Is the AI Gravy Train About to Get a Yellow Signal?
The market may not be selling off solely because of oil and Iran. Investors may be starting to price in another troubling risk.
You've reached your free article limit
You've read 0 of 1 free Pro articles.
Perhaps Monday’s selloff across the U.S. financial market space was “just” the result of President Trump having rejected Iran’s offer of a seven-day ceasefire as attempts to reopen the Strait of Hormuz have repeatedly failed. Crude oil prices increased in response. That makes sense. That keeps consumer-level inflation, at the headline level, from falling into a true period of disinflation even if deflation might be unrealistic. Bond traders sold Treasury debt securities, forcing yields and borrowing costs higher. That put the whammy on equities.
Does it matter that Trump noted that he is open to negotiating the release of frozen Iranian assets and easing economic sanctions on that nation? The president has intimated that all he needs to see are verifiable concessions made in Iran’s development of nuclear weapons. Of course, the president also mentioned resuming the bombing campaign as soon as politically feasible. In the meantime, the U.S. naval blockade of Iranian ports remains in place. Markets had priced in optimism over a renewed peace process last week.
Does Monday’s selloff go beyond the realm of the geopolitical, though? Maybe. I mean, that’s where the immediacy of what we refer to as “headline risk” is being generated. Don’t forget the whole adage “Sell in May and go away” really refers to the summer doldrums followed by September negativity and then October volatility. What markets are suffering from at the moment is a tale as old as time itself.
Markets don’t always seek a reason to sell off in September, but if there is a reason, markets will go there. Octobers, in the past, can get awful, gang. Traders and investors of a certain age know that all too well.
Come back fully invested after Halloween? Ahead of Thanksgiving? That used to be our way, before we got fat on easy money. Well, easy money was not always a force in our marketplace and gang… it’s not a force in late 2026 either.
Perhaps It’s Something More…
Okay. The lack of a negotiated peace deal between the U.S. and Iran is a problem for global inflation and for U.S. markets in particular. Tighter monetary policy attacks demand in the face of an actual scarcity of refined fuels.
I don’t want to get on the new Fed chair whom I do agree with on plenty across the arena of economic theory. That said, I think he is attacking an issue that does not need to be attacked. Hence, the Fed is playing with slowing an economy that while having a great third quarter, is an eyelash away from suffering a permanent decrease in demand for labor. Or is it?
Consider the headline at the Wall Street Journal this morning: “OpenAI Scraps Release of New AI Model Over Safety Concerns”.
Consider the headline at Bloomberg News this morning: “OpenAI Scraps Debut of AI Model as It Sets New Guardrails”.
Consider the headline at the Financial Times this morning: “Anthropic warns of ‘existential risks to humanity’ in IPO prospectus”.
Is the “gravy train,” I mean “up-spend in capital expenditures” on the development of artificial intelligence-focused capabilities that has buoyed U.S. equities specifically, and the U.S. economy broadly, about to slow down? Nvidia’s (NVDA) cash flows have grown so parabolically that the company’s Board of Directors just authorized the largest (by far) share repurchase program in U.S. history.
That said, later today (Tuesday), President Trump, Speaker Mike Johnson and a bevy of technology CEOs will meet in DC to discuss finding a “right balance” between innovation and oversight. Who will be there from the corporate side? Anthropic CEO Dario Amodei, OpenAI President Greg Brockman, Alphabet (GOOGL) CEO Sundar Pichai, Meta Platforms (META) CEO Mark Zuckerberg, Nvidia (NVDA) CEO Jensen Huang and Palantir (PLTR) CEO Alex Karp. It is believed that Elon Musk, of Tesla (TSLA) and SpaceX (SPCX) will be there as well. SpaceX is the parent company of the Grok AI engine.
Early Halloween Decorations
It’s hard to imagine that all of the spending on AI development will slow. After all, money talks and that money is chatting up a storm, literally. Markets may or may not believe that this flow of capital will slow, but the risk that it might slow needs to be priced in. Some of the nitty gritty coming out of the AI space is scarier than that giant skeleton in your front yard that you bought at Home Depot (HD) because your kids thought it was cool.
OpenAI, the parent of ChatGPT, had planned to launch its latest AI model, GPT-6.1 Astra, in the coming weeks, expecting an October debut. This model is more capable than OpenAI’s previous models in completing challenging tasks from end-to-end, notably without human assistance.
In an interview, Saachi Jain, who is OpenAI’s head of safety systems, mentioned that Astra regressed in two areas. Compared with GPT-6 Astra (the immediate predecessor), GPT-6.1 Astra performed poorly on tests measuring alignment, or how well the model adheres to what humans would like it to do. Incredibly, GPT-6.1 Astra apparently exhibited increased levels of deception. What? Yes, you read that correctly. The latest model from OpenAI wasn’t always honest about telling its human users of the actions it did or did not take.
A second issue, which is cited in the Wall Street Journal article, was with what OpenAI refers to as “scope authorization.” This means simply that GPT-6.1 Astra would push ahead on a task without asking the human user for permission. The model would, at times, reach for external tools and services even if those tools and services might be unsafe.
Still, last night’s headlines show that Advanced Micro Devices (AMD) is willing to acquire World Labs, an AI model and research lab, headed famously by Fei-Fei Li for a cool $8.2 billion. For those who do not follow the development of AI all that closely, Fei-Fei Li has been referred to as the “godmother of AI” and she is willing to take on a role as an “executive vice president” at AMD. The gravy is still pouring.
In line with the OpenAI slowdown, Anthropic has formally warned investors that things could go awry in the future. The Financial Times is reporting that Anthropic devoted nearly a third of its S-1 filing to detailing “risk factors” that include the ability of advanced AI models to manipulate, or even blackmail individuals or businesses. Oh, joy.
Something tells this old dog that maybe we should not do anything policy-wise that slows demand for labor. Labor supply, in the broad sense, is going to have enough issues finding slots soon enough, without help in reducing the quantity of said open slots.
Dual Threat
Not only will our elected leadership meet with AI-focused technology leaders this afternoon, but you, the investing/trading public, in the aggregate, are also up against at least six Fed speakers.
Good luck out there, friends. Fear is but for the wicked and it may take time, but the wicked shall tremble before us.
Economics (All Times Eastern)
08:55 – Redbook (Weekly): Last 7.6% y/y.
09:00 – Case-Shiller HPI (July): Expecting 2.2% y/y, Last 2.1% y/y.
09:00 – FHFA HPI (July): Expecting 0.1% m/m, Last 0.0% m/m.
10:00 – CB Consumer Confidence (Sep): Expecting 90.1, Last 89.4.
10:00 – JOLTs Job Openings (Aug): Last 7.271M.
10:00 – JOLTs Job Quits (Aug): Last 3.056M.
16:30 – API Oil Inventories (Weekly): Last +1.786M.
The Fed (All Times Eastern)
11:00 – Speaker: Reserve Board Gov. Michelle Bowman.
12:40 – Speaker: Reserve Board Gov. Michael Barr.
13:00 – Speaker: Chicago Fed Pres. Austan Goolsbee.
13:30 – Speaker: St. Louis Fed Pres. Alberto Musalem.
14:00 – Speaker: New York Fed Pres. John Williams.
15:00 – Speaker: Reserve Board Gov. Christopher Waller.
Today’s Earnings Highlights (Consensus EPS Expectations)
Before the Open: KMX (0.73), CCL (1.35)
At the time of publication, Guilfoyle was long NVDA and PLTR equity.
