Weekly Signal · · 8 min read
Three Rotations, One Week
One signal to define the week and shape the month. Three minutes, six questions. Every Sunday at 21:00 CET — scored the next.
19 July 2026 · closes through Jul 17.
0 · Last week, scored
Verdict: confirmed — the bear line fired. Last Sunday's signal said the market had priced the good outcome before seeing a single print, with SOXX the referee between 554 and 590. The prints came in good to record-setting — ASML's capacity frame, TSMC's record quarter, Netflix's growth — and the tape sold every one of them. SOXX broke the 554 floor and never saw it again: Friday ran the full round trip, 498.54 at the open of the washout, a kiss of 530.52 — the prior close, to the tick — and a fade to 521.81, 43 cents below the June washout intraday (522.24). SMH closed under its 566.83 triple-bottom line. QQQ, remarkably, held its own June washout by 39 cents (686.76 low vs 686.37). One asterisk over all of it: Friday was a third-Friday options expiry, which amplified both the violence and the reclaim. Monday — with Seoul reopening and a clean tape — is the confirmation session. The running ledger lives at the track record.
1 · The signal
Distribution is confirmed — so the question changes. It is no longer whether the AI trade corrects; it is where the money goes while it does. Three rotations are on the table, and this week names one of them. The name it picks shapes how the AI trade performs into year-end.
Rotation one — the handoff. Out of AI capex, into AI opex and beneficiaries: the build-out complex (Rubin) de-rates while the operating layer (Agentic Infrastructure) and the users (Agentic Winners) hold or advance. The AI trade doesn't end; it migrates downstream.
Rotation two — the internal. Inside the build-out itself: design de-rates while verification re-rates, tools give way to consumables, the memory complex washes out. Rubin remains the trade; its leadership changes.
Rotation three — the exit. Out of the AI trade altogether, toward what AI can't touch: every AI leg falls together and HALO is the best index of the four. Not a migration — a de-rating.
2 · What it is
Last week didn't just confirm distribution — it produced the first clean data for the rotation question. The cascade across our four indices, equal-weight, on the week: Rubin Build-Out −9.3% · Agentic Infrastructure −3.3% · Agentic Winners +0.3% · HALO −2.0%. The further from the capex checkbook, the better the week. And the market-side mirror says the same thing: SOXX −10.2% while software (IGV) finished green at +0.4% and the equal-weight S&P fell just −0.4%. This was a semis-specific unwind with breadth intact underneath — the signature of rotation, not (yet) of exit.
To measure exactly this, systematically, we put a new instrument live today: the AI Handoff Board — nine ratios of our own sub-indices tracking the handoff from building AI to operating it to using it, staged in the order a regime change would propagate, every ratio incepted at 1.00 on June 30. The first readings, closes through Friday:
| Ratio | Question it answers | Level | 1w |
|---|---|---|---|
| Use / Build (scoreboard) | From building AI to living off it? | 1.296 | +10.6% |
| Operate / Build | Building AI, or running AI? | 1.155 | +6.7% |
| Trust / Execution | Is trust the new bottleneck? | 1.135 | +11.0% |
| Verification / Design | Who checks the machines' work? | 0.938 | +9.1% |
Read those honestly: three weeks of history, and the big cross-index moves largely inhaled the semis correction — a falling denominator, not yet a rising numerator. But two readings carry real information. The Trust/Execution ratio rose +11% in a down week because the trust cohort was actually green — identity +10.9%, security +4.0% while the rest of the operating layer fell. And Verification/Design moved +9.1% the week an AI lab claimed machines can design silicon — EDA de-rated 15% while testing & metrology fell half as much. The two scarcity trades — who verifies the work, who is trusted to run it — outperformed everything around them. That is the K3 thesis showing up in relative price within days of the event.
3 · Why it matters
Because the three rotations produce three different year-ends, and portfolio construction for the back half depends on which one this is. If it's the handoff, the AI trade survives as a migration — the index-level story wobbles (the capex names carry the index) while the opex and beneficiary layers take leadership, and the right posture is rotation within the theme, not away from it. If it's the internal, the build-out itself remains the trade and the work is sector selection inside it — verification, consumables, the things that scale with utilization rather than expansion. If it's the exit, 2026's dominant theme de-rates into year-end, index earnings estimates built on the capex cycle come under review, and HALO — the part of our universe built to be insulated from exactly this — earns its keep. The tape has to choose; a distribution phase doesn't resolve into nothing.
