Why the global memory bottleneck may signal early innings in the cyber stock comeback

The comeback in the cybersecurity group has been nothing short of amazing. Over the course of just a few months, CrowdStrike and Palo Alto Networks have gone from being bucketed into the hated software trade to being anointed full-on artificial intelligence beneficiaries. From the outset of the selloff earlier this year, Jim Cramer was screaming from the rooftops that cyber should never have been lumped into run-of-the-mill enterprise software. That's why we bought CrowdStrike so heavily in March, shortly after the Club stock hit a 52-week low of around $86 per share in late February. We favored pressing the issue on CrowdStrike, while lightening up on Palo Alto Networks with some trims in April, about $40 per share above its 52-week low of $139 in late February. Fast-forward to this week, and CrowdStrike and Palo Alto Networks hit intraday all-time highs Monday, before retreating a bit Tuesday. While we're glad to see the stocks rebounding and scaling new heights, we must be mindful of the speed of these moves. We may think CrowdStrike is an amazing company, and the stock has further upside from here, but that doesn't mean we let it run unattended to the point of having too much influence on the Club portfolio's overall risk/reward profile. On Monday, we right-sized CrowdStrike back to about a 4% weighting, and booked a healthy profit to boot . Last week, we trimmed Palo Alto Networks, whose recent rally brought it back up to about a near 2% weighting. CrowdStrike is up 68% year to date, while Palo Alto Networks is up 88%. So, what has brought the market around to understand what we knew all along? At a high level, the move may not be that different from what has played out in memory-chip stocks over the past few months. When you think about it, it's just the software expression of the memory trade that has driven much of the market's returns this year. Both are tied to AI and the insatiable demand to adopt and implement it. On the hardware side, memory stocks have been absolutely red hot. Sandisk is up over 625% year to date, while Micron Technology is up 250%. The realization: While the world may need more central processing units (CPUs) than thought to run agentic AI workloads in high volume, it's the availability of high-bandwidth memory that is proving to be the real bottleneck for the infrastructure buildout. However, once the hardware is installed and the software applications are ready for adoption, enterprise organizations still need to be certain that they can implement AI safely, without compromising their own data, as well as that of their customers. In today's world, data is the new gold. That means that a data breach is the modern-day equivalent of a bank heist — except bad actors don't just get one shot; they can try, fail and keep trying. In fact, AI makes it so that they can attack a million times per day, and evading capture is much easier when the cybercriminals are behind a computer screen, oftentimes half a world away from their targets. Trying to implement a powerful new technology, such as artificial intelligence, is no small challenge. This is especially true for agentic AI systems, which give access permissions to thousands of digital agents necessary to complete tasks autonomously. All that access exponentially increases the points of vulnerability and plays right into the hands of hackers. It's an arms race with AI used on both defense and offense. To be sure, enterprises understand they have no choice but to adopt AI; failing to adapt to the AI era would be akin to lacking a mobile app following the launch of the iPhone, or a brick-and-mortar business failing to create an online presence as it became increasingly clear that Amazon was driving permanent changes in how consumers shop online. Either evolve and adapt, or be disrupted. Because this new technology is so powerful, it's the cyber companies that will determine the speed of adoption. CrowdStrike CEO George Kurtz said as much on the company's earnings conference call in June: "The inflection point is that every player in this value chain is experiencing hyper growth and every one of these technologies needs cybersecurity." Kurtz added: "The Mythos moment crystallized the reality for the market. And for the first time in my career, the market's view of cybersecurity's role has shifted from being viewed primarily through the lens of risk management, compliance, and protection to being recognized as a strategic accelerator and a critical enabler of AI adoption." That "Mythos moment" happened in early April when Anthropic unveiled Project Glasswing , which initially brought 11 organizations together, including CrowdStrike and Palo Alto Networks, to help secure the startup's new Mythos model, which proved in testing to be adept at spotting security holes. That's when cyber stocks really started to take off. Project Glasswing has since expanded to around 200 organizations. With all that in mind, there is one key difference between the cyber trade and the memory trade. Memory chips represent hardware sales, which can be somewhat lumpy and cyclical as supply/demand dynamics play a huge role in pricing power, even as players like Micron have been working to reduce that cyclicality with long-term contracts. Cybersecurity, on the other hand, represents an ongoing, recurring revenue stream, the kind that Wall Street rewards handsomely. It is also important to differentiate because the cybersecurity trade is more about volume growth over time, whereas the hardware story is about physical bottlenecks leading to massive pricing power for those selling the hardware. We shouldn't expect to see some incredible, unexpected jump in earnings for the cybersecurity companies, as we have for many of the hardware players. Micron's earnings per share last quarter, for example, soared an otherworldly 1,200% compared with the year earlier. Palo Alto's EPS was up 156%, while CrowdStrike's jumped 51%. The massive step up in hardware earnings was the result of buyers scrambling for a limited supply of a physical good (remember, more money chasing the same or less supply is how you get inflation). That dynamic doesn't really exist in software, where companies can scale with nearly no increase in marginal costs. Customers are looking to implement a security solution on hardware they already have installed. The more hardware that gets installed, and the more workloads that run on them over time, the greater the opportunity for the cybersecurity players, which again is why we think these are names to own longer-term, as AI proliferation increases. However, that newly installed hardware, on which AI applications will eventually run, is more about an expanded addressable market than a near-term earnings booster. Palo Alto Networks CEO Nikesh Aurora said as much on the company's latest earnings conference call: "If you thought that the terminal value of cybersecurity was gone, like many SaaS companies, this terminal value is here to stay. You actually just created a longer-term 'G' in your model for long-term growth rate for cybersecurity." He added, "I wouldn't get ahead of my skis and start throwing the kitchen sink at numbers for cybersecurity companies because there is still a process, a mechanism, a cycle that people buy in and there's execution and deployment. So, to the extent that, do I see good demand? Yes. To the extent do I believe that this demand will continue for longer? Yes. To the extent do I expect a windfall next quarter, the following quarter? No. I expect robust growth." So, investors in cyber names don't need to worry too much about the investment cycle — with hyperscalers Amazon , Alphabet , Microsoft and Meta Platforms projected to deploy a combined $750 billion this year , up more than 80% from 2025. If anything, the real demand for cyber companies comes once the AI infrastructure buildout is complete and workloads are being run on the installed hardware. In that world, the pervasiveness of AI will only be greater, the capabilities more cutting-edge, and the attacks from bad actors and resulting demand for cybersecurity unrelenting. (Jim Cramer's Charitable Trust is long CRWD, PANW, AMZN, GOOGL, MSFT, META. 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.