CJ’s Weekly Market Memo
A Change in the Weather
October 17, 2026
An old saying on Wall Street is “sell when the sky is cloudless”. For the AI trade there was nothing but blue sky until recently. Anticipating the change from clear selling to selling squalls is the job of technical analysis. Few of the people I have worked with over the decades have developed their own technical analysis tools or kept their own data archives to power those market barometers. I have, and I believe it gives me an edge over other advisors and investors. Last month on September 25th one of my models developed to signal a change in intermediate and longer-term price trend triggered such a sell signal. That model has triggered a sell signal within 10% of the high only five times in nearly five decades. Those years were 1978, 1980, 1987, 2000, and 2008. Those signals preceded sharp and sometimes deep selloffs and this September the signal triggered again. That is a good part of the reason for my continued caution toward this market. The prevailing view is that earnings which are growing more than 25% for a potential third straight quarter settle any question of where stocks go next. The feeling is the sky is clear, what could possibly go wrong? I don’t think current earnings are necessarily what the market looks for. Investors are forward looking. And my view is that they have begun to sense a slowdown in that earnings growth rate. So the market has begun to price that slowdown adjusting prices lower. AI leaders, stocks whose price is most dependent on that high growth rate, are where the repricing shows up first. This week the selling reached them. The semiconductors fell for eight sessions in a row through Friday, the longest losing streak since October 2008, and the chart below shows how rare that is.
Philadelphia Semiconductor Index, consecutive daily declines, 1994 to 2026. Source: Bespoke Investment Group.
Market Technicals
The model’s signal matters because this is what it is built to catch, a change in the trend of prices. It is slow by design. It ignores the ordinary pullbacks that come and go inside a bull market and fires only when the deterioration beneath the averages has run long enough to matter. That is why it has spoken so few times in five decades. Its signal arrived on September 25, weeks before the AI leaders cracked and while the headlines were still nothing but record earnings and extrapolations of those records for years to come.
The model is built much like Dow Theory, which is the source of that slowness. Over the years I recognized that sluggishness while markets were beginning to move with ever increasing speed. That is why I pursued the development of additional tools. Tools that were more sensitive to impending market trend changes. One such model mimics much of Lowry’s methods analyzing buying and selling pressure but more importantly whether narrowing markets were distribution by strong hands to the weaker later arrivals to the party. That model is the Lowry insight I have leaned on since March of this year. The signal from this shorter-term indicator has been in place since then.
The months long pattern of distribution has now moved to the leaders. As the unquestioned market leaders, the AI stocks carried the major indices through the months in which the average stock was already correcting. This week they gave way. This week technology was the worst sector, the Nasdaq 100 fell better than 5%, and Bespoke’s basket of neocloud companies, the firms building AI data centers from the ground up, is now down by more than half from its late June high. Seemingly without warning the clear sky clouded over. Three potential thunderheads arrived at once: debt that may prove unsupportable, doubt that profits will be large enough to honor the usage commitments already signed, and a general fear that the product itself is becoming a commodity. Stan Weinstein built his reputation on charts that separate accumulation from distribution, and the AI infrastructure stocks now show the pattern his method was designed to flag, a failure near the old high followed by selling that won’t let up. Of course, the S&P 500 itself has held up better than its leaders. It sagged below its 50-day average to the bottom of its recent range, helped by a recovery in single-stock breadth over the past few weeks, and that recovery is the one piece of the tape that argues against me. However, the slight new high the S&P 500 recorded on October 6 was not accompanied by increasing volume. According to methodology developed by both Weinstein and O’Neil it cannot be trusted. My technical tools are designed to signal selling points and conditions to buy back when the time is more propitious.
What I am waiting for is exhaustion. Selling in the market’s darlings should end when everyone who wants to sell them has sold, and that tends to arrive in a short, violent burst rather than a slow drift. The semiconductors are now firmly oversold and testing the lows of the past six months, which is where such a burst usually begins or ends. If my models turn up from here, with the cumulative advance-decline line rejoining the index and the percentage of S&P 500 members above their 200-day average turning higher, I’ll treat the sell signal as having done its job. If the semis break those six-month lows instead, the process has further to run. Long experience has taught me that the bottom in market leaders comes when the last committed holder gives up, and it rarely announces itself in advance.
Market Fundamentals
Investors look ahead, usually by about six months. According to a growing chorus of analysts the six months they are looking at now contain an earnings slowdown for the AI market darlings. Investors attempt to anticipate economic conditions over the next six months, because those conditions eventually reach corporate returns and so earnings. Those are the fundamental conditions I want to see in agreement with the market’s technical situation before I hold a strong opinion about the investable condition of equity markets. Right now I believe the warning my technical indicators are giving reflects a fundamental weakening that is not here yet but lies in the near future. Current conditions for corporate earnings have rarely been so strong, with potentially three back to back quarters of 25%-plus growth. That is phenomenal. It may also have led many investors to project those conditions far out into the future.
