The Buy Lists Arrive — One Week After K3, the Street Reaches Past Semis
A thesis read for the day the street showed its hand — companion to this morning's Morning 10. The three-part Kimi-K3 series this builds on: what happened · the chip chain · the Rubin map.
The buy lists arrive
It took one week. Kimi K3 landed on the 16th, the semiconductor complex had its worst week in over a year, and last night the first institutional buy-the-pullback notes hit client inboxes. UBS’s trading desk — Michael Romano, who runs hedge-fund equity derivative sales — argues the momentum unwind may be nearing its end: hedge funds have cut long positions in momentum and semiconductor names by roughly five percent of gross market value, a reduction the desk describes as among the largest on record, leaving net positioning in semis and software back at April levels.
The desk’s momentum gauge swung from −3.5% to +2.5% inside two hours on Friday — the kind of intraday reversal that historically marks late-stage unwinds — and the note expects the process to bottom by the end of July. The advice is not a table-pound but a posture: scaling in is prudent. Separately, CNBC reports a UBS screen — run on the bank’s HOLT framework — for technology names sold off since late June whose fundamentals kept improving through the drawdown.
Read the list, not the headline
The headline says semiconductor pullback. The basket says something more precise — and more interesting. UBS’s momentum names: Sandisk, Broadcom, Oracle, KKR, Datadog, Microsoft. Count the pure semiconductor plays: arguably one and a half. The rest of the list is storage and memory economics, custom silicon, AI operating capacity, observability software and the capital layer that finances the buildout. A week after a chip-design shock, the street’s first buy list is mostly not chips.
That composition is the quiet confirmation of the map Part II drew and Part III formalized: K3 splits the complex into a disrupted layer (design tooling — absent from UBS’s list, as it should be), a defended physical layer (memory, storage, custom silicon — present), and a demand side that gets cheaper inputs (AI opex and applications — present twice). Our own tape has been voting the same way for two sessions: Micron up while Synopsys sits within a few percent of its 52-week low, the Handoff Board’s Use-over-Build ratio up 29.6% in three weeks. When an independent screen built from momentum and fundamentals lands on the same asymmetry a structural argument predicted, that is not proof — but it is exactly what being early is supposed to look like.
The week we score it
So we make it operational. Through this week we run the ratios daily — Design/Physical from the Rubin map, Operate/Build and Use/Build from the Handoff Board — as a running tally, because five sessions of confirmation or failure will tell us more about the second half than any single print. What the ratios cannot supply is the demand line, and that arrives on schedule: Alphabet reports after tonight’s close, the first hyperscaler capex indication since K3 existed — with the rest of the large spenders following over the next two weeks.
A reaffirmed or raised spending line makes the rotation read the base case into year-end: design deflation feeding the physical bottlenecks while the savings migrate up the stack. A guide-down would mark the street’s buy lists as early and the week-one panic as information. Either way, by Friday the tally will have five sessions and one capex print in it — and the year-end map gets drawn from evidence, not from positioning.
The panic phase priced the curve. The buy-list phase prices the map. The difference between the two is where the second half gets made.
C · members block
Into tomorrow
The book moved this morning, before the street’s lists were news. As set out in today’s Morning 10 point 11: tactical starters in the K3 winners — Advantest as the template, Samsung and SK Hynix on the Korea rebound, TSMC after its record print — sized so that another leg down is an add, not a stop. Note the difference to UBS’s basket: they buy momentum broadly, including the opex layer; the diary buys the drawn-down physical layer the K3 map says wins either way. The structural trigger is unchanged: Alphabet capex at or above consensus and a SOXX close above 530–532 turns the starters into a full add at Wednesday’s open — SKHY 149 and MU 804 stay the invalidation floors.
Into tonight we watch: whether SOXX holds this morning’s gap into the 530–532 band at the close — the fourth attempt, first by air; the Design/Physical ratio print — a third session of physical-over-design would make the rotation the confirmed read before Alphabet even speaks; and after the bell, one number before everything else: the capex line against consensus, then the cloud growth rate it is meant to feed.