The Morning 10

What

SOXX closed Tuesday at 552.69, up 5.45% — it gapped through the 530–532 verdict band, printed a session low of 540 and never looked back. The pattern of three sold reclaims in four sessions is over: this was the fourth attempt, and it arrived with breadth — the equal-weight semiconductor ETF rose 5.44% and SMH 4.52%, so the move was the complex, not two megacaps.

If

Today's half-percent pre-market give-back holds anywhere above the old band, the break reads as normal digestion into tonight's prints. A full round-trip back inside 530–532 at the close would be the one print that converts breakout into trap.

Why

A band that rejects four recovery attempts and then breaks on a five-percent thrust with equal-weight participation is the signature of a seller who is finished. What it cannot survive is the catalyst behind it failing — and that catalyst reports tonight.

Then

The technical half of the two-part trigger is in ink at 552.69. Nothing today's tape does can un-ink it — only Alphabet's capex line can.

What

SPY closed 748.28, up 0.83% — back above its 50-day at 744.55 after an intraday kiss to 744.19 that held to the tick. QQQ rose 1.85% to 708.97. For the first time since the drawdown began, all three scoring lines — SOXX versus the band, SPY versus the 50-day, QQQ versus its floor — closed bullish on the same evening. The caveat: the equal-weight S&P added only 0.16%, so the advance was narrow.

If

This morning's dip holds above 744.55 on SPY, the repair survives its first pullback test with all three reads intact. A close back below the 50-day would re-open the disagreement the last two weeks traded on.

Why

Weeks of contradiction — epicenter repairing while the broad index lost its trend line — resolved in one session. Aligned reads are rarer and more informative than any single level, but a narrow advance means the alignment leans hard on tonight's confirmation.

Then

Score the same three closes again tonight, after the prints. That is the technical day in three numbers.

What

On a day the semi complex rose five percent, software fell: IGV down 1.25%, cloud down 1.69%. Inside the complex the same split — Micron up 12.17% to 970.82 while Synopsys managed 2.80%: green, but nearly ten points of relative lag, the third consecutive physical-over-design session. The sector board told the same story: Technology +2.89% led, while Staples (−0.94%) and Communications (−0.69%) were sold.

If

Software fails to confirm even through a strong capex answer tonight, the rotation stops being tactical and starts being structural — money leaving one layer of the AI trade to fund another, not de-risking.

Why

This is the exact split the K3 trilogy mapped: capex buys physical things — fabs, memory, power — while AI-assisted design compresses the value of tools and software economics get questioned. Three sessions of the same tape is positioning, not noise.

Then

Tonight's guide is the arbiter: a capex acceleration feeds the physical layer directly. Watch whether software even bounces on a good number — that is the tell for whether the divergence has further to run.

What

The Nasdaq listing of SK Hynix opened higher for the fourth time in the drawdown — and this time nothing was sold: it closed up 13.75% at 171.94, from a 149 shelf that was defended twice last week. Micron printed the same behavior at +12.17%. Advantest rose 8.5% in Tokyo, TSMC 5.55% here. The epicenter names didn't just bounce — they held their gaps, the behavior change every previous rally lacked.

If

The give-back morning gets bought in these names — dips absorbed rather than gaps sold — the regime change is confirmed at the single-stock level, where it matters most.

Why

Three gaps sold into a defended floor was a stalemate: support without demand. One sustained gap doesn't build a trend, but it is the first evidence of demand above the floor since the drawdown began — and it arrived the session before the fundamental verdict.

Then

Memory is the purest play on hyperscaler buildout budgets. Tonight's capex guide either funds this move or strands it at 171.94.

What

At Tuesday's close the Build layer massively outran the Use layer: Rubin Build-Out (equal weight) rose 5.2% on the day while the Agentic Winners cohort slipped 1.1% — the Use/Build ratio fell roughly six percent in one session, its sharpest one-day reversal since the June 30 base, after three weeks of one-way movement that had carried it up nearly thirty percent.

If

The ratio keeps falling through a confirmed capex night, the last three weeks read as defensive positioning for the drawdown — now unwinding. If it stabilizes and resumes rising while semis stay green, the handoff is a durable leadership change that survives a Build rally.

Why

This is precisely the test the trend needed. A ratio that only rises when the market falls is a hedge; a ratio that rises through both tapes is a regime. One session decides nothing — but this week decides which one it is.

Then

Two closes to watch through Friday: Use/Build versus this week's low, and whether tonight's capex answer feeds the denominator again.

