Q3 positioning — the two markets problem

RE: Q3 positioning — the two markets problem
DISTRIBUTION: Investment committee only. Do not forward.

Let's stop pretending we're looking at one market, because we aren't. We're looking at two, wearing the same ticker tape.

Market One closed the holiday week at a record high. The Dow put up 595 points on Thursday alone, its second all-time high in three sessions, carried by Apple (+4.8%), McDonald's (+4.1%), Walt Disney (+3.8%), Visa and Walmart. This is the market of cash flows, dividends, and companies that sell hamburgers and payment rails to people who still have jobs, mostly. It does not care what a GPU costs. It has never heard of a token.

Market Two is somewhere else entirely. Micron dropped more than 10% on Wednesday and another 7% Thursday — a stock still up over 260% year-to-date, which tells you everything about how far it had to fall before anyone noticed the air thinning. Applied Materials slid 7.4%. AMD gave back 4.3%. Nvidia, Broadcom, Intel, KLA, Lam Research all traded lower on the same tape that sent the Dow to a new high. Tesla fell 7.5% on a strong delivery report, which is the kind of dissonance that should worry you more than a bad number would.

Here is the memo's actual point, stated plainly: the June jobs report did this, and almost nobody is pricing it correctly.

Nonfarm payrolls came in at 57,000. Consensus was north of 100,000. ADP's private read for June landed at 98,000, itself a miss, on top of a downward revision to May. Unemployment sits at 4.2%, inching up, participation softening at the edges. This is not a collapse. It is a fed being handed the first genuinely inconvenient data point of Kevin Warsh's tenure — a chairman who spent his first meeting stripping the statement down to nine words and daring markets to fill in the rest, who told a room in Sintra that "prices are too high" while declining to acknowledge that jobs might be too few.

The equity market's initial reaction to the payrolls miss was textbook: soft labor data reads as dovish, dovish reads as bullish, buy the Dow. Fine. Except the same data point that eases pressure on the front end of the curve should terrify anyone still underwriting the AI capex thesis on the assumption of an economy running hot enough to justify $400 billion a year in data center spending. You cannot have it both ways. Either the consumer is fine and the Fed stays tight, pressuring the multiple on anything priced for perfection, or the labor market is actually rolling over, in which case the hyperscalers spending like it's 2021 are about to discover that enterprise software budgets do not expand when payrolls shrink.

The tape spent Monday and Tuesday of this holiday week trying to convince itself of a tech relief rally. It did not survive Wednesday. Caterpillar, of all names, pulled back nearly 7% intraday on the same session the Dow hit its high — a reminder that even the "safe" index has a landmine or two wired into it. And underneath the semiconductor rout sits a detail that deserves more attention than it got: reports that OpenAI is in talks to sell the US government a 5% stake, alongside Meta's disclosure that it may start monetizing excess compute capacity it built for itself. Read those two sentences together. One of the best-funded private AI labs in the world may need a government check. One of the best-funded public ones has more chips than it currently knows what to do with. Neither of those is what a supercycle sounds like at its peak. Both are what one sounds like six months past it, when the people closest to the spending finally say the quiet part.

None of this shows up in the Dow. It cannot. The index is price-weighted and stuffed with companies that don't touch a wafer. It will keep printing highs on Apple and Visa long after the AI trade has finished digesting its own excess, and every headline will say "stocks hit record" while a fifth of the market that actually matters for growth expectations quietly bleeds out.

Positioning note: treat Dow strength as a distraction, not a signal. The number that matters this quarter is the ISM services print investors are waiting for post-holiday, and whether it corroborates or contradicts the payrolls miss. If services confirm the slowdown, the Fed's "curt" statement gets tested fast, rate-hike dots from June start looking like a chairman's overreach rather than a forecast, and the semiconductor names that got hit this week become the cheap side of the trade rather than the falling one. If services data comes in hot, Warsh has his excuse, the hawkish path holds, and the AI-adjacent names have further to fall before anyone starts calling it a buying opportunity out loud.

Either way, stop reading the Dow as a market signal. It's a museum exhibit that happens to still trade.

— filed ahead of Monday's reopen