The AI trade has left the hyperscalers in the dust. What will it take for that to change?

We are starting to see the real weaknesses of these hyperscalers. Amazon, Alphabet, Microsoft, and Meta Platforms may have money, but they've hit a brick wall in this stock market. That brick wall is a hardware. I should have realized how serious the shortage was when we saw how memory chip stocks galloped towards much higher stock prices. The chip shortage, caused by just a handful of players in the high-bandwidth memory (HBM) category (SK Hynix with a share of about 60 percent, followed by Samsung and Micron with about 20 percent), is a bottleneck they cannot overcome. HBM is a special type of dynamic random access memory (DRAM) that is vital for AI computing. We know this because Apple has had to endure price increases under pressure from memory manufacturers shifting more of their capacity from consumer-grade DRAM to HBM. Stocks of another class of memory chips (Sandisk, Western Digital, and Seagate), all focused on long-term data storage, also appear to be capped. They do their best to offer innovations. I wish they would expand directly with new factories, but according to storage companies, this seems like the wrong approach. The opaque nature of the cost of these chips in a business-to-business context is undeniable. We can't open online chats to find out. It just looks like a big black box. Maybe that's why we didn't know how severe the penalties were becoming for these hyperscalers with billions of dollars in capital expenditures. We know that both Microsoft and Meta have cited higher component pricing in earnings calls as one of the factors behind their big investment numbers. Shares of all four hyperscalers fell last month, while the tech-heavy Nasdaq rose almost 1%. Meanwhile, a basket of memory stocks is up 41% in one month. Of the four hyperscalers, only Meta is almost entirely consumerized with an advertising model. This dependence on advertising budgets severely restricts the company in the minds of the market. Meta needs a web services business just like the other three. If Meta had one, I could see its shares doubling and many people, including us, couldn't wait to get one. Having a cloud business will make it easier for Meta to show a clear return on investment from all its AI investments. The commodity is down 12.55% since the beginning of the year. MU YTD mountain Micron's year-to-date stock performance. I thought HBM tightness would be alleviated by additional chip production facilities, known as mills. But they can't get online fast enough or get more out of their existing machines fast enough. That's because the real intellectual property in this supply chain is neither the hyperscalers nor the memory chip makers, but the capital equipment companies: Applied Materials, Lam Research, and KLA Corp. We need more of what they produce, but we won't be able to do it in time to determine which hyperscalers can win. Capital equipment companies are all dual and triple focused, so they are considered more dangerous than Micron or Sandisk. But I think that's probably wrong, because Applied Materials CEO Gary Dickerson told me last month that the company has "unprecedented visibility" from customers because demand is so strong, so I don't think they'll run short based on Wall Street forecasts anytime soon. This whole memory complex has thrown a monkey wrench into hyperscalers' growth plans. It's no longer whether they have enough Nvidia chips, as we saw in the early stages of the AI boom. Hyperscalers have tried to get around Nvidia's woes by working with Marvell Technology and Broadcom to co-design custom AI chips. We chose Broadcom to work with and are almost three years into our ownership. Still, I can't believe Marvell is so high this year, with its shares more than tripling. It's interesting to see Nvidia CEO Jensen Huang embrace Marvell, first with a $2 billion investment in March, and then calling it the next trillion company earlier this month. Why is there so much curiosity? Because Marvell is working with Amazon to beat Nvidia and build its own semiconductor business. Already, Amazon says its chip business would generate $50 billion in annual revenue if it were an independent entity. It's incredible to see how Broadcom's shares crashed even as it continued to work with Alphabet's Google to break Nvidia's bottleneck. While I don't think there's a weakness in this partnership, Broadcom's latest conference call left us confused. The stock was trading at $479 per share ahead of earnings on June 3. It clawed back some of the 22% post-earnings decline over several sessions but still finished Thursday at $411. Fortunately, we took some profits on June 2nd at roughly $480 because the stock experienced a parabolic move in earnings and my long-standing discipline tells me to cut back on parabolic moves. Believe me, as a money manager, I wish I'd bet on Seagate, Western Digital, and Sandisk, or even an exchange-traded fund that follows the Korean market, where Samsung and SK Hynix are by far the biggest names. I could, I should, I could. SK Hynix plans to go public in New York in an effort to expand its investor base and improve its public profile, but the stock has had such a rise that it feels foolish to follow. On the other hand, when you consider the passion versus rigor argument I laid out during our Monthly Meeting on Wednesday, you might buy it. I also regret not owning Applied Materials or Lam Research, because we bought them in "Mad Money" and they are phenomenal companies. Of course, acquiring Arm Holdings this year was a complete success for us. But because of these names in memory and half-capacity equipment, we were constrained by our decision to adopt underperforming hyperscalers. They underperform precisely because of these