AI chip stocks have cooled fast. The SOXX fund, which tracks the semiconductor index, sits about 15.7% below its June high, and after a long run of dip-buying, retail traders have started selling.
That flip is the rare signal Scott Rubner, Head of Equity Derivatives Strategy at Citadel Securities, just flagged. One that has marked past selloff lows, or rather, local bottoms.
What Rubner Flagged
In a July investor note, Rubner said retail clients turned net sellers of chips on two down days, July 2 and July 7, as the Philadelphia Semiconductor Index (SOX), the benchmark for major chip makers, fell about 5%. Selling into a falling SOX is rare.
Note: We chart SOXX, the exchange-traded fund that tracks the SOX index, because the index itself cannot be traded.
Moreover, he counted only about eight such episodes over the past year. Nearly all arrived late in a selloff, just before chips bounced. That’s the AI chip bottom thesis this piece chases.
Citadel sees this through payment for order flow, the arrangement that lets it handle retail trades and read their positioning. That data is not easily accessible.
Why We Rebuilt the AI Chip Signal
Because that order flow is private, we rebuilt the signal from public data. Our proprietary Retail Capitulation Radar (RCR) tracks two leveraged chip funds, SOXL and SOXS, which aim to move two or three times the semiconductor index each day.
Retail traders dominate them. The RCR is our own bottom signal detector.
When retail dumps the bullish fund or crowds into the bearish one as chips drop, the behavior shows up in that trading. On the test, the strict signal fired twice, both in early March 2026.
The chart shows why that matters. SOXX has dropped about 16% from its June high, yet it still trades roughly 80% above that March base, where the signal last fired.
Here is the honest part. Citadel counted eight episodes, yet the public proxy (our metric) confirmed only two, and it did not reproduce the exact July signal on the chart. That gap cuts both ways. Either our proxy runs too tightly, or public data missed what Citadel’s private order book saw.
Another Historical Pattern Shows Similarity
Still, both datasets point the same way. In Rubner’s retail-selling episodes since February, chips rose over the next five to ten days every time, with a median gain near 18% over ten days, and the March case rose about 29%.
The proprietary radar above is deliberately strict, which is why it fired only twice. So we also ran a second, loser test that flags any two-day drop with broad chip weakness. That wider net catches more cases, ten in all, and it broadly agrees, with a median gain near 7% over the next ten days.
However, this test is noisier. One late-February episode kept sliding for three weeks before recovering, so the rebound is a direction, not an immediate rule.
What the AI Chip Signal Says Now
Timing matters here. Citadel flagged the move in early July, and chips have rallied since, so the setup is aging rather than fresh.
For now, the radar reads idle. It fires only when heavy retail selling meets a falling market. Today the selling pressure is elevated but still short of that mark, and the latest session jumped 5.45% (from the Tradingview chart), an up day the tool ignores.
Yet the pressure on the AI chip stocks has not cleared. Nvidia and AMD absorbed the selling best, holding buying support while their prices slipped, unlike most peers, so they would likely lead any turn back up.
The next trigger is close. Intel reports earnings on July 23, and options traders are leaning bearish into it. Puts outnumber calls on both volume and open positions, and the market braces for a 5.2% swing around the report.
So the story is not over. A weak Intel print could send AI chip stocks lower again. That would re-arm the bottom signal that sits idle today. That is why the options crowd is paying for protection rather than trusting the bounce.