Weekend Decode: AI Was the Story. It Was Also the Worst Sector.

Chapter 01

The week built around AI, and AI finished last

In the week the calendar branded AI’s, Technology had the worst median 5-day return of all 11 sectors at −8.6%. The leaders were Real Estate, Healthcare and Utilities.

This was supposed to be AI’s week. Broadcom reported, the custom-silicon debate was front and center, and every desk note was about the profit pool three years out. Then the five days happened. Of the eleven broad sectors, Technology finished dead last with a median five-day return of minus eight point six percent. Not mid-pack. Last. The three sectors that led were the ones AI was supposed to make quaint: Real Estate up three point two, Healthcare up three, Utilities up two point six. Six of the eleven sectors finished green, and every one of them was a place you go for dividends or defense, not growth. The widest gap of the week was eleven point eight points, from Real Estate at the top to Technology at the bottom. The marginal dollar didn’t leave the market this week. It rotated — out of the long-duration AI trade and into the corners that pay cash today. Everything else in this Weekend Decode is a different angle on that one fact.

"Six of the eleven sectors finished green, and every one of them was a place you go for dividends or defense, not growth."

Chapter 03

The rotation map: same quality scores, opposite week

Plot every name by 5-day return and composite quality score and the week splits cleanly: high-quality AI names sold (left), high-quality defensives bought (right). Apple was the only AI name that didn’t get sold.

Here is the cleanest way to see what happened. Take the names that score highest on our composite quality engine — the engine doesn’t care whether a name is AI or a soap company, it just ranks fundamentals and trend — and plot each one by how it did this week. The picture splits in two. On the left, negative returns, sit the high-composite AI names: Microsoft at a sixty-seven composite but down nine point six, Nvidia sixty-four and down eight point three, Meta sixty-six and down one point seven, Broadcom fifty-one and down fifteen point three. Same quality, sold anyway. On the right, positive returns, sit the high-composite defensives: Abbott seventy and up three point eight, Philip Morris sixty-eight and up three point three, Visa sixty-seven and roughly flat, Wells Fargo and JPMorgan both up five-plus. The quality scores barely differ. The direction is opposite. The one name that broke the pattern is Apple — composite sixty-eight, up zero point seven on the week, the only Magnificent Seven name to finish green. It is also the one with a catalyst Monday.

Chapter 05

How far it ran: the complex snapped below trend

Most of the AI complex closed the week below its 20-day line — Broadcom −9%, Vertiv −11%. The two names still stretched well above it, Marvell (+31%) and Arm (+17%), are the unfinished business into Monday.

The last lens is technical: how stretched is each name versus its own twenty-day trend line, the level that tells you whether a stock is extended or washed out. A week ago much of this complex was riding well above trend. By Friday most of it had snapped below: Broadcom closed nine point two percent under its twenty-day line, Vertiv ten point five under, Nvidia six point two under. That is what a fast de-rate looks like — the extension came out in five sessions. But two names did not reset. Marvell is still thirty-one percent above its twenty-day line even after Friday’s thirteen-percent drop, and Arm is still seventeen percent above its own. Those are the two with the most room left to fall if the rotation continues — the overhang the rest of the complex already worked off. The reset is mostly done; the cleanup is not.