Asia’s AI trade sell-off: A warning sign or healthy reset?

• South Korea's leveraged semiconductor ETF unwind triggered a sharp market sell-off on 23 June, highlighting the risks of crowded positioning and excessive retail leverage. The decline was further amplified by a more hawkish Federal Reserve and growing concerns over the rising financing costs of AI infrastructure investment.

• Despite the correction, AI fundamentals remain intact. Record hyperscaler capital expenditure, strong semiconductor demand and persistent supply constraints in advanced chips and high-bandwidth memory continue to support the sector's growth outlook.

• The recent pullback appears to be a valuation reset rather than a deterioration in earnings prospects. The next phase of the AI cycle is likely to reward companies that can deliver earnings growth, rather than those relying solely on narrative-driven valuations.

• Investors should focus on high-quality AI beneficiaries and avoid leveraged products. Our preferred vehicle for accessing Asia's semiconductor opportunity remains the Global X Asia Semiconductor ETF (HKEX:3119), capturing the region's AI supply chain while reducing single-market risk.

• Diversification remains important in a more volatile environment. Singapore provides exposure to domestic and financial sector growth drivers that are less dependent on the AI cycle, while China offers attractive valuations and a relatively self-reliant technology and semiconductor ecosystem that can help cushion global technology market volatility.

Just two weeks after Asia's sharp sell-off on 8 June, investors were forced to confront another bout of extreme volatility.

Related article: The Asia bull stumbles – but we believe it’s a reset, not a regime change

On Monday evening, comments from South Korea's Financial Supervisory Service (FSS) Governor triggered a wave of selling that quickly spiralled across regional markets. Circuit breakers were activated twice on the Korea Exchange on 23 June 2026, as investors rushed to unwind positions in what had been the world's best-performing equity market this year. By the closing bell, South Korean equities had plunged 9.5% in SGD terms — their worst single-day decline in months.

The shock did not stay confined to Seoul. Volatility spread rapidly across Asia, dragging down technology and semiconductor stocks from Tokyo to Taipei and Singapore. The global equity fear gauge surged as investors once again asked a familiar question: Is this the beginning of the end for the AI trade?

Figure 1: Performance of Asian markets on 23 June 2026

Multiple headwinds converged to trigger the sell-off

As with the correction on 8 June, Tuesday's sell-off was not driven by a single event. Rather, it reflected the concerns about the sustainability of AI trade, and growing macroeconomic uncertainty.

The Spark: Korea's leveraged ETF frenzy unravels

The immediate trigger emerged on Monday evening when FSS Governor Lee Chan-jin publicly expressed regret over the late-May launch of 16 single-stock leveraged ETFs linked to Samsung Electronics and SK Hynix. Originally introduced to encourage Korean investors to remain in domestic markets and support the weakening Korean won, the products quickly evolved into vehicles for speculative trading.

Margin loans in South Korea — a key indicator of retail leverage — first exceeded KRW38 trillion on 29 May and remained near record highs throughout June, according to Korea Financial Investment Association data. The 16 leveraged ETFs, comprising both spot and futures-based long and inverse products, grew from approximately KRW4.1 trillion in assets at launch to nearly KRW14 trillion by 22 June — a staggering 241% increase in less than a month. Even more striking was the composition of investors. Retail traders accounted for roughly 92% of holdings, while average daily turnover reached 122.5%, implying an average holding period of just three to five days.

Once regulators signalled concern, the unwind was swift and brutal. As prices fell, leveraged investors faced margin calls, forcing additional selling. Those forced liquidations pushed prices lower still, triggering further margin calls and creating a self-reinforcing feedback loop that accelerated the decline.

The impact quickly spread beyond the leveraged products themselves. With Samsung Electronics and SK Hynix accounting for roughly half of the KOSPI's market capitalisation and the index, any disorderly unwind in these names was bound to reverberate through the broader market.

The Accelerant: A more hawkish Fed and rising questions around AI financing

The Korean sell-off landed on already fragile market sentiment. At its 17 June FOMC meeting, the Federal Reserve left interest rates unchanged. However, the revised dot plot lifted the median year-end 2026 policy rate projection to 3.8%, up from 3.4% in March. The updated forecasts effectively removed expectations for rate cuts this year and priced in possibilities of more than one rate hikes.

For high-growth technology companies, higher interest rates matter. Much of their valuation depends on future earnings, and a higher discount rate reduces the present value of those cash flows. In other words, the AI sector's valuation premium becomes harder to justify when rates remain elevated.

