The global AI investment cycle has become one of the dominant macro drivers of country-level equity returns in 2026 — but not in the way most investors assume. The countries best positioned to capture AI-driven growth are not the ones building the models. They are the ones supplying the hardware that makes those models possible.

On March 19, 2026, Bank of America economist Stephen Juneau published the first report in the bank's new AI Matters series — "AI Matters #1: The tip of the AI-ceberg?" — projecting that aggressive AI capital expenditure plans will add 0.4 percentage points to US GDP growth in 2026, with the largest spillover benefits flowing to AI hardware suppliers: Taiwan, South Korea and Mexico. The report specifically names these three countries as "the biggest beneficiaries of US and, to a lesser extent, Chinese investment," and notes that BofA's Asia economics team raised its Taiwan 2026 GDP forecast by 3.5 percentage points to 8.0% on the basis of AI demand alone.

This analysis maps the BofA AI thesis to the U.S.-listed country ETFs that provide the most direct equity exposure to each country's AI supply chain role — and examines the critical complication the Iran war introduces for the two most important markets in that supply chain.

The BofA AI Investment Framework: Key Number

s

Before mapping the thesis to country ETFs, the BofA data points that matter most for country-level equity allocation deserve enumeration:

• US private AI investment: $109 billion in 2024 — more than 10x China ($9.3B) and 5x Europe ($19.4B)
• Hyperscaler capex 2026: Projected to reach 2.1% of US GDP — nearly 16% of all non-residential US investment
• Data center investment: Expected to rise to 1.7% of US GDP in 2026, adding 0.4pp to growth
• US AI compute dominance: US controls 74% of global high-end AI compute; China 14%; EU 4.8%
• Taiwan GDP upgrade: BofA raised 2026 forecast by 3.5pp to 8.0% on AI demand
• Iran war supply chain risk: Taiwan and Korea rely on the Middle East for nearly 70% of crude oil imports and 20–25% of LNG imports

The supply chain geography is the critical insight. Servers and GPUs are primarily assembled in and imported to the US from Mexico, Taiwan, Vietnam and Thailand. Electronic circuits come mostly from Israel, Taiwan, Malaysia and Japan. This distribution is not a geopolitical coincidence — it is the product of decades of specialised manufacturing investment concentrated in a handful of countries that each dominate a specific layer of the AI hardware stack.

Tier 1: The Semiconductor Core — Taiwan (EWT) and South Korea (EWY

)

Taiwan: EWT ETF — The Irreplaceable Foundr

y

The iShares MSCI Taiwan ETF (EWT) is the most direct country ETF exposure to the AI semiconductor thesis. Taiwan Semiconductor Manufacturing Company (TSMC) accounts for 21.3% of the fund's weight — the single largest position in any country ETF in the universe, and the single most critical company in the global AI supply chain.

TSMC manufactures the chips that power every Nvidia GPU, every Apple processor and the vast majority of advanced semiconductors used in data center AI infrastructure. No other company in the world can produce chips at the 3nm and 2nm process nodes that AI model training requires at scale. That monopolistic position is not a short-term competitive advantage — it is the product of $40 billion in annual capital expenditure, decades of process engineering and a workforce of 60,000+ engineers that cannot be replicated quickly in any alternative geography.

BofA's Asia team raised Taiwan's 2026 GDP forecast by 3.5 percentage points specifically on AI demand, bringing the full-year projection to 8.0%. That is the single largest AI-driven GDP revision BofA has made to any country in its forecast universe. For a $760 billion economy, an 8% growth rate powered primarily by semiconductor export demand represents one of the most direct GDP-to-equity transmission mechanisms available.

EWT's +12.29% YTD return as of March 19 reflects a pre-war AI momentum trade that has been partially offset by the Iran war's energy import penalty. Taiwan imports approximately 70% of its crude oil requirements from the Middle East — a dependency that BofA specifically flags as an Iran war supply chain risk. The fund has held up better than EWY since February 27, because Taiwan's equity market is more purely a semiconductor/AI story and less exposed to the broad consumer and industrial sectors where energy cost increases hit margins most directly.

The breaking news from GTC 2026 (March 17) adds a live catalyst: Micron completed its acquisition of Powerchip Semiconductor Manufacturing Corporation, further consolidating Taiwan's position in the DRAM supply chain that underpins AI memory infrastructure.

South Korea: EWY ETF — The HBM Memory Monopol

y

The iShares MSCI South Korea ETF (EWY) is the AI memory trade expressed through a country ETF. Samsung Electronics (005930.KS) at 22.8% and SK Hynix (000660.KS) at 19.9% together represent nearly 43% of the fund — the highest combined weight of two AI-critical companies in any country ETF globally.

