Is the AI Trade in Asia Running Out of Steam?

The data in the graph above points to a clear shift in positioning. After 18 months of strong investor demand for AI-related semiconductor stocks, short interest across Asian semiconductor names has fallen back to levels last seen before the trade started to gather pace. Securities lending data tracking the percentage of market capitalization on loan for Asia's Semiconductors & Semiconductor Equipment sector shows the metric at 0.679% as of mid-July 2026, down 37.6% from its peak of 1.087% in June 2025.

This is not only a technical data point. It suggests a change in how institutional investors are positioning around what had recently been one of the most crowded trades in global equities.

The Peak That Was

In June 2025, the AI infrastructure buildout was running at full speed. Demand for advanced chips continued to exceed available manufacturing capacity at TSMC, while SK Hynix's high-bandwidth memory attracted premium pricing as hyperscalers sought additional supply. The average percentage of market capitalization on loan during that period exceeded 1.03%, as hedge funds borrowed shares actively, not primarily for bearish positions, but to hedge investments and execute trading strategies around volatile names.

In Q2 2026, the market backdrop changed. Year-to-date, the lending metric has averaged 0.68%, with the 30-day average remaining at 0.67%. The year-over-year decline from July 2025 stands at -25.7%, indicating a reduction in positioning that has coincided with the sector's recent price correction.

A Tale of Two Markets

Individual stock data points to a more differentiated market. TSMC (2330.T), a leading company in advanced chip manufacturing, currently carries the lowest short interest in the sector at 0.06% of market capitalisation on loan. This suggests limited short positioning in a company whose technology supports many major AI systems. MediaTek (2454.TW), Taiwan's mobile chip company, also has relatively light short positioning at 0.38%.

By comparison, the sector's most-shorted names are led by Montage Technology Co Ltd (6809), with 20.05% of market capitalization on loan, followed by Flat Glass Group Co Ltd (6865) at 14.01%, Techwing Inc (089030) at 11.46%, Realtek Semiconductor Corp (2379) at 10.42%, and Omnivision Integrated Circuits Group Inc (501) at 10.27%. This concentration of short interest suggests investors are taking a more targeted approach, differentiating between companies with direct exposure to the AI supply chain and those where earnings expectations, valuations, or demand sensitivity may be more vulnerable.

July's Reality Check

The past fortnight has provided a test for AI-related semiconductor positioning. On July 2nd, the KOSPI recorded its steepest single-day decline in 17 years, falling 8% after reports that Meta (META) was planning to monetise excess data centre capacity. Markets interpreted this as a possible indication that AI chip demand growth may be moderating. Samsung Electronics (005930.KS) fell 9%; SK Hynix (000660.KS) declined 14.6%, its largest daily fall since the 2008 financial crisis.

The weakness was also seen outside Korea. Japan's Nikkei dropped 3.4%, while Taiwan's TAIEX declined by more than 3%. By mid-July, further pressure emerged: SK Hynix fell 11% and Samsung declined 8% on July 16th as investors awaited TSMC's earnings and looked for evidence that AI capital expenditure would continue to grow through 2027.

Samsung's quarterly profit surpassed both Nvidia (NVDA) and Apple (AAPL), but its shares still fell 8%. This indicates that, in the current market environment, earnings beats alone may not be sufficient. Investors appear to be looking for evidence that AI-related spending can continue to support earnings growth over a longer period.

A More Discriminating Market

The data suggests not the end of the AI trade, but a more selective phase. Broad-based buying across semiconductor names appears to have moderated. The decline in securities lending activity, from over 1.08% to below 0.68% of market capitalization on loan, reflects institutional investors reducing risk, unwinding hedges, and rotating out of crowded positions.

Looking ahead, the market may continue to differentiate between companies with stronger fundamentals and those with less direct exposure to AI demand. TSMC's July 16th earnings call, SK Hynix's US listing, and Big Tech's AI investment guidance will help indicate whether this correction represents a reset in valuations or the start of a more prolonged consolidation.

AI-related investment in Asia remains significant, but returns may become more dependent on company-specific fundamentals and evidence of sustained demand.