The Consumer Cracks as July Retail Sales Fall 0.6%. Applied Materials Posts a Record and Still Falls 5%.

Friday, August 14, 2026, written at 9:20am Eastern, fifty minutes after the retail sales release and ten minutes before the bell, with University of Michigan sentiment still to come at 10am. The S&P 500 closed at a record 7,798.99 last night, its twenty-seventh record close of the year, after trading above 7,800 intraday for the first time. This morning the American consumer posted its worst month in more than a year and the futures barely moved. That gap is the note.

Top Call: The Consumer Broke and the Bid Did Not Care

July retail sales fell 0.6% against expectations of a small gain, the steepest monthly decline since May 2025, with total sales of $763.6 billion against a revised $768.1 billion in June. Excluding autos and gasoline the drop was 0.2%, and excluding gasoline alone it was still 0.6%. The Census release and the follow-up commentary point at the same cause, which is that the spring tax refund impulse that lifted April and May has now been fully spent.

The tape’s response was almost nothing. S&P futures sit at 7,828.25, up 0.07%. Nasdaq 100 futures are up 0.28% and lead. Only Dow futures are lower, at minus 0.18%. An equity market that shrugs off the worst consumer print in fifteen months is telling you something specific about what is actually driving it, and the answer is visible in the other half of this morning’s news. Applied Materials printed a record quarter overnight and guided the October quarter more than $700 million above consensus. Korea closed at an all-time high after touching 7,000 on the index for the first time in three weeks. Neither of those things depends on an American household buying anything.

Our take: this is the cleanest evidence yet that the 2026 rally is a supply-side capital cycle rather than a demand-led expansion, and it changes what the macro calendar means. Weak consumer data used to be a market problem because earnings ran through consumption. Right now weak consumer data is a rates input and nothing more, which is why softness gets bought instead of sold. The genuine risk in that structure is not that the consumer keeps deteriorating. It is that the capex cycle at some point has to be paid for out of somebody’s revenue, and if the household is the terminal customer for a meaningful share of it, the two halves of this market are not as independent as today’s price action implies. That reconciliation is a 2027 question. It is not today’s.

The Tape

The internal composition is worth a sentence because it repeats yesterday’s pattern in reverse. On Thursday the Dow led and the Nasdaq lagged because Cisco is a Dow name that fell 6%. Today the Nasdaq leads and the Dow is the only index lower. The rotation keeps happening inside the technology complex rather than between technology and everything else, which is what a market with one theme and no alternative looks like.

July Retail Sales: The Refund Runs Out

The year-on-year line is the part that stops this from being a recession print. Sales are still running 5.0% above July 2025 and the three-month window through July sits 6.3% above the same stretch a year ago. What broke is the sequential path, not the level, and the mechanism is identifiable rather than mysterious. Refunds landed in April and May, spending front-ran into those months, and July is where the payback shows up.

That framing matters for how much weight the Fed should put on it. A committee that spent July worrying about energy pass-through now has a labour report that printed minus 23,000 with 103,000 of downward revisions, a CPI that landed in line, a PPI that was flat on the headline, and a consumer that just contracted. Four consecutive data points have gone the dovish way. The calibers we track on September have been spread messily between roughly 40% and 65% all week and this morning’s print pushes the whole distribution lower without settling the disagreement.

Our take: we would not extrapolate the monthly number, and we would not dismiss it either. The honest read is that the consumer is decelerating from a level that was artificially boosted rather than falling off a cliff, and that the composition matters more than the headline because a 0.2% decline excluding autos and gasoline is a soft month, not a break. What has changed is the direction of the rates risk. A week ago the debate was whether oil would force a September hike. This morning the burden of proof has moved onto the people arguing for one.

Applied Materials: We Said Watch the Margin Guide, and the Margin Went Up

Yesterday this note said Applied Materials was the cleanest test of whether the order books everyone is reporting had reached the capital equipment layer, and it said to watch the gross margin guide at least as closely as revenue, on the argument that a fourth consecutive margin-down guide would make margin compression the defining feature of this phase of the cycle. The company answered in the opposite direction on every line. Semiconductor Systems operating margin expanded almost five points to 38.0% as DRAM, leading-edge logic and advanced packaging orders accelerated. Revenue set a record. The October quarter guide clears consensus by roughly $700 million on the top line and 8.4% on earnings. Management pointed at capacity investment to support demand through the end of the decade.

We got that call wrong, and the way we got it wrong is instructive. The margin story we built on Wednesday night held across networking, optics and specialised silicon, and it does not hold at the tool layer. That is not an accident of one quarter. Cisco, Coherent and Cerebras all sell into a customer with negotiating power, whether that is a hyperscaler or a model developer, and Applied sells into a shortage where the buyer’s binding constraint is capacity rather than price. Margin compression in this cycle is a function of who is short and who is long, and equipment is the shortest link in the chain.