Samsung and SK Hynix are not peripheral participants in the AI supply chain. They are the only two companies in the world capable of producing High Bandwidth Memory (HBM) chips at the volumes required by Nvidia's AI GPU architectures. HBM is not interchangeable with standard DRAM — it is a specialised, stacked memory architecture that delivers the bandwidth speeds required for large language model training. Without HBM from Samsung and SK Hynix, Nvidia's H100 and B100 GPUs cannot be manufactured. The dependency is total.

The GTC 2026 conference in San Jose on March 17 made this explicit in real time. Samsung Electronics, SK Hynix and Hyundai Motor announced landmark deals with Nvidia — SK Hynix unveiled its custom HBM4 architecture at the conference while Samsung was named as the manufacturer of Nvidia's new Groq 3 LPU inference chip, with shipments targeted for Q3 2026. These are not incremental supplier relationships — they are strategic production partnerships at the frontier of the AI hardware stack.

EWY's +35.65% YTD return is the highest in the country ETF universe, driven by the pre-war AI and semiconductor momentum. The Iran war has introduced a severe energy penalty — South Korea imports approximately 98% of its energy needs, and the fund has fallen approximately -16.7% since February 27. The result is a market trading at 10.29x forward P/E — a 45% discount to the MSCI ACWI — with the AI earnings story structurally intact and a valuation that reflects the energy shock as a temporary headwind rather than a permanent impairment.

Tier 2: The Server Assembly Layer — Mexico (EWW

)

Mexico: EWW ETF — The Server Export Powerhous

e

Mexico's role in the AI supply chain receives far less investor attention than Taiwan and South Korea, but BofA's data is unambiguous: Mexico is one of the largest exporters of AI servers to the US, alongside Taiwan and Vietnam. US hyperscaler capex of 2.1% of GDP flows heavily through Mexican assembly facilities, and the nearshoring trend — accelerated by US-China supply chain diversification — has made Mexico the preferred location for server manufacturing operations serving US data centers.

The iShares MSCI Mexico ETF (EWW) does not hold pure-play semiconductor or server manufacturing companies — its portfolio is dominated by Grupo México (mining, 13.3%), Grupo Financiero Banorte (banking, 10.7%), América Móvil (telecoms, 8.1%) and FEMSA (consumer, 6.8%). The AI supply chain exposure is indirect, flowing through infrastructure demand, logistics, energy and the broader economic multiplier from Mexico's manufacturing export sector.

EWW's +4.40% YTD performance reflects the mixed picture: AI server nearshoring is a genuine tailwind, but Mexico's equity market is dominated by financial, consumer and telecom companies whose earnings are more sensitive to Mexican domestic growth and the peso than to semiconductor export volumes. The AI thesis is real for Mexico's economy — but it expresses through EWW with significant dilution.

Tier 3: The State-Led Challenger — China (GXC

)

China: GXC ETF — A Different AI Architectur

e

BofA's report describes China as the US's primary competitor in the AI race, operating under a fundamentally different model. China's AI investment should add less than 0.1 percentage points to GDP growth in 2026 — a smaller immediate impulse than the US — but the 15th Five-Year Plan earmarks $280 billion for national data infrastructure between 2025–2030, and China's AI investment is projected to grow at a 29% CAGR from $90 billion (0.5% of GDP) in 2025 to $330 billion (1.2% of GDP) by 2030.

The SPDR S&P China ETF (GXC) provides the broadest available exposure to China's AI ecosystem. Tencent (0700.HK) at 11.5% and Alibaba (9988.HK) at 8.8% are both hyperscaler-equivalent companies — building cloud infrastructure, training proprietary large language models and deploying AI across their consumer and enterprise platforms. Xiaomi (1810.HK) at 2.1% adds exposure to China's physical AI ambitions in robotics and autonomous vehicles.

GXC's -2.00% YTD return is driven by continued geopolitical risk premium and US export control headwinds rather than any deterioration in China's domestic AI investment trajectory. The BofA report explicitly notes China's strategic advantages: cheaper industrial electricity (30–60% below US/EU rates), dominance in processed critical minerals (>60% of global cobalt, lithium and rare earth supply by 2030), and a state-led coordination mechanism that can accelerate infrastructure deployment faster than market-driven models.

The valuation at 11.80x forward P/E — a 37% discount to the ACWI — prices in the geopolitical risk but arguably overstates the probability of a clean break in US-China tech supply chains. China's DeepSeek and Kimi models have closed the capability gap with US frontier models to within months, according to BofA's Epoch Capabilities Index data.

