Worst AI Landmine Stocks: Chaikin’s Avoid List
“The Worst AI ‘Landmine Stocks’ to Avoid Right Now” is the fear-side companion report in Marc Chaikin’s American Atlas offer, and it is the half of the package most conversations about the presentation skip. The buy side, the American Atlas Portfolio of 17 companies, gets the attention. But for investors who already hold AI positions, the avoid list is arguably the more immediately useful document, because Chaikin’s thesis is a rotation story: some of today’s AI darlings benefit from the coming FP64 shift, and some are on the wrong side of it.
The premise
The report’s reason to exist is stated plainly in the presentation: “a lot of AI darling stocks are actually doomed right now.” The host, Molly Hendrickson, delivers the sharper framing early: “Today’s AI darlings are, in some cases, about to get clobbered… Along with anyone holding shares,” personalized as “if you have a dime invested in Tesla… Or Nvidia… Or Microsoft… or Alphabet… or Meta… You need to understand what’s coming next.”
The mechanism behind the warning is the floating point problem. Today’s generative AI runs on low-precision FP16 hardware, which rounds numbers in ways that are harmless for chat and disqualifying for science. Chaikin’s demonstration, ChatGPT failing to count the C’s in “broccoli,” and his explanation, “the floating point problem isn’t a software problem. It’s a hardware problem,” establish the ceiling on what FP16 infrastructure can honestly do. The DOE’s FP64 machines, starting with Lux at Oak Ridge, “expected before the end of this year,” open the scientific layer of AI that the current darlings are not built to serve.
What “landmine” means here
The honest reading of the landmine concept is not that these companies go to zero; it is that the market has priced some AI stocks for a version of the AI future that precision hardware redirects. Chaikin’s own gentler formulation, offered to holders late in the presentation, is the accurate summary: “if you’re already positioned in today’s top AI plays, that’s a good thing. What I’m showing you today is a different layer of the same opportunity. I want you to have both.” A landmine, in this construction, is a stock whose AI premium assumes it captures work its hardware architecture cannot honestly perform, like fusion simulation or drug molecule modeling, work that migrates to the FP64 supercluster instead.
How the Power Gauge factors in
What gives the avoid list standing beyond opinion is the tool behind it. Chaikin’s Power Gauge, the 20-factor rating system he built after a career that included clients “George Soros, Michael Steinhardt, Paul Tudor Jones, and Steve Cohen,” and whose Chaikin Money Flow indicator sits “in every Bloomberg and Reuters terminal on the planet today,” rates stocks on factors including money flow, earnings quality, and expert revision trends. The landmine screen is the bearish output of that system applied to the AI sector, and the backtest claim in the presentation, that it “would’ve accurately flashed bullish on at least 7 out of the top 10 stocks of the year since 2016,” cuts both ways: the same factors that find winners early flag deterioration early.
The honest caveats
Two cautions belong next to any avoid list. First, the urgency framing around it, “Do not buy or sell shares of a single AI stock until you’ve heard what he has to say,” is manufactured pressure; no honest analyst freezes an entire portfolio pending one presentation. Second, “doomed” is promotional scaling of a defensible rotation thesis, since FP16 generative AI and FP64 scientific computing coexist, solving different problems, and the presentation qualifies its comparisons with “regarding major breakthroughs.” The useful version of the landmine report is discrimination: knowing which of your AI holdings are geared to precision workloads and which are not.
The report is included with the Power Gauge Report membership, normally $499/year, discounted for the promotion, alongside the American Atlas Portfolio and full Power Gauge access, all covered by the 30-day full-refund guarantee.
Questions readers ask about the landmine report
What qualifies a stock as a “landmine”? In the presentation’s construction, a stock priced for the FP16 consumer-AI regime while the breakthrough work shifts to FP64 hardware at the Department of Energy. The claim: “a lot of AI darling stocks are actually doomed right now.” The companion bullish report is the 17-company American Atlas Portfolio, and both are Power Gauge outputs.
Is this a short-selling report? No, it is an avoid-list, which is the honest and practical form. The average subscriber uses it to sanity-check holdings before adding, not to build a short book. The freeze instruction around it, “Do not buy or sell shares of a single AI stock until you’ve heard what he has to say,” is the one part we encourage readers to ignore outright.
What is the strongest and weakest part of the case? Strongest: the precision argument is real computer science, FP16’s 65,000 values versus FP64’s 18 quintillion, and the DOE buildout at Oak Ridge is real policy on a real calendar. Weakest: the “doomed” framing, since FP64 machines solve a different class of problem than ChatGPT and the promo’s own subtext concedes it. Rotation risk is real; doom is stage language. Our full review covers the whole thesis.
Where to learn more
Start with our End of Elon? American Atlas review, then the FP64 explainer and the 17-company supercluster universe. The full presentation and reports are at the Chaikin Analytics offer page.
NewsletterVetter is an independent publication. We receive compensation from some of the services we review through affiliate links. Nothing on this site is investment advice. Always do your own research.