On Monday, at least, we’re back in a market where Anthropic talks, and we all freak out. The selling pressure hitting AI infrastructure stocks on Monday feels reminiscent of sell-offs in software stocks earlier this year, when investors feared that technological advances from Anthropic (and others) would disrupt their lucrative business models. Now, the infrastructure stocks — chipmakers and all manner of data-center suppliers, including gas turbine makers and electrical equipment vendors — are tumbling on concerns that demand for their products will weaken in a world where AI model development is intentionally throttled back, as Anthropic CEO Dario Amodei called for this weekend due to safety concerns. The declines this time around do feel different, perhaps more serious, because so many AI industry thought leaders co-signed Amodei’s slowdown advocacy, including Sam Altman of OpenAI and Elon Musk of SpaceX, which makes the Grok chatbot. So, has Wall Street traded in a “SaaSpocalypse” narrative — fueled mostly by the AI labs — for an “AI apocalypse” narrative, also fueled mostly by the AI labs? Monday’s action would have you believe the exchange has been made. The stocks getting crushed are the hardware names that had been anointed winners, while we’re seeing rallies in the software stocks that had, in the past, been victims of disruption fears. Investors would be well served by taking a deep breath and avoiding premature conclusions. It’s way too early to declare that the apocalypse baton has, officially, been passed onto the infrastructure stocks. Earlier this year, we argued the bearishness on enterprise software was overblown, especially when it came to cybersecurity. Our view was that AI wouldn’t result in the complete destruction of the businesses that helped make the corporate world what it is today. That didn’t prevent sellers from way overshooting to the downside. That’s just what happens when the herd jumps on a trend and presses the bet until something — in this case, hedge fund Situational Awareness — blows up. Just as the market tends to overshoot the downside, it often overshoots the upside as well. That’s crucial to keep in mind on Monday: We don’t want to chase software stocks to the upside, opening us up to risk that we’ve bought the “overshoot.” In practice, this means we’re not looking to add to CrowdStrike and Palo Alto Networks on Monday with shares up double-digit percentages; same goes for Salesforce, which is up more than 4%. At the same time, we cannot call a bottom in the AI infrastructure stocks just yet. Though we’re encouraged by signs of stabilization in afternoon trading and nibbled on some Micron as a result, we weren’t too aggressive with the purchase, acknowledging the potential for more downside ahead, which gives us the optionality to make further buys at lower levels. The reality is, despite the scary warnings and headlines, nothing has fundamentally changed in our view of the AI buildout. There are a few reasons why: Amodei called for a training slowdown, not an inferencing one. Adoption of existing models isn’t going anywhere. All that demand for compute driven by the desire to implement already released models remains, while new agentic applications built on existing models — like xAI’s Grok Bot or Meta’s Muse — further adds to aggregate demand for compute. China is already referring to this as a “Cold War playbook” tactic, according to the state-backed Global Times. Its argument is that Amodei’s call is being made specifically with the hope and intention to slow down Chinese AI research, so that U.S. frontier labs can pull further ahead. The implication is China won’t be falling for it and, therefore, won’t look to join in on the effort pace training progress. Of course, if all players don’t moderate their development, none will. Amodei himself acknowledged the low likelihood that China would co-sign the initiative. And President Donald Trump coming out Monday against the idea of a slowdown certainly won’t tempt China to slow its own efforts. Open-weight models are just that — open. Anyone can access, download and modify them. The training data may be the key to unlocking them, but the point is that increasingly capable models are being developed outside the controls or risk management frameworks of any one regulatable entity (like Anthropic or OpenAI). China, in particular, has been a hub of open-weight development. That’s why we do not expect Monday’s “AI apocalypse” trade to have the same legs that the “SaaSpocalypse” trade had. We think it will be shorter-lived, though we have to be measured in our approach. We don’t want to act too aggressively because we have two additional, very important dynamics at work that we didn’t have to worry about at the beginning of the year: the midterm elections in November and the possibility of an imminent Federal Reserve rate hike. According to the CME Group’s FedWatch tool , traders are pricing in a roughly 93% chance of a hike Wednesday at the conclusion of Wednesday’s Federal Open Market Committee meeting. The bottom line? Monday is not a day to overreact. We don’t think the updates over the weekend did much to break the underlying thesis driving the AI hardware trade: the world is starved for more compute (even based only on currently available model capabilities). At the same time, though, with a rate hike likely to hit markets this week and midterms in view, this is not a time to make statement buys. Rather, we want to keep a cool head, pick our spots, and continue to build up positions in the long-term winners we think are best able to bounce back when headline-based bearishness runs its course. Should things start getting really overextended and seemingly irrational to the downside, we may want to implement some technical analysis and wider scales , where we look for larger-than-normal declines before putting more money to work. But it is the fundamental argument — demand for compute continues to grow, whether or not the industry, globally, decides to join Amodei in his pacing efforts — that will support every move we make. (Jim Cramer’s Charitable Trust is long MU, CRWD and PANW. See here for a full list of the stocks.) As a subscriber to the CNBC Investing Club with Jim Cramer, you will receive a trade alert before Jim makes a trade. Jim waits 45 minutes after sending a trade alert before buying or selling a stock in his charitable trust’s portfolio. If Jim has talked about a stock on CNBC TV, he waits 72 hours after issuing the trade alert before executing the trade. THE ABOVE INVESTING CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY , TOGETHER WITH OUR DISCLAIMER . NO FIDUCIARY OBLIGATION OR DUTY EXISTS, OR IS CREATED, BY VIRTUE OF YOUR RECEIPT OF ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTING CLUB. NO SPECIFIC OUTCOME OR PROFIT IS GUARANTEED.
Is a ‘SaaSpocalypse’-like sell-off coming for AI hardware stocks? Not so fast