During periods of market panic, it is essential to analyze the data with a clear and rational perspective.
Recent market declines have been driven by a negative liquidity feedback loop and a concentrated release of pessimistic sentiment. In reality, there have been no significant fundamental changes since this downturn began. Even the release of a record-breaking quarterly report from TSMC this week, which also included a substantial increase in capital expenditure guidance, failed to lift the prevailing market gloom. The core issue remains that liquidity-driven feedback has amplified market volatility and negative sentiment, turning even positive news into a perceived negative.
However, it is precisely during such panicked times that a rational examination of the data becomes most critical.
First, the sentiment indicators we track have fallen back into bottoming territory, suggesting that the adjustment phase driven purely by liquidity and sentiment may be nearing its end.
The RSI indicator, which measures market sentiment, has dropped into the oversold region below 30. Historical instances where it reached this level (September 2024, January 2025, April 2025, December 2025, March 2026) have all presented significant bottom-fishing opportunities.
Secondly, the "margin financing purchase ratio" indicator, which gauges sentiment in the margin trading sector, has declined to its lowest level since July of last year. Following the recent concentrated outflows, risk has been substantially released.
This is further supported by the recent significant net inflows from market-stabilizing funds, which help to anchor market expectations and mitigate the risk of a negative funding feedback loop.
Secondly, the magnitude of the current correction in the primary AI investment theme has approached or even exceeded historical extremes for such thematic adjustments.
For the US market, the Philadelphia Semiconductor Index has fallen over 20% since the adjustment began on June 23rd. This decline is second only to the mid-2024 correction driven by weaker-than-expected fundamentals and the trade war-induced correction in early 2025.
For the A-share market, the TMT index has declined 23.72% since the adjustment started on July 1st. This exceeds the magnitude of all previous AI-themed corrections since 2023 and is historically second only to the correction in the new energy theme in early 2020 caused by the pandemic.
Considering that this correction is not due to systemic risk or a substantive slowdown in fundamentals, the current extent of the decline appears relatively sufficient.
Finally, the market's sharp single-day decline has sparked fears among some investors that the uptrend has ended. However, a review of history shows that such sharp pullbacks are not uncommon during bull markets, and a single-day plunge does not signal the end of a trend. Following a sharp decline, the probability of a short-term oversold rebound actually increases. The long-term sustainability of the trend still depends on fundamentals.
We reviewed the performance of the ChiNext Index following single-day declines exceeding 5% during bull markets since 2010. First, the probability of a short-term oversold rebound increases. On average, the index rose by 1.2%, 3.9%, and 4.68% one, five, and twenty trading days after such a large drop, respectively. Second, "bull markets often experience sharp corrections." A single-day plunge does not constitute a signal for the end of the trend; the long-term trajectory of the index still hinges on fundamentals. Unless, as in 2015, there is a lack of fundamental support leading to prolonged weakness, such short-term sharp declines often provide buying opportunities for subsequent market moves.
As for fundamentals, there is currently insufficient evidence to suggest a slowdown in AI sector momentum.
In the A-share market, mid-year earnings previews confirm that AI-related demand remains the most significant driver of corporate performance, and its positive impact is continuously spreading along the upstream and downstream supply chains.
Overseas, the earnings expectations for the 25 key AI companies we track, as well as the long-term capital expenditure expectations for hyperscalers, are still being revised upwards against the market trend.
The core verification ahead lies in the US earnings season in late July. After the emotional volatility, the market is poised to return to a rational assessment of fundamentals. Following communications from major North American tech giants, right-side signals regarding the global computing power cycle and the earnings sustainability of AI hardware companies will become clearer.
Additionally, late July will also see policy meetings from major global central banks and China's Politburo meeting. Once key signals regarding AI industry demand, the global monetary policy path, and domestic policy directions become clearer, making directional and sectoral choices will be much more straightforward.
Sectors Deserving Attention After the Sell-off
Following the recent panic-driven sell-off, technology and growth sectors have become more attractively valued. First, our tracking of sector crowding shows a "rotation from old to new" and "rotation from high to low": low-positioned consumer sectors like pharmaceuticals, beauty & personal care, and animal husbandry have seen their crowding levels rise to relatively high levels, while most technology sectors have seen their crowding levels fall back to lower levels. Second, crowding in core segments of the AI industry chain has retreated to moderate or even relatively low levels. In particular, the core North American computing power chain, represented by optical modules, fiber optic cables, and PCBs, has seen its crowding level fall to historical lows.
Furthermore, by using four key metrics—the magnitude of the current correction (performance since July), the degree of being oversold (RSI indicator), crowding (short-term trading crowding), and the strength of earnings revisions (earnings revision situation since July)—we have identified sectors that are significantly oversold, have seen continuous earnings upgrades, and may have been "unduly sold off."
The screening results primarily include: TMT, manufacturing, and cyclical sectors.
Risk Warnings
Potential risks include fluctuations in economic data, policy easing falling short of expectations, the Federal Reserve's interest rate cuts being less aggressive than anticipated, and an escalation in geopolitical tensions.
Comments