This week’s Barron’s carries a column showing this quarter’s strength followed by a sharp slowdown in the rate of earnings growth over the next three quarters. Earnings would still be rising! But for growth investors, and for their more extreme members known as momentum investors, a deceleration is poison. William O’Neil put current quarterly earnings first in his CANSLIM method, and the discipline he taught a generation of investors was to want growth that is accelerating. I started using his Daily Graphs in 1979 or 1980, and I have watched that discipline move markets ever since. The rule’s failure mode is plain enough. Followed blindly, it sells a company growing at half its former pace as though it had stopped growing, and it sells at the same moment everyone else following the rule does the same. That is how a slowdown in the rate of growth becomes a sharp decline in price long before it becomes a decline in earnings themselves.
The other thing investors are looking ahead to is the cost of money, and this week’s inflation data made that picture worse. Consumer prices came in firmer than forecast, and producer prices surprised to the upside across the board, with two of the three main measures at new highs for this cycle. The PPI level matters more to me than the CPI, because producer prices are where the pipeline starts. One of my first economics classes at The University of Texas at Austin taught me a point I’ve never forgotten: when an economy runs close to the limit of its capacity, any attempt to expand production further shows up in prices before it shows up in output. The latest capacity utilization figures indicate industry still has room to expand before adding new capacity. Bespoke reads the broader supply chain as easing and sees some evidence of disinflation ahead. However, the industries that currently matter most to this market have no such room. ASML’s management said this week that supply as currently planned won’t fully meet demand in 2027 and 2028, tanker rates have broken prior records as Iran steps up attacks in the Strait of Hormuz, the diesel crack (its premium to crude) is back above one hundred dollars a barrel. In addition, computer prices have turned sharply higher as AI demand for chips and memory passes through to consumers. Higher interest rates can’t fix bottlenecks like these. They restrain demand without adding a single fab, tanker, or refinery.
PPI final demand and PPI ex food and energy, year over year. Source: Bespoke Investment Group.
The Fed makes the timing harder. October’s hike is off the table, but futures market pricing still holds one more increase in December and two more in 2027. Bespoke’s work on policy lags suggests this year’s first hike may not show up in the economy until next summer, which falls squarely inside the six months investors are trying to anticipate. So the market is being asked to price a slower rate of earnings growth and the arrival of tighter money in the same window. The next signal will be this quarter’s round of corporate earnings and sales guidance. If the trillion-dollar companies reporting over the next two weeks raise their earnings and capital spending outlooks, the slowdown in the Barron’s column may prove shallower than the tape implies.
Prognostication
My two confirmations, fundamental and technical, agree about the next several weeks and split about the year ahead. The technicals own the timing, and my own model says the decline isn’t finished. The fundamentals own the destination: earnings still growing, but at a slower rate. In my last letter I said the sharp selloff, if it came, would arrive later in October around the reports from the trillion-dollar companies. It started a week early, in the AI leaders, and the reports are still ahead of us.
Leverage always leverage. In his book A History of the United States in Five Crashes: Stock Market Meltdowns That Defined a Nation, Scott Nations argues that major breaks begin when leverage meets a new vehicle the market doesn’t yet know how to price. CoreWeave, with its leveraged GPUs, is that type of vehicle. Its bond, the only publicly traded high yield issue among the neoclouds, fell from 93 cents on the dollar to 84 this week. That, and the negative commentary beginning to appear about private credit and alternative investments, worries me. I don’t believe this is one of his five crashes. But it’s the reason I expect the washout in the leaders to be sharp, because leveraged holders don’t get to choose when they sell.
I may yet be proven wrong, the AI leaders may find their footing at these six-month lows and the slowdown in the rate of growth may prove shallow. Should that happen, I will have been early and my caution excessive. Nothing would please me more.
My list for the coming weeks is short: the guidance from the companies carrying the S&P 500, whether the semiconductors hold their six-month lows, and signs that the selling has reached exhaustion. When my models say the sellers are finished, I intend to put excess cash to work in the better stocks my fundamental analysis helps me select. The leaders usually fall last, and they are often the first to recover.
CJ Brott
Chairman Emeritus, Capital Ideas
The material presented is for informational purposes only and is believed to be accurate. Sources include but are not limited to publications by FactSet, Dow Jones, Yardeni Research, and Bespoke Investment Group. All expressions of opinion reflect the judgment of the author as of the date of publication and are subject to change. The author may hold positions in securities mentioned in this commentary.