What

TLT slipped again to 83.66 — down almost three points since the end of June (86.42), a fresh July low. IEF closed at 93.31, sitting directly on its major support at 93; the next shelf below is 91. All of this while the Fed path trades as a hold — plausibly no move in either direction this year. The dollar firmed 0.32%.

If

IEF loses 93 on a close, the long-end selloff graduates from drift to break — with the next support a full two points lower, and every rate-sensitive corner of the equity tape repricing against it.

Why

If the Fed sits still, the long end is doing the tightening on its own. An equity market re-rating on an AI-capex acceleration while duration sells off is a pairing that eventually collides — buildout capex is financed at exactly the maturities that are cheapening.

Then

Watch 93 on IEF through tonight's prints. A capex-acceleration answer that also breaks the bond floor is only half a green light — the multiple pays for the second half.

What

The broad tape's fuel on Tuesday came from outside tech: 3M rose 7.32% and GM 4.91% on morning beats, and after the close Capital One printed $5.81 against a $4.67 consensus — a 24% earnings beat on in-line revenue of $15.85 billion, having gone into the number flat. Intel, which reports Thursday, rose 8.64% in sympathy with the complex.

If

Capital One's beat gets paid this morning and Financials — trading near their 52-week high — stay green, the market has earnings breadth underneath it while tech argues with itself.

Why

A tape where industrials, autos and card lenders beat while the AI complex debates its capex cycle is a market with a floor: the drawdown was a sector repricing, not an earnings recession. That floor is what makes tonight's binary survivable either way.

Then

The non-tech calendar keeps printing: Intel Thursday after the close, American Express Friday. Breadth of beats is the quiet confirmation to track.

What

US futures are down roughly half a percent pre-market after Asia followed New York higher overnight. The rest of the board is calm: gold rose 1.96% Tuesday to 374.81 and is firmer again this morning, bitcoin is flat after Tuesday's 2.1% gain, VIXY fell 2.68% to 20.66, and Money Temperature reads 51 — risk-on, not stressed.

If

The dip gets bought before midday, the regime change is confirmed in mirror image: for three weeks every gap up was sold — a give-back that gets absorbed is the same behavior change as yesterday's held gaps, from the other side.

Why

Breakout days are graded by what happens next. A half-percent digestion above a broken band is textbook; what would not be textbook is vol bidding and the old band trading again. The cross-asset board — gold up, crypto flat, vol falling — says hedging demand is coexisting with risk appetite, not replacing it.

Then

A red close is fine; a red close above 530–532 into the prints is even textbook. The only failing grade today is a close back inside the band.

What

All four report after tonight's close. Alphabet: consensus near $2.87 EPS on roughly $116.5 billion revenue — but the capex guide is the number the entire complex trades on (call 22:30 CET); the stock went in down 1.38% at 347.15. Tesla: consensus near $0.50 on about $26.1 billion. IBM: the first guide since its −25% pre-announcement session — the floor question. ServiceNow, the closest thing to an agentic-AI bellwether: backlog, RPO and AI contract value — it went in down 2.52%.

If

Alphabet's capex number and trajectory land at or above consensus, the Build layer gets its demand confirmation from the largest spender one session after the chart broke out — technicals and fundamentals aligned for the first time in a month. A guide-down strands a twenty-point breakout.

Why

The whole three-week debate — decelerating AI spend versus migrating AI spend — gets four primary-source answers in one evening: hyperscaler capex, enterprise-agent demand, an enterprise-AI floor, and Tesla's spend. Every level in points 1 through 5 is downstream of the first one.

Then

Read the four in order of weight: Alphabet's capex guide, ServiceNow's AI contract value, IBM's floor, Tesla. The first number sets the open on Thursday; the other three set the breadth of it.

What

Gene Munster's current framing holds that AI infrastructure spending is entering a second, deeper capex wave driven by inference demand — and that this wave benefits semiconductor and systems names before software. Tuesday's tape traded exactly like that sequence: the physical layer up five percent, software down, design tools lagging memory by ten points.

If

N/A

Why

His thesis and our Handoff Board are pulling in opposite directions this week — the board's three-week Use-over-Build trend just mean-reverted hard in favor of his sequencing. Both cannot keep winning, and both get graded by the same number tonight: a capex guide at or above consensus funds his second wave; a guide-down feeds the handoff read instead.

Then

Read his latest Deepwater notes, then set his infrastructure-before-software sequence against tonight's actual capex and AI-contract numbers — not against anyone's summary of them.

Gene Munster · Deepwater Asset Management ↗