Week In Review: America Found 44,000 Jobs It Never Lost

Financials closed the week at 58.10. Financials started the week at 58.10. Not approximately. Exactly. An entire industry of very expensive people spent five days producing a number identical to the one they started with, and somewhere a strategist is writing a note explaining why that was the correct outcome.

Had $1,000 riding the S&P on Monday morning? You have $1,001.09 now. That is a dollar and nine cents for a week that included a shooting war, a jobs shock, and a rate-hike scare. Buy yourself most of a gumball.

Last week everybody saw the tide chart and built anyway. This week everybody searched a pond for a man who was sitting on his porch.

The week moved everything except the number everybody watches.

A month ago, the July jobs report said America lost 23,000 jobs. We wrote a whole edition about it. Wall Street rallied on it because a weak job market means the Fed stops raising rates, and a market that has learned to root against its own neighbors has a personality problem.

On Friday, the government came back and said: our mistake. ” July was not minus 23,000. July was plus 21,000. That is a 44,000-job swing in a single revision. June got moved up too, from 20,000 to 31,000. Add them together, and 55,000 jobs that everyone had written off turned out to be sitting on the porch, drinking lemonade, watching the search party drag the pond.

Then August printed 162,000 new jobs. The forecast was in the mid-50-thousands. Unemployment stayed at 4.1%. Average hourly pay rose 10 cents to $37.75 an hour, up 3.1% over the year, which is real money in a real paycheck and the first honest raise in this whole paragraph.

Here is what makes it worse, or funnier, depending on how much of your retirement is in this thing. Every single piece of evidence during the week pointed the other way. On Wednesday, the ADP report, a private payroll company’s count that lands two days before the government’s, showed 38,000 jobs, against 47,000 expected. Thursday, Challenger job cuts, a tally of announced layoffs, jumped to 52,881 from 33,429. The employment component of the big services survey came in at 47.8, and anything under 50 indicates contraction. We covered the soft stuff live on Thursday and, like everyone else, treated it as a trend rather than a rumor.

Three separate sources spent the week agreeing with each other and being wrong together. That is not a data problem. That is a group chat.

Broadcom reported Wednesday night. Revenue was $29.6 billion, up 86% from a year ago. Its AI chip business did $16.7 billion, up 221% from last year and up 54% from just three months ago. Adjusted profit was $3.32 a share. Free cash flow, meaning the actual money left over after running the business, rose 95% to $13.7 billion.

The stock fell 2.74% the next day and finished the week down 2.95%.

The reason is the kind of thing that makes normal humans close the tab. Broadcom told everyone next quarter would bring in about $34.8 billion, which would be 93% growth. Some analysts had penciled in $35 billion to $35.4 billion. So the company promised to nearly double in a year, missed the mark by a couple hundred million dollars on a thirty-five-billion-dollar quarter, and got marked down for it. That is a shortfall of well under one percent on a forecast for a number that does not yet exist.

Management also raised its outlook for AI chip revenue next fiscal year to roughly $115 billion, up from the previous “more than $100 billion.” Nobody cared, because the market had already decided what the day was about.

While everyone argued about Broadcom, Micron closed the week above $1,000 a share.

Micron makes memory, the chips that hold information while a computer works on it. It is the least glamorous corner of the semiconductor business, the drywall of computing. It rose 8.98% on the week and 6.10% on Friday alone. SanDisk, which makes storage, led all Nasdaq gainers with more than 17%. Had $1,000 in Micron last Friday? You have $1,089.76, which beat the entire S&P 500 by about eighty-nine dollars in five days.

Meanwhile, the famous names split apart. Meta rose 6.70%, the best performer among the megacaps, climbing over four straight sessions. Nvidia added 5.89%. But Microsoft fell 2.69%, Amazon fell 2.97%, and Alphabet fell 2.35%. The seven stocks everyone treats as one stock stopped moving as a single entity.

That matters more than it sounds. For three years “tech is up” and “the big seven are up” meant the same thing, and most people’s index funds quietly became a bet on seven companies. This week those seven went in different directions while the money underneath them moved sideways into memory and storage.

The United States and Iran spent the week trading fire, and the oil market noticed. The oil fund we track rose 9.45% over the five sessions, including a 5.46% gain on Monday, when the strikes started. We covered that day as it happened.

The stock market’s response to a 9.45% move in the price of energy was to finish the week up a tenth of a percent.

Energy stocks were the best sector of the eleven, up 2.20%, which sounds like a win until you notice the barrel moved more than four times as far. When the commodity rips and the companies that pull it out of the ground barely shuffle, the market is telling you it does not believe the price sticks. Nobody is building a drilling budget around a war headline.

And the thing built for exactly this week did nothing at all. Gold fell 2.86% on Monday, the day the shooting resumed. It bottomed Tuesday, climbed back Thursday, and finished the week down 0.52%. A war, a jobs shock, and a rate scare, and gold ended almost exactly where it started. We made this argument on Wednesday, and the week did not take it back: this conflict has been reclassified in the market’s head from a scary story into an inflation story, and gold cannot compete with a bond paying 4.77%.

