Week In Review: Everyone Saw the Tide Chart - Everyone Built Anyway

Last week we told you the headliner bombed and the opening act went up 177%. This week the whole venue got a new landlord.

Look at the top three rows, then look at the bottom two. That is the entire week in one table. The giant companies went up, the small companies went down, and healthcare went from the best sector in the market last week to the worst sector in the market this week without anybody throwing it a going-away party. Volatility, the number that prices in how nervous everyone is, spent the week declining and finished at 14.43. That is the market's way of saying "nothing bad can happen here," a sentence that has historically aged like milk in a hot car.

Kevin Warsh has been Fed Chair since May. Friday was the first time he stood up at Jackson Hole, the annual mountain retreat where central bankers go to say important things in a fleece vest, and told everyone how he intends to run the place.

He did not announce a rate hike. He did something that moved more money than announcing one would have. He said the Fed's "predominant focus right now should be on prices," that "inflation is running above our 2 percent target," and, in the line that should be tattooed on every forecaster's forearm, that "price stability is not self-executing, nor is inflation necessarily mean-reverting." In plain English, inflation does not fix itself just because everyone has gotten bored of it.

Then he took away the market's favorite toy. Forward guidance is the practice of the Fed telling you in advance roughly what it plans to do, so traders can position for it. Warsh said that in normal times its role "should be limited and circumscribed," and added that "we should not indulge a regime in which market participants are looking primarily to the Fed for their next trade." He closed by saying he stands "committed to a discipline, not to a decision."

The market heard all of that and immediately went looking for its next trade. Fed funds futures, which are just bets on where interest rates land, moved the odds of a September hike from about 35% to about 57% by Friday's close. The 10-year Treasury yield finished near 4.72%, up about four basis points, and a basis point is one hundredth of a percentage point, which sounds like nothing right up until it is attached to your mortgage. Short-term yields jumped harder than long-term ones, which is the bond market's way of saying it thinks the next move is up and it is coming soon.

Nvidia reported Wednesday after the close and the numbers were not close. Revenue came in at $96.2 billion against the $92.5 billion Wall Street expected. Adjusted earnings were $2.22 a share against $2.09. Data center revenue, the part that sells the actual AI machinery, came in at $89 billion against $85.8 billion.

Thursday the stock rose 8.74%, and we watched the market cancel every worry it owned. Everyone wrote the same headline. The AI trade is back; the doubters were wrong; the buildout is real.

On Friday, it fell 4.57% and gave back more than half of the whole thing. The broader semiconductor group fell about 3.5% on the day, which is a lot of money to lose two days after the biggest chip company on earth told you business is fantastic.

Nothing about Nvidia changed between Thursday and Friday. What changed is that borrowing money got more expensive in everyone's imagination. Chip companies are what the trade calls long-duration assets, which is a fancy way of saying most of their value is supposed to show up later, and anything whose value shows up later is worth less today when interest rates rise. So the market did not sell Nvidia because Nvidia disappointed. It sold Nvidia because a man in Wyoming said the word "prices."

Salesforce rose 22.39% this week. That is the best week of any large company on our watch list and it is not remotely close. Second place was Microsoft at 6.27%.

The report behind it: $11.35 billion in revenue, $5.90 in adjusted earnings per share, both above what analysts expected, free cash flow up 81% to $1.1 billion, and full-year guidance raised. Agentforce, its AI product, is now doing $1.5 billion in annual recurring revenue, which is the subscription money a company can reasonably expect to collect every year without selling anything new.

Now the part that did not make the headline, because good news travels faster when you leave the footnotes at home. Three of the eleven percentage points of that full-year revenue growth are not organic. They came from the acquisition of a company called Informatica. And remaining performance obligations, which is the pile of work a company has signed contracts for but not yet delivered, and the best single tell for whether growth keeps going, grew only 10%. So the quarter was excellent, and some of it was purchased.

Put Salesforce next to Nvidia, and you have the whole rotation in two tickers. Same week, same AI story, opposite outcomes: the company that sells AI software as a subscription rose 22%, and the company that sells the machines making AI possible finished the week up 1.32%. $1,000 in Salesforce at Wednesday's close was worth $1,245 by Friday. The same money in Nvidia that day earned you a slightly nicer sandwich.

Two things happened this week that cannot both be true, and both are.

Thing one: the consumer is falling apart. The University of Michigan's consumer sentiment index for August came in at 51.0, down 4.2 points. The forward-looking piece, which asks people how they expect to feel later, dropped to 50.6 from 55.4. The Conference Board's separate confidence measure slipped to 89.4. New home sales for July fell 10.5% in a single month when the forecast was for a dip of about 1%. People are miserable, and they are not buying houses.

Thing two: the consumer is fine. Redbook, which tracks actual money crossing actual registers at big chain stores, ran 9.1% above a year ago. Real consumer spending in the second quarter came in at 3.4%. Corporate profits for the quarter grew 8.2% when the forecast was 0.7%, and corporate profits are mostly just the receipts from everything the rest of us bought.

So Americans are telling surveys they have never felt worse and then going out and setting spending records anyway, which is either economic resilience or a nation collectively coping, and the data genuinely cannot tell those two apart.

Underneath all of it, Wednesday's data dump landed and the index barely twitched. Core PCE, the price gauge the Fed actually targets and which strips out food and energy, held at 3.3% from a year ago. Headline PCE came in at 3.7%, a tick hotter than the 3.6% expected. Economic growth for the second quarter was revised down to 1.5% from 2.1% the quarter before.

PayPal fell 12.71% on Friday. Not on earnings, not on a product failure, not on anything the company did. It fell because a takeover bid that was never signed, never agreed and never announced by either party went away.

