S&P 500 and Nasdaq 100 futures are starting the week off on a rough note as the “bad momentum” in oil and yields returns amid the continued inability of the US and Iran to broker a deal to reopen the Strait of Hormuz.
News of Akamai’s cloud deal with Anthropic, announced after the close on Thursday, spurred volumes on Friday that were 8,495% above the trailing one-month daily average.
This was a case of Robinhood traders buying the rip in a stock that had only been seeing lukewarm purchases ahead of this revelation.
Meanwhile, Costco was dumped even on heavy volumes after rising nearly 3% following its quarterly results, paring some of its 18% decline since mid May. The members-only retailer appears to have been suffering from a similar problem as peer Walmart: it has run-of-the-mill revenue growth and a very high valuation, with a forward price-to-earnings ratio north of 40.
It’s not good to be an expensive stock that isn’t wowing investors with top-line growth in a world where top-line growth is ample. In some respects, safer consumer-oriented behemoths like Costco and Walmart have a similar “ARR” model as software companies (if you squint). And they’ve both lost some of their allure as those tech companies haven’t been immediately disintermediated by AI tools.
Microsoft’s big advance on Friday was also an excuse to hit the exit button for Robinhood traders, who continued their recent trend of selling the megacap tech giant.
As a subsidiary of Robinhood, Sherwood Media is restricted from writing about any company in which Robinhood is or was a selling group member of the IPO during the regulatory "quiet period" for that company.
Mixed signals?
Over the past weekend, I’ve been hanging out with friends to do fun stuff doing a lot of reading about the seemingly puzzling combination of diverging cross-asset volatility (implied vol high in US bonds, low in stocks) along with brutal breadth in US stocks.
These are, in my mind, two sides of the same coin.
The loose conceit has been “how long can the S&P 500 hold up and equity vol stay low when most stocks seem to be under pressure and the bond market is going berserk?" (And the answer may well be “no longer!”)
“Nobody makes money on the long side with any consistency or regularity when NYSE Cumulative Breadth is negative and NYSE New Lows outnumber NYSE New Highs,” quipped John Roque, head of technical analysis at 22V Research.
I did a little data validation on the first part of Roque’s statement: Friday gave us our 24th instance of “it’s been 29 sessions since NYSE cumulative breadth peaked” since May 2007.
During those stretches, the average S&P 500 return has been -2.2%. This time, it’s held up surprisingly well, falling just 0.5% since August 14.
.png)
But how has this apparent dichotomy persisted?
Well, for starters, by not really being a dichotomy. Implied volatility in the S&P 500 has stayed low because realized volatility has been low. Implied volatility in bonds has been rising because realized volatility has been rising.
Surging yields (and, often, oil prices) have catalyzed a big deterioration in breadth. Equity index vol has stayed subdued because (most) tech heavyweights and semiconductor stocks haven’t cared.
Why? Take your pick/<shrug emoji>. You can chalk it up to Muse-related enthusiasm. Taking a longer lens, you could argue that stocks with better profit revisions than the broad US market can’t be persistently punished with underperformance, especially after the steep swoon to start Q3. Or something else.
To be sure, these pockets of the market have faced stiff headwinds from rising yields and oils multiple times this year — at the onset of the Iran war in March, again near the 2026 bottom for the S&P 500, and also in early June following a hot May payrolls report.
One factor going forward that might also help keep a lid on implied volatility in stocks absent a crescendo in macro fears:
It may feel like earnings season just ended, or is still going on <cough> Micron <cough>, so it gives me no great pleasure to report that the next round of quarterly reports for most of the Magnificent 7 will be in the 30-day forward window — that is, the period used to calculate the VIX — later this week.
Earnings season is a time when stocks tend to move for their own unique reasons, so dispersion is high and correlations are low. Single stock vol had already started to rise relative to index level vol as the meatier parts of earnings season start to come into view.
I am not aMused
US banks have been one of the so-called “consumer inertia” pockets of the market that have been whacked amid the enthusiasm around Meta’s Muse.
The thinking: customers will be more able to move deposits out of banks into higher-yielding accounts at fintech alternatives.
Apollo Global Management chief economist Torsten Slok threw some highly-priced diesel on that fire this weekend, publishing a note titled, “Is an Agentic Bank Run Coming?”
Apropos of nothing: a tip of the cap here to former colleague John Authers, who once self-censored from reporting on firsthand evidence of bank runs in New York City’s financial district in 2008, believing “ there was the risk of a fire, and we might have lit the spark by shouting about it.”
Also apropos of nothing: “The Financial Select Sector SPDR Fund is flagging as an optimal entry as it is now oversold on our Oscillator. The deterioration in Financials is concerning but the sector is still in a bullish trend,” writes Jeff deGraaf, head of technical research at Renaissance Macro.