The timing sharpens the stakes: Rubin-generation systems ramp in the second half of 2026 — the hardware that is supposed to move agentic AI from technically possible to economically rational. If that transition is real, the handoff is the rotation that should win over quarters, whatever this week does. This week tells us whether the market is starting to price it now, under pressure, or still treats every AI dollar as one trade.
4 · What to look for
| When | What | If | Then |
|---|---|---|---|
| Mon | The confirmation session — Seoul reopens (SK hynix traded +7.6% in the US while Korea was shut), first clean tape after OpEx | SOXX reclaims 530–532 | Friday's fade was mechanical; the distribution verdict is overturned and the rotation question is postponed |
| Mon–Fri | The cascade — Rubin vs Agentic Infrastructure vs Winners vs HALO, daily | the ordering of last week repeats | handoff; if all four fall with HALO best — exit |
| Wed | Tesla prints after the close — earnings season week two, on a tape that just sold record numbers | another good print sold | the reaction function from last week persists; the bar is the ceiling |
| Fri | SOXX weekly close + the Handoff Board week-over-week | see the lines below | the branch gets named — and scored here next Sunday |
The lines this signal is scored against: SOXX above 530–532 on Friday's close overturns distribution; below 522.24 extends it. With SOXX below the zone: Operate/Build and Use/Build rising on the week = the handoff; HALO the best index of the four = the exit; Rubin stabilizing while verification and consumables lead inside it = the internal. They are not mutually exclusive — but the week will rank them.
5 · Bull case / bear case
| Bull | Bear |
|---|---|
| Monday confirms the OpEx artifact: Seoul reopens with a gap up, SOXX reclaims 530–532 and holds it into Friday. The washout goes in the books as the low of the correction, QQQ's 39-cent hold marks the degree split that held, and the rotation question resolves the friendly way — leadership broadens within a rising tape. | Monday opens heavy and the 522 zone becomes resistance. The cascade repeats with everything down — including the opex and beneficiary layers — and HALO finishes the week best of the four. That is the exit rotation: not a handoff but a de-rating, arriving mid-season with the heaviest AI prints still ahead. |
6 · How it may define the next weeks
Whichever branch the week names becomes the frame for August: the handoff regime says buy the migration — the operating and beneficiary layers on weakness, the build-out only at its structural floors; the internal regime says the build-out remains the core position but its sector mix does the work; the exit regime says reduce the theme and let the insulated side of the barbell carry into the seasonal window. We will not have to guess: the cascade, the board ratios and the SOXX close are all measurable, and next Sunday's edition opens by scoring them. Probability, not prophecy — the referee wears a ticker.
The week, for the record
| Closelook indices (EW, wk) | Markets (wk) | ||
|---|---|---|---|
| Rubin Build-Out | −9.3% | S&P 500 (SPY) | −1.5% |
| Agentic Infrastructure | −3.3% | Nasdaq-100 (QQQ) | −4.2% |
| Agentic Winners | +0.3% | Equal-weight (RSP) | −0.4% |
| HALO | −2.0% | Semis (SOXX / SMH) | −10.2% / −8.9% |
| Rubin storage / EDA | −26.4% / −15.0% | Software (IGV) | +0.4% |
| Agentic identity / security | +10.9% / +4.0% | Volatility (VIXY) | +6.4% |
Source: closelook.net Functional Indices and data lake; index and sub-index figures computed from the live index workers (equal-weight, Jul 10 → Jul 17); ETF figures verified against vendor EOD closes through Jul 17. Tesla's print date per our earnings calendar (Wed Jul 22, after close). Friday Jul 17 was a third-Friday options expiry — flows that day were partly mechanical. This is an investment diary, not investment advice — Look Investment GmbH is not a licensed adviser.
The signal is free, and stays free. What we do with it in our own books — the branch playbook — continues for members. Join the Look — it’s free.