The week’s story is easy to state and harder to sit with: the market got paid for believing things that have not happened yet. Stocks rallied to fresh all-time highs — every major index except the Nasdaq — and the move was set off at the very start of the week by a simultaneous drop in oil prices and Treasury yields after the Treasury Secretary suggested a ceasefire with Iran was close. By Friday there was still no deal. The only agreement visibly in the works appears to be one with Oman, and a report late in the week described Iran’s draft plan as banning the United States and Israel from the Strait entirely — the opposite of the “freedom of movement” language that started the rally on Monday. Crude and the curve are trading as if the probability of a resolution is high. That’s a position, not a fact.

The cleanest thing that happened was technical. After roughly three months of chopping inside a 7,250–7,600 band, the S&P broke out to fresh all-time highs on Wednesday. A three-month range resolving upward is an incrementally bullish development on its own terms, and it tends to bring a second wave of buying that has nothing to do with fundamentals: performance chasing by managers who are behind the index, and short covering by anyone who spent the summer positioned for the range to hold. Call volume in the S&P complex stayed elevated all week, which is what it looks like when the fear of missing a move overtakes the fear of the move. The counterweight is just as mechanical — a market that gains this much this fast is stretched near-term, and a breakout that needs chasing to sustain itself is the one most exposed if the catalyst behind it never arrives.

Tech was the engine. The Nasdaq gained roughly 4% and technology as a group added about 6%, the best week for the sector since early May. Software was extraordinary — the IGV put in a five-day move of nearly 9% — and semis had their best week since mid-June. Microsoft added around 7.5% and turned green for the year; Meta gained roughly 6%. Two things pushed it. Last week’s collapse of a heavily levered AI-focused hedge fund now looks like a capitulation event rather than the start of something, the kind of forced selling that clears the deck and then stops. More substantively, the season’s two largest cloud reports delivered actual evidence of AI monetization alongside forecast acceleration, which is the one thing this trade has been missing, and a solid beat-and-raise from Palantir on Monday carried that sentiment across the whole software cohort — ServiceNow closed above its 200-day for the first time since Halloween, Atlassian jumped better than 30%, Snowflake recovered much of its downdraft. Notably this wasn’t indiscriminate: good prints were rewarded hard and bad ones punished just as hard, which is a healthier tape than one where everything rises together.

Rates reversed, and that’s the second pillar. Last week the 30-year was making fresh cycle highs and the 10-year was at 20-month highs; this week yields pulled back across the curve. The long end still had its scare, touching roughly 5.22% during the Chair’s press conference before backing off, but the front end did the real work Friday. July payrolls came in at +23,000, the second-weakest reading of the year, with government down about 53,000, private payrolls up roughly 30,000, manufacturing up only 5,000 and restaurants shedding 26,000. Annual wage growth was the softest in years and the real wage figure, near 3.2%, sits just south of inflation. The two-year fell about ten basis points after having touched 4.26% earlier in the week, and the read was immediate: the case for a September hike got materially harder. Equities took a weak labor report as good news, which tells you what the marginal buyer is actually trading.

Two honest readings of that print, neither settled. The benign one is distortion — a World Cup hiring unwind that will look different in the next two reports, and nobody can adjudicate it until those reports exist. The structural one is that this is a tight labor market rather than a strong one: the labor force has shrunk by something like 1.3 million people over the past year, so businesses can’t find workers and workers can’t find raises at the same time. That second reading is, oddly, the friendlier one for inflation — without wages moving there’s no wage-price spiral available, which is the mechanism by which an oil shock becomes an inflation regime. June was soft too, so this is two consecutive months of less-than-ideal job growth, and the argument that a wait-and-see Fed was the prudent choice is stronger now than it looked a fortnight ago. Jackson Hole later this month becomes the venue for saying something about September, because when a central bank doesn’t fill in the narrative, everyone else does.

