole position.

A quick word on SOXX versus SMH if you go the US-chip-ETF route, because the two are not interchangeable. SOXX caps individual holdings near 10% and tracks a broader index, so it is the more diversified, lower-volatility choice — closer to "own the semiconductor industry." SMH runs a tighter ~25-name basket with high single-stock caps, which currently means a heavy NVIDIA weight (around 17%) and a meaningful TSMC ADR slug (around 10%). SMH is therefore really a concentrated bet on the largest winners — superb in an AI-led melt-up, more painful if leadership rotates. For a diversification-seeker, SOXX is the cleaner default; for someone who explicitly wants to lean into NVIDIA-and-TSMC dominance, SMH does that, though at that point you might ask whether you should just own those two names directly. The lower 0.34% expense ratio on SOXX is a further mark in its favour for a buy-and-hold core.

If your real conviction is the single foundry name, remember that none of these baskets beats simply owning the TSM ADR for pure TSMC exposure — the ETFs dilute it (deliberately) and add a fee. Use the ETFs when you want the industry; use the ADR when you want the company.

Whatever stance you take: avoid double-counting TSMC, prefer total-return over yield, file Form 67 for any dividends, disclose every holding in Schedule FA, run remittances through the LRS/TCS calculator, and keep an eye on your aggregate US estate-tax exposure.

The Taiwan hub collects the full Taiwan series, the Korea guide covers the memory side in depth, and the markets overview sets all of this against the other global markets we write about.

---

This is general information, not investment or tax advice. Fund weights, expense ratios, withholding rates, and treaty terms reflect the position as understood in early 2026 and change over time; verify live figures before investing. The India–Taiwan and India–Korea dividend rates should be confirmed against the operative agreements.