Throughout 2026, positive catalysts for the AI trade have often been treated as zero-sum. Finding victims to shoot first and asking questions later has been the normal state of affairs (just ask software companies in Q1!). We’ve flagged how many of these “inertia” stocks were already rolling over and underperforming long before Muse captured popular attention!
US banks have meaningfully underperformed their global peers since the launch of Muse. Are we to believe that US banks are that much more vulnerable to agentic-driven deposit flight than their peers?
An alternative explanation looks a little something like this: US stocks stand to benefit more than their global peers from an extension of the AI boom (and, thanks to Meta, from this product in particular). Therefore the need to find and punish presumptive “AI losers” is also larger in the US than for global markets.
Are AI agents coming to the rescue to boost consumer welfare by helping to compete away outsized rents, putting some downward pressure on bank net interest margins? Probably.
“If agentic AI is genuinely changing behavior, evidence should emerge through higher deposit costs, lower non-interest-bearing balances and greater competition for marginal funding,” wrote BofA analysts led by Ebrahim Poonawala. “Until deposit costs rise faster than can be explained by rates or competition, the disruption thesis remains conceptual.”
BofA expects that banks will engage in some “self-cannibalization” to keep depositors happy, flagging that JPMorgan has already started to do so.
The idea of a “run” both likely exaggerates the immediacy and intensity of the near-term threat to US financials, and their ability to mitigate tail scenarios with carrots (like higher deposits) and sticks (blocking agentic access).
In my opinion, all of our jobs involve some mix of sales and quality assurance. Frictions exist on both functions — sometimes for genuine reasons, other times as an excuse to extract an extra ounce of flesh.
If you want to buy a cheap shirt and Amazon’s blocking Muse, you can still get something pretty close on Shopify fairly easily. I don’t think a similar alternative would currently exist if banks were to take measures to stem agentic deposit sweeps. While that’s likely not a long-term solution for banks, it’s a lever that could be pulled to require more leg work from you to realize higher deposit rates.
The eventual implications for banks from AI agents may seem obvious (and may turn out to be!), but I’d like to add a couple of notes to the “making predictions is hard, especially about the future” file:
Is AI going to help us save money on insurance by finding cheaper plans and helping recover more from claims? Or…
And the all-timer for this year: the imminent launch of Anthropic’s Mythos was originally a powerful catalyst to dump cybersecurity stocks. A couple months later, the cyber risks raised by that and other new powerful models have been a phenomenal boon to those very same companies!
Sales later > sales now
Goldman Sachs has a pair of fantastic notes out on what’s priced in to the AI trade (some skepticism on earnings delivery), how much hyperscaler capex has been contributing to S&P 500 earnings growth (roughly half of it), and how high a bogey they need to hit to break even on this spend (more than what’s in their backlogs now).
Now that we’ve gotten those high level observations out of the way, one element that really jumped out at me from one of these notes was the discussion of what investors are willing to pay up for, from Goldman’s Ben Snider:
“Revenue growth, and especially longer-term growth, is generally the most important determinant of most stocks' valuations. Today, however, investors are paying an above-average premium for 3-year ahead sales growth and focusing less than usual on 1-year ahead sales growth. This dynamic likely reflects the uncertainty investors face surrounding the durability of current earnings, whether due to the ‘over-earning’ dynamics tied to the AI investment cycle or fears of future AI disruption risk.”
Near-term sales growth doesn’t really matter at all, while medium-term sales growth matters the most!
On that note, 22V Research’s Dauvin Peterson highlighted CoreWeave — a relatively highly-shorted AI company — as a battleground stock with a big gap between its high and average medium-term sales estimates (which, in his view, remain too low):
“While there continue to be rate concerns, we did see analysts start to recognize the need for higher revenue per MW realizations for neoclouds. CRWV 2030 consensus versus high consensus revenue estimates imply a gap for revenue realization per MW of 10–12M/MW/year on the low end and~14M/MW/year on the high end. One example of upside potential is that the market is now north of $25–30M/MW/year (and even $16M/MW/year for long-term base deals such as Anthropic’s); therefore, given their open future capacity, these estimates likely trend higher.”
Seen on Socials
Via @RitholtzWealth on X:
What to watch
Monday:
Jefferies Financial Group and Vail Resorts slated to release earnings after the close.
Tuesday:
CarMax and Carnival Cruise scheduled to release quarterly results premarket.
August Job Openings and Labor Turnover Survey due out at 10 a.m. ET.
MongoDB hosts investor day.
Wednesday:
Conagra and Jabil earnings slated for release ahead of the open.
August PCE inflation report due out at 8:30 a.m. ET.
Micron slated to publish quarterly results after the close.
Thursday:
Accenture and McCormick slated to release quarterly results premarket.
JPMorgan hosts software forum through Friday.
Nike results due out after the close.
Friday:
September non-farm payrolls report due out at 8:30 a.m. ET.