Buried in Thursday’s services survey was a prices reading of 72.6 against 66 expected. On Tuesday, the manufacturing version came in at 71.1. Both measure how many companies are paying more for the stuff they buy, and both are at levels that mean the answer is “almost all of them.”

Businesses that pay more eventually charge more. That is not a theory; that is arithmetic with a delay built in.

This is the number that decides whether the Fed raises rates on September 16, and it landed on a Thursday afternoon between a war and a jobs report, which is the data equivalent of a fire alarm going off during a parade. We wrote about it on Friday morning, hours before payrolls made everyone forget it existed.

Real money, real odds, no forecasting genius required.

Ninety-six million dollars of real money says this month’s Fed meeting is a coin flip. Meanwhile, the reporting on the futures market had the odds of a hike near 66% on Monday, at a dead heat right before the jobs report, and back near 60% after it. Two betting markets, tens of millions of dollars, and the collective wisdom is “could go either way, ask us Thursday.”

Also worth sitting with: the market gives a rate cut this year a 7% chance. Anyone still waiting for cheap money in 2026 is waiting for a bus that has been canceled since spring.

The 30-year mortgage rate came in at 6.71% on Thursday, up from 6.66% the week before. The 15-year moved to 6.04% from 5.98%.

Five basis points is nothing. A basis point is one hundredth of a percent, and five of them on a $350,000 loan is roughly $10 a month, which is one pizza or half a tank, depending on where you shop.

The interesting part is underneath. Mortgage applications to actually buy a house rose 2.2% on the week, while refinancing applications fell. Translation into English: nobody is refinancing at 6.71% because nobody’s existing loan is worse than that, but people are still buying, because at some point you need a house more than you need a good rate. That is not a housing recovery. That is a housing surrender, and it is the most human number in this whole newsletter.

News sentiment on the S&P 500 reached its highest reading of the week on Friday, the day the index fell. The crowd felt best about the market on the session it lost money. Put that on a poster.

Meta generated 240 news stories on Wednesday, roughly four to six times its normal daily count, and finished the week as the best-performing megacap. Broadcom’s coverage spiked the day after earnings, which is when the coverage always spikes and never when it would help you.

The most recent positioning report from the futures regulator, which is released weekly and reflects the prior week, showed speculators were net short about 68,000 S&P 500 contracts. Betting against the index has been a losing trade all year, and the professionals are still doing it, which is either conviction or a personality disorder. Bitcoin, meanwhile, quietly added 3.96% to finish near 80,913, and the software company that mostly exists to own bitcoin rose more than 12%. Nobody mentioned it. There was a war on.

The appetizer was Monday’s manufacturing survey, served cold, with the employment piece already going soft at the edges. Tuesday brought the war back to the table and the oil ran, which is the one dish that always arrives hot. Wednesday was Broadcom, an absolutely magnificent main course that got sent back to the kitchen because the garnish was two hundred million dollars short. Thursday was a palate cleanser that turned out to be the actual poison, a services price reading of 72.6 that nobody tasted because everyone was already looking at the dessert menu. And Friday was the check: 162,000 jobs, a 44,000-job apology for July, and the sudden realization that the entire week had been ordered off the wrong menu.

The market ate it all and finished up 11 hundredths of a percent. That is not a meal. That is a tasting flight where every glass cancels out the last one, and you leave sober and confused.

Here is the arc of the week in one sentence. Every soft, fast, unofficial measure of the job market said things were falling apart, and then the slow, official, boring one said they were not, and it also said the previous month’s disaster had never happened.

That should be humbling for everybody. The private payroll count missed by a mile. The layoff tally screamed. The services employment gauge said contraction. Three independent sources, all pointing the same direction, all pointing at nothing. Meanwhile, the number everyone treats as gospel, the government’s monthly report, just revised itself by 44,000 jobs on a figure that half of Wall Street had already built a Fed forecast around.

The honest read is not “the data is fake.” It is that all of this is an estimate, all of it gets corrected, and the confidence with which it is delivered on the day is entirely theater. A market that trades on a first draft is going to spend a lot of time apologizing.

Which brings us to the only question that matters for next week. The Fed meets September 15 and 16. Employment just stopped being a reason to be gentle. So the entire decision now rests on one inflation report on Thursday morning, and the survey data going into it is already flashing that prices are climbing. Ninety-six million dollars of betting money currently calls that meeting a coin flip.

Everybody spent this week looking for a jobs crisis that was never missing. Thursday, they find out whether the inflation problem is still standing right behind them.

Everybody spent the week searching for a crisis that was sitting on the porch the whole time.

Markets don’t wait, and neither do we. Follow Tracking the Trade for fast takes, the numbers that matter, and the occasional meme that hits a little too close to your portfolio.