At the same time, SpaceX's inaugural bond issuance on 22 June introduced fresh questions around the financing sustainability of the AI infrastructure buildout. The company is reportedly seeking to raise at least USD 20 billion, with proceeds primarily intended to refinance a bridge facility arranged earlier this year.

The transaction surprised some investors, given reports that SpaceX holds in excess of USD 100 billion in cash following its IPO. While the rationale remains unclear, markets interpreted the move as a potential indication of the enormous capital requirements associated with the next phase of AI-related infrastructure investment. More importantly, the issuance highlighted a broader concern facing the technology sector. As AI spending scales, even the largest technology companies may increasingly rely on external financing rather than internally generated cash flows. Such a shift could place downward pressure on returns on capital through higher financing costs, rising depreciation expenses and longer payback periods. In this context, the bond offering became more about what it may imply for the economics of the wider AI ecosystem.

The Catalyst: Micron earnings Shadow

Hovering over Tuesday's chaos was the looming earnings report from Micron Technology, due 24 June 2026. Micron is an important single data point for the AI memory trade as its results carry implications for every HBM supplier, every DRAM maker, and every hyperscaler capex budget globally. Pre-earnings profit-taking is common but this time it is amplified by a Korean meltdown and a hawkish Fed, is something else entirely. Micron shed over 13% on Tuesday in USD terms, despite following the announcement of a strategic supply partnership with Anthropic.

Is the AI trade broken? We believe the answer is no

The recent correction has reignited concerns that the AI boom may be approaching its peak. Those concerns are valid and should not be dismissed lightly. Semiconductor valuations remain elevated, investor positioning has become crowded, and higher interest rates create a more challenging environment for long-duration growth assets.

Yet the evidence today does not resemble the conditions that typically precede a technology bust.

Historically, major technology cycles peak when earnings growth slows, inventories accumulate, order books weaken and capital expenditure plans are cut. None of those warning signs are visible at scale today.

Instead, the AI investment cycle remains remarkably robust. Hyperscalers continue to commit unprecedented sums to AI infrastructure. Microsoft spent USD 30.9 billion on capital expenditure in fiscal Q3, up 84% year-on-year, while its AI business surpassed a USD 37 billion annualised revenue run rate. Meta recently raised its full-year capital expenditure guidance to USD 125-145 billion, citing continued demand for AI infrastructure, data centres and supporting hardware.

More importantly, spending is translating into real demand rather than speculative projections. SK Hynix has reportedly committed most of its 2026 HBM3E production capacity to key customers including Nvidia. Meanwhile, TSMC's advanced-node manufacturing capacity remains heavily booked through 2028 as demand for leading-edge 3nm and 2nm chips continues to accelerate.

The earnings outlook reflects this strength. We estimate the FactSet Asia Semiconductor Index could deliver earnings growth of approximately 169% in 2026. TSMC recently raised its revenue growth outlook to above 30% and is expected to spend toward the upper end of its USD 52-56 billion capital expenditure budget to expand capacity. Likewise, both Samsung Electronics and SK Hynix continue to highlight the possibility of HBM shortages extending into 2027 as demand continues to outpace supply.

In our view, the recent sell-off represents a valuation reset rather than a fundamental deterioration in the AI story. Markets are beginning to recognise that even the strongest secular growth themes remain subject to the realities of valuation, interest rates and risk management. The era of pricing semiconductor companies for perpetual exponential growth may be ending. What follows is likely to be a more mature phase of the cycle — one where earnings delivery matters more than narrative, and where stock selection becomes increasingly important.

What should investors do now?

The recent correction is uncomfortable, but it does not fundamentally alter our view on the AI investment cycle. Investors should focus on separating short-term market dislocations from long-term business fundamentals.

Immediate actions: Avoid panic selling, reduce leverage exposure

1. Avoid panic-selling quality semiconductor leaders

The latest sell-off in technology stocks was driven primarily by positioning and leverage rather than a meaningful deterioration in fundamentals. The forced unwinding of Korean leveraged ETFs created indiscriminate selling pressure across the semiconductor ecosystem, dragging down companies whose earnings outlook remains intact.

For long-term investors, this distinction matters. Companies such as TSMC, which fell sharply despite no material change to its business outlook, represent the type of high-quality franchise that investors should evaluate for accumulation rather than abandon during periods of market stress.