shortcomings and Nvidia's cost and, most importantly, materials, workmanship and layout. It is now becoming clear to me and everyone else that these companies and Nvidia are in a huge fight to figure out who will win. Meanwhile, memory and storage semi-finished products, formerly considered total commodities, are now so different that they are incomparably different, driving down prices. Despite their run, they're probably better buys than hyperscalers for now. I feel the same way about club names like Corning and Qnity Electronics, winners on both ends of the AI business. Corning's fiber is coveted in data centers for its data transfer speeds. Qnity's special materials are required for making and packaging the chips. Both stocks have more than doubled this year. If I had my druthers, I'd pick one or two more of these quasi-companies that we don't have in our portfolio, but that would involve selling Salesforce, which we want to give another quarter, and perhaps Microsoft, which is being assassinated by vendors who think their business could be disrupted by up to 50%. I'm talking here about Microsoft's enterprise software business and its reliance on an armchair model. The same situation applies to Salesforce. Stay or go? So the question is: Why stay with hyperscalers? There are several reasons. One, or perhaps both, of these will turn a blind eye to AI spending. This causes the other two to roar. I don't think Alphabet will blink because it just raised money and has a valued partner in Apple, which is using Google's Gemini models to provide Siri with much-needed AI capabilities. Microsoft desperately needs to merge with OpenAI. As far-fetched as this idea may seem, I'm starting to believe it's the only way out for Microsoft. Meta needs to build a cloud business or it will become completely irrelevant in this competition. Amazon is very competitive and doesn't stop spending. Assuming Anthropic follows through on its plans to go public, it will be a large company filled with hubris, which would normally lead to collapse. It hasn't happened yet. This means a war between Amazon, Alphabet, Anthropic and OpenAI. Four. Anthropic will have a similar war chest to Alphabet. OpenAI will need much more capital than it would from an initial public offering because it has a more consumer-focused revenue base than Anthropic, which has a business-to-business bias. OpenAI wants to get closer to a 50-50 revenue split, but for now Anthropic is better liked by the investor class. The key for Amazon is to get its AI business to profitability as quickly as possible, which they think could happen next year. Of course, the missing link for all these companies is exactly that: profitability. Capital markets will determine which of the four wins, because I don't think we need them all. In this period, I can no longer see how the shares of these companies can exceed the performance of their suppliers, which is the purpose of this article. I'm laying the groundwork for a few techies to be thrown in to get something off the incredibly lucrative food chain when we face some sort of general market decline caused by the president's erratic behavior. I probably won't have to wait long. Even if I think the Broadcoms, Cornings, and Qnities are adequate, it's hard to admit you're on the wrong horse. That's why Broadcom's recent decline was so painful. Any of these (Sandisk, Seagate, Micron, Applied Materials, Lam or KLA and Marvell) were superior to what we had. We need to reposition, and given that hyperscalers are still spending like crazy, we need to do this soon or until the two blink and then the build rate slows down. But I'm betting that this slowdown won't happen for a while as the stock market easily absorbs the Alphabet share sale and SpaceX's IPO. As you know, I was very concerned that there wouldn't be enough demand for this new supply of stock. I don't know if SpaceX was a sui generis company (there was money for a special occasion, like owning another Elon Musk startup) or if we just had more cash than I thought. I think SpaceX's upcoming acceptance into the Nasdaq 100 would be an interesting narrative and another millstone around Nvidia's neck. The good news is that, as the vaults appear to be replenishing, currently only Anthropic's upcoming IPOs (which will be oversubscribed) and OpenAI, which will be much more stagnant, stand in the way of the market's next run. Why am I not worried about the influence of a potentially more hawkish Federal Reserve Chairman, Kevin Warsh, on all of this? Don't interest rates determine some things? I think they help set prices for Nvidia, Amazon, Microsoft, Google and Apple. But not as much as those mentioned here, they are all about scarcity and price increases. The biggest thing I have to worry about is the thing I'm most dependent on: Nvidia. If Nvidia's shares are going to rise further, it will need to follow Apple's lead and buy back fistfuls of its own shares because the supply is too large. I think a hedge fund would abandon Nvidia, Microsoft, Alphabet, Amazon, Meta, and even short the stock on any rally. I'm just putting it out there. It remains on a path of trust. (See here for a complete list of stocks in Jim Cramer's Charitable Trust.) When you subscribe 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 trading alert before buying or selling a stock in his charitable foundation's portfolio. If Jim talked about a stock on CNBC TV, he would wait 72 hours after issuing the trading alert before executing the trade. THE ABOVE INVESTMENT CLUB INFORMATION IS SUBJECT TO OUR TERMS AND CONDITIONS AND PRIVACY POLICY, TOGETHER WITH THE DISCLAIMERS. NO CIVIL OBLIGATIONS OR DUTIES EXIST OR SHALL BE RESULTING FROM YOUR RECEIVING ANY INFORMATION PROVIDED IN CONNECTION WITH THE INVESTMENT CLUB. NO SPECIFIC RESULT OR PROFIT CAN BE GUARANTEED.