The Iran War Complication: A Supply Chain Stress Tes

t

BofA's report directly addresses the Iran war as a supply chain risk for the AI thesis — and the data is stark. Taiwan and Korea, the two most critical AI hardware suppliers, rely on the Middle East for approximately 70% of their crude oil imports and 20–25% of their LNG imports. The Strait of Hormuz disruption has therefore simultaneously:

1. Raised energy costs for Taiwan and Korea's semiconductor manufacturing operations
2. Increased shipping costs for AI hardware components transiting between Asia and the US
3. Driven up LNG prices, which power the data centers that are generating AI demand

BofA's conclusion is measured: the Iran war "could lead to less real activity due to rising costs" but "wouldn't lead to less nominal spending on AI." The distinction matters for equity investors. Nominal AI capex plans from Microsoft, Amazon, Google, Meta and Oracle remain intact at 2.1% of GDP — but if input costs rise, the real output of that spending is lower, margins at AI hardware manufacturers compress, and the GDP multiplier for AI-supplying economies is reduced.

The Polymarket probability framework provides the duration context: Hormuz normalisation by April 30 is priced at just 26%, suggesting the energy cost headwind for Taiwan and Korea persists for at least another six weeks. BofA explicitly notes it remains "comfortable" with its 8% Taiwan GDP forecast "unless geopolitical tensions are more protracted" — a phrase that maps directly to the Polymarket Hormuz disruption probability.

The Country ETF AI Scorecar

d

The scorecard reveals a bifurcated picture. Taiwan and South Korea sit at the top of BofA's AI supply chain hierarchy — identified explicitly as "the biggest beneficiaries" — but both absorb the energy import penalty from the Iran war most severely. Mexico avoids that penalty and carries a simpler AI story. China operates in a parallel AI ecosystem with genuine scale advantages in energy cost and critical minerals.

Conclusio

n

The BofA AI Matters report released March 19 provides the clearest macro framework yet for mapping country-level AI supply chain exposure to equity returns. Taiwan (EWT) and South Korea (EWY) are the Tier 1 plays — semiconductor foundry and HBM memory, respectively — with valuations that partially reflect the Iran war's energy penalty rather than any structural impairment to the AI demand story. Mexico (EWW) offers the cleanest Iran war-insulated exposure with indirect AI supply chain exposure. China (GXC) represents a deep-value bet on a parallel AI ecosystem that BofA acknowledges is closing the capability gap with the US faster than most investors recognise.

The critical variable in determining how much of the AI supply chain thesis is captured by each of these ETFs in the near term is the Hormuz timeline. A normalisation — currently priced at 74% probability of remaining disrupted through April 30 — would relieve the energy headwind for Taiwan and South Korea and restore the full AI earnings narrative for EWT and EWY without competition from the geopolitical risk discount.

Track live performance and compare EWT, EWY, EWW and GXC across any time period at countryetftracker.com. Use the Compare Tool to model the AI supply chain basket versus the S&P 500 from any start date.

Frequently Asked Question

s

Which country ETF has the most direct AI exposure?

The iShares MSCI Taiwan ETF (EWT) provides the most direct AI exposure through TSMC's 21.3% weight — the world's only manufacturer of advanced AI chips at 3nm and below. The iShares MSCI South Korea ETF (EWY) is a close second through Samsung (22.8%) and SK Hynix (19.9%), which together monopolise HBM memory supply for Nvidia's AI GPUs.

What does Bank of America say about AI and country ETFs?

In its March 19 AI Matters report, BofA economist Stephen Juneau identifies Taiwan, Korea and Mexico as "the biggest beneficiaries of US and, to a lesser extent, Chinese AI investment." BofA raised its Taiwan 2026 GDP forecast by 3.5 percentage points to 8.0% on AI demand, and projects AI capex will add 0.4pp to US GDP growth in 2026.

How does the Iran war affect the AI country ETF thesis?

Taiwan and Korea rely on the Middle East for approximately 70% of crude oil imports and 20–25% of LNG imports. The Hormuz disruption has raised energy costs for semiconductor manufacturing and AI hardware shipping. BofA notes this could reduce the real economic output of AI investment without reducing nominal spending plans.

Why is South Korea trading at such a deep discount despite leading AI supply chains?

South Korea's iShares MSCI South Korea ETF (EWY) trades at 10.29x forward P/E — a 45% discount to the MSCI ACWI — because the market is pricing the energy import shock as a near-term earnings headwind. Samsung and SK Hynix's AI earnings power (HBM demand, GTC 2026 Nvidia deals) remains structurally intact. The valuation discount reflects the current macro regime, not a change in the competitive position of Korea's AI hardware manufacturers.

What is Mexico's role in the AI supply chain?

According to BofA's analysis, Mexico is one of the largest exporters of AI servers to the US alongside Taiwan and Vietnam. The nearshoring trend has concentrated server assembly capacity in Mexico as US hyperscalers seek to reduce China dependency in their supply chains. The iShares MSCI Mexico ETF (EWW) captures this indirectly through infrastructure, logistics and manufacturing sector exposure.