Underneath all of it sits earnings, which is why nothing has actually broken. With 442 S&P 500 companies reported, 69% have beaten on the top line and 87% on the bottom line, with headline EPS growth tracking at 51.44% and revenue growth at 14.70%. That headline carries a large asterisk — strip out the one-time investment gains inside a couple of mega-cap results and underlying growth lands somewhere in the 26–29% range depending on whose adjustments you use. Hold both framings. A number that shrinks by nearly half under inspection is a thin foundation for a breakout, and it may explain the sour reactions to genuinely good prints all season: the tape appears to be pricing the adjusted figure, not the advertised one. Equally, 26–29% is an outstanding quarter by any historical standard, forward quarters are tracking high-20s and mid-20s rather than falling off a cliff, and while that line holds, the targets now clustered at 8,000 and above don’t get revised down.

The cost side is where the enthusiasm gets expensive, and the plumbing is where it gets violent. Alphabet, Meta and Amazon alone have issued roughly $170 billion of investment-grade paper this year against about $38 billion for all of last year, with the year not over — real competition for dollars, and one plausible reason the long bond can barely get below 5.2% no matter how weak the labor data looks. The market rewarded the fact that no hyperscaler signaled a slowdown in spend; the other side of that is every additional basis point raising the hurdle those projects have to clear while returns are still being demonstrated rather than banked. Meanwhile there are now roughly 540 levered or inverse ETFs against about 75 in 2023, and something like 64% of all options volume is zero-day expiry — the mechanical reason hundred-billion-dollar single-stock moves now happen at a frequency with no precedent. Margin debt is the highest ever recorded, and the reported figure understates it because much of the same leverage travels under other names. A widely followed sentiment gauge closed the week at its most extreme bullish reading since 2021. Against all that, the July momentum unwind was the fastest and deepest on record and was met by an equally fast recovery — either evidence of a self-correcting bull market that clears excess in days instead of months, or evidence the same excess is being rebuilt at a higher price. This reset came through rotation, deleveraging and improving fundamentals rather than deterioration in the macro backdrop. That’s the strongest single argument for why the tape absorbed a weak jobs print and made new highs anyway.

The broadening is real and not confined to tech. Financials are going for a tenth consecutive weekly gain, the longest streak since 1989, taking a sector that was down 10% on the year at one point to up nearly 5%; Goldman is back above $1,000 and an IPO pipeline that looked questionable ten days ago feels open again. Healthcare has quietly been the best sector over the past three months despite a rough July. Housing may be consolidating. The consumer is where the disagreement is genuine, same day and same data: one case says real wages below inflation plus almost no job growth makes the back half hard for anyone selling into the ordinary economy — housing, consumer basics, hospitality, retail — and that the large discount retailers have been sniffing exactly that out; the other says the K-shape is a balance-sheet story rather than an income-statement one, since unemployment is low, people who want jobs have them, and 90-day credit card delinquencies are behaving. Both can be true at once. Next week’s CPI and PPI, plus another round of AI-complex earnings, will arbitrate more than anything said this week.

So the tape hands us a divided picture, not a verdict.

On one side: a three-month range resolved upward, all-time highs across most of the majors, the best week for tech since May, an earnings base still growing high-20s ex-noise, yields and oil lower, a labor print that pushed hike risk further out, and a deleveraging event that now looks like it cleared rather than broke something.

On the other: a rally built on a ceasefire that doesn’t yet exist with headlines pointing the wrong way, an earnings headline that loses half its size under inspection, a long end that won’t come down while $170 billion of new paper competes for the same dollars, sentiment at 2021 extremes, record margin debt with an unmeasured shadow, a microstructure that manufactures violence in both directions, a near-term overbought condition, and an August that is seasonally among the thinnest stretches of the year.

The Iran war remains a live wild card, and the pain trade still looks higher, because performance chasing may still be lurking under the surface. I’m biased toward the tape. I follow what the market does, not what I think it should do — and for now that means the same discipline as always: watch the levels, respect the rotation, and let price lead.