2. Reduce exposure to leveraged semiconductor products

While the broader AI investment theme remains attractive, leveraged single-stock ETFs have become increasingly vulnerable. The FSS has openly signalled discomfort with the rapid growth of these products, and further regulatory measures cannot be ruled out. Combined with the high concentration of retail investors and elevated turnover rates, the risk of future forced liquidations remains significant.

Investors seeking exposure to semiconductor leaders should favour unleveraged vehicles or direct equity investments instead of attempting to chase short-term rebounds through leveraged products.

Our preferred vehicle to particate in Asia's semiconductor opportunity remains the Global X Asia Semiconductor ETF (HKEX:3119). The fund provides exposure across South Korea, Taiwan, Japan and China, reducing single market risk while maintaining participation in the broader AI supply chain that spans across the whole value chain in Asia.

3. Pay close attention to Micron's earnings

Micron's upcoming earnings release may provide the most important near-term read-through for AI infrastructure demand. More than the headline numbers, investors should focus on management's guidance regarding HBM demand, DRAM pricing and customer order visibility. The results will offer valuable insight into whether AI-related memory demand continues to exceed supply or whether growth is beginning to normalise. Given Micron's central role in the AI memory ecosystem, its outlook could significantly influence sentiment across the broader semiconductor sector.

Medium-term positioning: Focus on fundamentals, not momentum

1. Differentiate ruthlessly within the semi complex

The semiconductor sector is no longer moving as a single trade. Investors should increasingly distinguish between companies benefiting directly from AI infrastructure spending and those still reliant on cyclical consumer demand. The strongest opportunities remain concentrated in areas where demand visibility is highest, including AI accelerators, high-bandwidth memory, advanced packaging and leading-edge foundry capacity. Companies such as SK Hynix, Micron and TSMC continue to enjoy direct exposure to these structural growth drivers through their positions in HBM, AI memory and advanced semiconductor manufacturing. By contrast, segments tied primarily to smartphones, PCs and traditional consumer electronics may continue to face a more challenging demand environment.

2. Prepare for a higher-rate environment

Build in a rate-sensitivity buffer. With the Fed's dot plot shifting higher and the market pricing in a more hawkish rate outlook, long-duration growth assets are facing increasing valuation pressure from higher discount rates. Prioritise companies where earnings growth can outpace multiple compression. Names trading at elevated forward earnings multiples should be approached with extreme caution in this environment.

3. Diversification remains an underappreciated source of resilience

While AI remains one of the most compelling structural growth themes globally, portfolios should not become overly dependent on a single investment narrative. Singapore and China offer attractive diversification benefits for different reasons.

Singapore's equity market remains heavily weighted toward financials, industrials and domestic economic beneficiaries. This makes the Straits Times Index less dependent on global semiconductor cycles and AI-related capital expenditure. Higher interest rates continue to support bank profitability, while Singapore's stable economic backdrop and ongoing capital market development initiatives should provide resilience during periods of heightened volatility.

For income-oriented investors, the Amova Singapore Dividend Equity SGD offers exposure to high-quality dividend-paying companies. Investors seeking broader participation in Singapore's economic growth, including exposure to small- and mid-cap companies, may consider the iFAST-Amova Singapore Equity A SGD.

China provides a different form of diversification. While still influenced by global technology trends, China's semiconductor and technology ecosystem is increasingly driven by domestic innovation, localisation efforts and policy support. This creates a degree of independence from global liquidity cycles and may help cushion portfolios during periods of volatility in international technology markets. Chinese equities also continue to trade at relatively attractive valuations compared with global technology peers.

For active exposure, the Fidelity China Focus A-SGD provides access to China's long-term growth opportunities through a diversified portfolio. Investors seeking more targeted exposure to China's technology and AI ecosystem may consider the iShares Hang Seng Tech ETF (HKEX:3067), which provides exposure to leading internet, AI application and platform companies. For investors looking to participate in China's semiconductor self-sufficiency and hardware buildout, the GF All-Share Information Technology ETF (SZSE:159939) offers access to many of the country's key semiconductor and technology supply chain players.

Declaration:

This research report was prepared with the assistance of artificial intelligence (AI) tools. iFAST Financial Pte Ltd does not rely exclusively on AI for content generation; the content of this report – including all investment theses, ratings, price targets and conclusions – has been independently reviewed and verified by the research analyst(s) to ensure accuracy and professional integrity.

For specific disclosure, at the time of publication of this report, IFPL (via its connected and associated entities) and the analyst who produced this report hold a NIL position in the abovementioned securities.