Written by: Rita
Tide Guide
The Philadelphia Semiconductor Index (SOX) has fallen approximately 19% since hitting its all-time high on June 22, nearing the 20% threshold for a bear market. AI stocks are still declining, and JPMorgan advises: Don’t rush to buy the dip.
This assessment comes from J.P. Morgan’s quantitative team’s “AI Bubble Sentiment Score,” which tracks how intensely AI bubbles are discussed in global media. The score peaked on June 29 and is now declining. Historically, such narrative cycles take about 1.5 months to shift from hot to cold, with the lowest point expected as early as mid-August.
The issue is that the absolute score remains at 687, within the top five percent of historical levels. Even if cooling begins now, as long as it stays in this high range, there is still a 53.8% probability that the SOX will experience a pullback of more than 8% over the next 20 days.
JPMorgan's quantitative model is clear: the AI alert has not been lifted. Buying AI semiconductors before mid-August carries high risk.
How the alarm works
JPMorgan tracks narratives, not stock prices. This score counts the number of negative media reports globally about the AI bubble—the more reports, the higher the market anxiety, and the higher the score.
Historically, the cycle of this anxiety subsiding has been regular: averaging 43 trading days from peak to trough, or about 1.5 months. This peak occurred on June 29; according to historical patterns, the bottom is unlikely to be reached before approximately August 12.
But this score depends on two factors: direction and level.
The direction has moved downward, which is positive, but the level is still at 687, placing it in the top quintile (Q5) of historical readings. J.P. Morgan’s backtesting of all historical samples shows that when the score is in the Q5 range, the average 20-day return for the SOX is -0.7%, with a 53.8% probability of drawdown. Only when the score falls into the low-to-mid range does the average return turn positive (+5.1%) and the drawdown probability drop to 25.4%.
In other words: the alert is shifting, but has not been lifted.
Three specific thresholds provided by JPMorgan
JPMorgan translated "all clear" into three specific numbers:
- Score drops below 629: SOX drawdown probability decreases to 30%
- Below 560: Drawdown probability reduced to 25%
- Below 418: Drawdown probability reduced to 20%
Current score: 687. A drop of 8.5% is needed to reach 629, 18.6% to reach 560, and 39.2% to reach 418.
JPMorgan recommends entering in two phases: first near 629, and second near 560. Until then, any rallies may be pulled back by narrative-driven panic.
Current stage: Positions in Japanese AI stocks are nearly fully liquidated, but panic in U.S. markets hasn't fully subsided.
The situation with Japanese stocks is different from that of U.S. stocks.
Position adjustments for Japanese AI momentum stocks are nearing completion. While the Nikkei 225 has been weighed down by semiconductors, TOPIX futures have remained resilient, as capital shifts from semiconductors to banks, financials, and materials. CTA (Commodity Trading Advisors) continue to hold long positions in TOPIX futures.
The situation in U.S. equities is closer to the "narrative" stage. Since hitting an all-time high on June 22, the SOX has declined approximately 19%, nearing the 20% threshold for a bear market. However, the score remains at 687, indicating that market anxiety is primarily driven by the narrative that "the AI bubble is about to burst," while positioning issues are secondary.
JPMorgan Chase's historical data highlights a key distinction: when scores begin to decline but remain high, SOX still faces significant retracement risk. The true condition for a rebound is not "AI is good again"—it’s when the topic of the "AI bubble" has completely cooled down, signaling the optimal time to buy the dip.
Tide View
The most valuable information in this report is turning the vague market sentiment of an "AI bubble" into three actionable numbers: 629, 560, and 418.
For U.S. stock investors, the quantitative alert has not yet been lifted—this is not the time to add positions. Use this framework to monitor; only consider entering positions in batches once the score drops below 629. Until then, hedge with puts or defensive sectors.
For A-share investors, sentiment toward the A-share AI sector is more fragile, and domestic AI sectors (such as optical modules and servers) exhibit higher volatility. J.P. Morgan’s “two-stage entry” strategy can be referenced, but it must be adjusted to align with domestic policies and the logic of China’s domestic computing power. The August earnings season may be a critical window for China’s AI ecosystem to reclaim its leadership.
The long-term logic of AI has not changed. However, the short-term quantitative alerts have not yet been lifted.
Disclaimer
This article is a summary and interpretation by Chaoxiang Research of a third-party brokerage research report (J.P. Morgan, July 16, 2026). The ratings, price targets, earnings forecasts, and related judgments cited herein reflect the views of the brokerage’s analysts and represent the position of their respective institution, not the views of Chaoxiang Research, nor do they constitute any investment advice.
The market carries risks; make decisions independently. This article should not be used as a basis for buying or selling any securities.