And the stock fell anyway, roughly 5% from a $534.54 close to around $508 after hours, with sell-side price targets moving up to $578.91 as the shares moved down. That is now four AI-adjacent names in two sessions that beat and were sold, two with margin falling and two with margin rising. When a variable flips sign and the outcome does not, the variable was not the cause.

Our take: this is the strongest single data point of the week for the durability of the buildout, because a toolmaker guiding to $10.25 billion with expanding margin and a smaller China mix is describing demand that is both larger and better quality than a year ago. The reason it sold is that the stock is up triple digits this year and still 27% below its peak, which is a chart that has already been asked to price a boom and a scare and now has nothing cheap left in it. Position for the fundamentals and expect the tape to keep punishing good prints until the crowding clears.

Silver Lake and Workday: the AI Discount Becomes a Takeout Premium

Reuters reported that Silver Lake is in talks to take Workday private at a valuation around $43 billion, which would rank among the largest software buyouts ever completed. The stock surged as much as 21% and was halted for volatility before closing up 17.8%, erasing a year-to-date loss and leaving it up 3% in 2026. Before the report Workday was down roughly 15% this year and more than 40% from its 2024 peak, on the thesis that seat-based enterprise software is structurally threatened by AI agents that reduce the number of seats.

This is the most interesting corporate item of the week and it deserves more than a line, because it is the first large-scale test of a question the public market has been answering one way for eighteen months. Public investors have been de-rating traditional software on the view that AI erodes the seat count. Private capital is now willing to write a $43 billion cheque on the view that a business with Workday’s retention and cash conversion is worth more than the public market says. Both cannot be right about the same asset.

Our take: the take-private bid is the natural clearing mechanism for a group that has been sold on narrative rather than numbers, and we would expect this to be the first of several rather than an isolated event. Legacy software with high recurring revenue, real free cash flow and a compressed multiple is precisely the profile that leveraged buyers want when credit is available, and the AI-disruption discount has done the work of making the entry price attractive. The read-through is broader than Workday. Every large-cap application software name trading at a discount to its own history because of an agentic-AI story now has a floor under it that did not exist yesterday, and that floor is private equity rather than fundamentals.

Korea: A Record Close, and a Second Look at 7,000

The Kospi rose 2.41% to close at a record 6,977.34, its fifth consecutive gain, after trading as high as 7,010.86 in the morning session, the first time above 7,000 in fifteen trading days. The index has now added roughly 11.5% on the week. Samsung Electronics rose 2.4% to 274,500 won and SK Hynix gained 3.26% to 1.65 million won, with Kioxia up around 7% in Tokyo. Japan followed, with the Nikkei up 0.6%, or 405 points, to 68,713.80.

The Korean move is not a sentiment rally and it is worth being precise about why. SK Hynix’s chief executive has said publicly that 2027 will be the worst year on record for memory supply shortage and that demand will exceed the company’s capacity beyond 2030. HBM is sold out through 2026 across all three manufacturers. The Korean government is backing a plan to double national memory capacity within five years with roughly 400 trillion won of investment behind it. This week added strategic joint-venture structures at SK Group designed to manage the capital intensity of that expansion. Every one of those is a supply fact, and supply facts in a shortage translate directly into pricing power.

The link back to Applied Materials is the whole reason both stories appear in this note. A memory industry that needs to double capacity has to buy the tools from a very short list of vendors, and Applied’s DRAM and advanced packaging strength is the same shortage viewed from the seller’s side. Korea rallying 11.5% in a week while a toolmaker expands margin five points is one phenomenon reported twice.

China went the other way for the third session running. The Hang Seng fell around 1.1% to 25,116 and Shanghai was roughly flat at 3,927, with the CSI 300 soft. The regional semiconductor bid is not reaching Chinese equity, which is consistent with Applied’s disclosure that China fell to 28% of revenue from 35%. Export controls are visible in both markets at once.

Europe: Germany Carries It

The DAX rose 0.7% to 26,476.21 while the FTSE 100 slipped 0.1% to 10,761.60 and the CAC 40 fell 0.1% to 8,654.97, leaving the Stoxx 600 near 659 and holding the record shelf it has occupied all month. The dispersion is the point. Germany is the index most levered to industrial and semiconductor capital spending, and it is the only one of the three that participated in this morning’s chip strength.

There is a second European angle in this morning’s geopolitical news that has not been priced yet. Washington’s pressure campaign on Iran is migrating from the Strait to the financial system, and the second-order effects of a sanctions programme that reaches into correspondent banking land disproportionately on European institutions rather than American ones. That has not moved a single bank stock this morning, and we think it will eventually.

Oil and Geopolitics: Two Tankers and a One-Two Punch

Brent rose about 1% to roughly $87.31 and WTI added about 1% to roughly $81.83, reversing part of yesterday’s decline, and the catalysts are specific rather than atmospheric.