The reported story, first out of Bloomberg and picked up by at least five other outlets the same day: private equity firm Advent and payments company Stripe had been circling PayPal with an offer somewhere around $50 billion. PayPal's board reportedly thought it was too low. Then PayPal turned in a good quarter, the stock rose sharply on the results plus the takeover chatter, and a more expensive PayPal turned out to be a less attractive PayPal. So the group walked.

Notice what actually got priced here. PayPal's business on Friday was identical to PayPal's business on Thursday. What evaporated was a rumor. Roughly one dollar in every eight of the company's value on Thursday night was there only because people believed somebody else might pay it.

Live odds off Polymarket's economy board, pulled Friday. This is people betting real money, which is a very different instrument than people answering surveys.

Read the second row twice. Eighty-nine percent of that money says there will be no rate cuts at all this year. A year ago the entire market conversation was about how many cuts we would get. Now the popular bet is a hike, and September is a genuine coin flip. The gambling crowd arrived at Warsh's conclusion before Warsh did, which is either impressive or unsettling depending on how you feel about the gambling crowd.

Gold had a magnificent week right up until it did not. It ran to its highs on Tuesday, then fell 3.43% on Friday to $4,504 an ounce. $1,000 of gold Thursday night was $966 by Friday's close.

Here is the part that matters more than one Friday. Gold is still up roughly 10% for the month of August and up about 28.6% from a year ago. That is a spectacular year for a rock that produces nothing, pays nothing, and cannot be eaten. Grandma's coins are not in trouble. Grandma's coins had one bad afternoon at the end of an extraordinary run, which is a different thing entirely and a much better problem to have.

The mechanism is simple. Gold pays you no interest, so its competition is anything that does. When the market decides rates are going up and staying up, holding a metal that yields exactly zero starts to feel like a choice rather than a strategy, and some people stop making it. Nobody decided gold was fake on Friday. They just noticed the savings account got a raise.

The 30-year mortgage rate came in at 6.66%, up one basis point from 6.65% the week before. Steady, boring, and stubbornly parked in the mid-sixes for what now feels like a geological era.

The number that actually moved: new home sales fell 10.5% in July, against a forecast of roughly a 1% dip. Building permits rose 4.3%, so builders are still lining up future projects. Home prices per the Case-Shiller index were up 2.1% from a year ago. Purchase mortgage applications slipped slightly.

So builders keep building, prices keep creeping up, and buyers just walked away from the table in the biggest one-month drop in a while. That is a standoff, not a crash. And the standoff gets worse rather than better if Warsh meant what he said Friday, because every honest conversation about your housing payment starts with the 10-year Treasury, and the 10-year Treasury just went the wrong way.

Our news-sentiment read on the S&P finished the week at 0.94 out of 1, the highest reading of the entire week, recorded on Friday. That was the same Friday the index closed lower, small caps fell 1.4%, gold fell 3.43% and the odds of a rate hike jumped 22 points.

Sentiment is a rear-view mirror with a really excellent sound system.

The word-frequency scan of the week's market coverage put "yield" ahead of both "nvidia" and "ai" by weight, which is the most honest summary of this week that anybody produced, and it was produced by a machine counting words.

On positioning, the most recent published futures data has speculators sitting net short both the S&P 500 and the Nasdaq 100 while holding a very large net long position in gold. So the professional money was leaning against stocks and leaning hard into the metal. You can work out for yourself how Friday afternoon went in many very expensive offices.

The appetizer arrived Monday, and it was chips, in the worst sense: semiconductors got dumped before anyone had ordered anything. Tuesday brought the palate cleanser nobody enjoys, a plate of sagging consumer confidence with a 10.5% drop in new home sales on the side. Wednesday was the main course and it was enormous: inflation, growth, corporate profits and Nvidia's entire quarter, served all at once, still steaming.

Thursday was dessert, and everybody had seconds. Nvidia up 8.7%, Salesforce up 22.6%, jobless claims at a lovely 203,000, and the fear gauge down to 14.51, because who needs insurance at a party like this.

Then, Friday morning at nine o'clock, the man who owns the restaurant walked out of the kitchen, wiped his hands on a towel, and said that prices were going up. And everyone remembered, all at once, that there is a bill.

Here is what actually happened this week, underneath that green number on the S&P.

Money did not get more optimistic. Money got more selective. Eight of eleven sectors fell. Small companies dropped 1.40% while the index rose 0.47%, and that gap is the entire story, because small companies borrow at floating rates and mega-cap technology companies sit on mountains of cash. When the market decides borrowing is about to get more expensive, it does not sell everything. It sells the borrowers and hides inside the lenders. $1,000 spread across small American companies last Friday is $986 today. The same $1,000 in the S&P is $1,005.

The tide chart was on the post all week. Warsh's Friday keynote has been on the calendar since the middle of August. Everyone knew the hour. Everyone knew he had already called inflation too high in public, that July's meeting minutes came out hawkish, that three regional Fed presidents voted to raise rates back in July and lost. None of that was a secret. The market spent Monday through Thursday building anyway, and then acted startled when the water arrived at precisely the hour the chart predicted.

The one question that decides next week: does the jobs report give Warsh the cover to actually do it? Friday morning brings August payrolls. July was minus 23,000 jobs. The forecast for August is plus 45,000 with the unemployment rate ticking up to 4.2%. A hot number hands him a hike. A cold number hands him a genuinely awful choice between an economy that is slowing and prices that will not.

Nobody builds a sandcastle expecting it to last. They build it because the afternoon is nice and the tide will be a problem later. This week, later showed up at nine in the morning.

Markets don't wait, and neither do we. Get the quick hits between editions.