A rotation signal missed. The volume data says that's a different kind of miss than it looks.

Ten trading days ago (Oct 2), one of AI007''s tracked rotation signals fired: the Semiconductor Equipment layer jumped +2.33% in a single session, led by Axcelis Technologies (ACLS) +8.11%. Same day, the Pure-Play Semiconductor ETF layer confirmed it — XSD +2.17% on its own tracked basket, with XSD itself up 3.59% on 1.78x its normal volume. That's the exact setup this signal is built to catch: equipment-layer strength early in the AI chip supply chain, confirmed the same day by real volume elsewhere, often leads downstream spending — fabrication, chip design, servers, data center buildout, power infrastructure — by days to weeks.

The thesis: that strength rotates into those downstream layers within 10 trading days. We track every signal like this to a real, recorded outcome — hit or miss, no cherry-picking which ones get written up.

This one missed

Tracked outcome, averaged (true compounded return, not a simple average) across chip fabrication, AI chip design, networking, AI servers, data center construction, and power/grid layers over the next 10 trading days: −3.63%. Not flat — a real reversal, opposite direction from the thesis.

A miss is a miss. But the more useful question isn't just "did it work" — it's "what kind of move was it." A -3.63% drift that happened because real sellers showed up with conviction is a very different signal than the same number happening on a quiet Tuesday with nobody really trading.

So we checked volume instead of assuming

Across the 172 of 180 tracked companies in those target layers with usable data, comparing the 10-day outcome window to each company's own prior 20-session average volume:

• Average volume ratio: 0.99x — essentially flat against baseline, not elevated
• Only 14 of 172 (~8%) traded at 1.5x or more of their baseline volume
• Only 4 of 172 (~2%) traded at 2x or more
• 109 of 172 (~63%) actually traded below their own baseline volume
• Total volume summed across the group: ~1.454B shares in the window vs. ~1.457B in the baseline period — essentially a wash

That's not the signature of real distribution. It reads more like broad de-risking and reduced conviction than institutions actually selling into the move.

The handful of names that did see real volume pick up cluster in one specific theme, not broadly across the target layers: Vistra (VST, 2.26x baseline), Constellation Energy (CEG, 2.05x), and Applied Digital (APLD, 1.98x) — all power/data-center-for-AI names — alongside a few Korean semiconductor equipment suppliers. The power-for-AI-infrastructure trade saw real participation during the pullback; chip design, networking, and servers broadly didn't.

A second data point pointing the same direction: long yields are pulling back

AI007's own Treasury yield data (sourced from FRED) shows the 10-year at 5.22%, down from a 5.28% prior reading, and the 30-year at 5.60%, down from 5.67%. Both have pulled back off recent highs. Falling long-end yields generally support longer-duration growth names — including AI infrastructure and semiconductors — since they lower the discount rate applied to future earnings.

What's driving it matters, though, and we're not going to guess at a story we can't verify. What we can point to, from AI007's own tracked macro data, is a real and already-confirmed input: nonfarm payrolls fell sharply (133K to 29K) and unemployment ticked up (4.1% to 4.2%) in the most recent reading. A weak labor print is a textbook driver of falling long yields on its own — it lowers the market's growth and inflation risk premium, independent of anything else. Worth noting plainly: this yield pullback predates this week's inflation data entirely, since CPI hadn't printed yet when it started.

What we're actually watching next

CPI releases October 14, PPI releases October 15 — both after a run where CPI YoY had already accelerated (3.54% to 3.71%) and PPI YoY jumped hard (4.80% to 5.41%) in the prior reading, against that same weakening labor backdrop. That's a genuine two-sided setup, not a clean one:

• Cool CPI/PPI prints, yields keep drifting down: consistent with the low-volume, non-distribution pullback resolving as a normal pause, not a reversal.
• Hot CPI/PPI prints despite the weak labor data: the stagflation-ish scenario, and the one that could turn a quiet pullback into something with real volume behind it, particularly for rate-sensitive, long-duration names like semiconductors and AI infrastructure.

We'll track the actual prints and what the market does with them, the same way we tracked this signal — reported straight, whichever way it goes.