Two tankers owned by ADNOC, the Abu Dhabi state oil company, were attacked by drones on Thursday evening while transiting the Strait of Hormuz. British military reporting put the damage as minor. The UAE foreign ministry blamed Iran directly, called the strikes a flagrant violation of freedom of navigation and described them as acts of piracy by the Revolutionary Guard Corps. Separately, Treasury Secretary Scott Bessent said the administration will announce measures next week that will apply, in his words, economic isolation of a kind never seen in history, describing it as a one-two punch alongside the continued blockade of Iran’s ports. Defense Secretary Hegseth said the blockade can run indefinitely. Iran’s central bank governor said the country will soon join the BRICS New Development Bank, which is the clearest signal yet that Tehran is building a parallel settlement channel rather than negotiating for relief.

Our take: the crude response is far smaller than the headlines warrant, and that tells you the market has stopped pricing escalation risk and started pricing a frozen conflict. Six months into the blockade, with the Iran and Oman accord still unsigned, both sides have found a level of pain they can sustain. That is a bearish setup for the front month and a bullish one for the tail, because a stalemate suppresses volatility right up until the moment a drone hits something that does not sustain minor damage. We keep the $85 to $90 band and hold the view that the risk is skewed higher. The financial escalation Bessent described is also worth watching in the currency rather than the oil, since a sanctions programme of that scale is a dollar-plumbing event before it is an energy event.

Also Worth the Time

Reddit jumped more than 10% on confirmation it joins the S&P 500 effective August 18, the mechanical index bid rather than anything fundamental, though the inclusion itself marks how thoroughly the market has re-rated user-generated content as AI training and licensing inventory. SanDisk held its investor day on Thursday and closed up 13.67% after trading up more than 17% intraday, taking market capitalisation to roughly $227.7 billion, which extends the storage repricing this note has been following since the August 6 print and is the same memory shortage expressed in NAND. International Money Express rose 25% after its acquisition by Western Union cleared approval, a small deal but the second M&A completion in a tape that had been quiet for two weeks. Uber and Pony.ai agreed to put 2,000 Chinese-built robotaxis on European roads, which is the first meaningful Chinese autonomous vehicle deployment in a Western market and a regulatory story worth tracking rather than a stock story today.

Key Events Today

• 8:30am: July retail sales, released at minus 0.6% against roughly plus 0.1% expected, ex autos and gas minus 0.2%
• 10:00am: University of Michigan preliminary August sentiment, July final was 55.2 with one-year inflation expectations at 4.2% and five-year at 3.3%
• 10:00am: June business inventories
• 1:00pm: Baker Hughes rig count, more relevant than usual with the blockade in its sixth month
• Next week: the Treasury measures on Iran that Bessent flagged, and Reddit’s S&P 500 inclusion effective Monday the 18th
• September 11: August CPI, the last inflation print before the meeting
• September 15-16: FOMC, with September odds pushed lower by four consecutive soft prints

What We Are Actually Watching

Three threads into the weekend.

The first is that we were wrong about the margin thesis and the way we were wrong is more useful than being right would have been. Wednesday night’s framework said AI mix compresses gross margin everywhere it touches, and Applied Materials expanded margin five points on exactly the mix we said would hurt it. The correct version is narrower and better. Margin in this cycle accrues to whoever is capacity-constrained, which means the toolmakers and the memory manufacturers keep it, while the people selling boxes and inference capacity into buyers with alternatives give it away. That is a sorting rule with a testable edge, and it explains Coherent’s expanding margin as easily as it explains Cerebras compressing ten points. What it does not explain is why all four stocks fell. The answer there is price, not profit, and a market that sells four consecutive good prints is describing its own positioning rather than the companies.

The second is that the macro tape has fully inverted in eight days. On August 6 this note was tracking a market pricing a September hike because oil was running. Since then payrolls printed minus 23,000, CPI landed in line, PPI was flat on the headline, and retail sales just contracted 0.6%. Four consecutive prints have come in on the dovish side of expectations and the S&P has responded with a record close. The thing to be careful about is that the market is now long the same trade from both directions, holding equities because growth data is soft enough to stop the Fed and holding the AI complex because its demand is independent of that growth. Those two positions are compatible only as long as the softness stays confined to the household. The moment weak consumption shows up in an enterprise budget, they conflict.

The third is Workday, which we think is the most underrated item on the tape this morning. A $43 billion take-private bid for a company the public market had marked down 40% on an AI-disruption story is private capital calling the public market’s judgement wrong in size. Whether or not the deal closes, the bid establishes a reference price for an entire cohort of software names that have been sold on narrative. We will be watching for the second one, because a single approach is an opinion and a second one is a repricing.

Marks above are pre-market and will move by the open. University of Michigan sentiment at 10am is the last data point of the week and the one most likely to confirm or contradict the consumer story this note just built.