The data center is suddenly suspect. Can Big Tech fix it?
I sure didn’t want what happened to occur, at least not during my vacation, of all things. But it did, and we have to deal with the fallout head-on because it came as such a surprise to many. Imagine for a moment there were five geese — Amazon , Alphabet , Microsoft , Meta Platforms , and Apple — and they were laying golden eggs like mad. No other stocks have consistently produced eggs like these five. It was like a fairy tale. Now that the fairy tale ended a long time ago, their incredible businesses, tremendous leadership, and franchises — moats — keep them a big part of the S & P 500 and the biggest part of tech, which is the biggest part of everything. As long as nothing encroaches on their territories, these mega-caps can continue to invent and profit with their core businesses generating consistent gains. All you had to do was feed them your money, and they turned it into gold. Suddenly, it seems like the geese are being slaughtered. But that’s just exchange-traded funds talking. The pain the geese feel — and they are still very much alive — seems existential. The fact is, however, that it’s not. Not at all. What’s happened is that they are producing okay eggs, good for scrambling. But you didn’t buy them to own egg producer Cal-Maine , did you? Last week, I wrote a piece about what I would do if I were a hedge fund manager — basically, go short ahead of Micron’s earnings (last Wednesday) and then figure out how to get back in at a lower price. I, too, didn’t want to take the gains. I, too, recognize that these companies aren’t any less great, even though they do a lot of poaching. Someone should step back and realize that Meta CEO Mark Zuckerberg, while quite brilliant, has not been able to put together a dream team. In fact, anyone who knows pro sports knows that a dream team is the kiss of death. It’s now all hate and kisses. What really did happen here? Let’s review. First, this whole space-race build-out for data centers has caught the eye of every politician and populist group, and these companies, save Meta and Apple, which have put together institutions designed to build things here on Earth — a commendable gesture, but not one that is getting enough praise. I know it: I study Apple’s Advanced Manufacturing Fund and its efforts to onshore silicon, and I am suitably impressed. Billions of dollars make a difference. As well as the billions of dollars Meta is investing in its America’s Workforce Academy, which will help train the people we need to build data centers. Decades of four-year liberal arts grads can’t turn a screwdriver to hook up the transformers to the electrical system. I digress. The problem is that nobody, at this point, cares at all what they do. They can give away free candy at the corner of Wall and Broad, and no one would take it. They certainly wouldn’t take it on Mission Street in San Francisco or anywhere in Seattle. And, yes, it looks like they aren’t taking the candy in that warren of streets around Congress. These companies didn’t realize the consequences. They are just beginning to figure out the consequences, and now everyone — I mean everyone, including the people who should like them — is against them. So, they are now viewed as a national plague. Of course, it doesn’t have to be like this, but for many investors, it’s too late. And that’s because when you combine it with my second point — the coming white-collar job loss tsunami that almost no one has seen because there are still jobs galore — you get a genuine Stephen King boogeyman, a Pennywise the Dancing Clown. Pennywise is a fantasy. So it is the job loss situation. We happen to have two things going that are being conflated: a severe shortage of blue-collar workers, because of generations of looking down on those jobs or exporting them to other countries, and a white-collar workforce that is scared to death every time they read that a hyperscaler has come out with a program that will take their job away. Which brings me to my third point: These hyperscalers are phenomenal, and they can do things faster and make us smarter, but if we are not at the tiller, they will make mistakes that will end our careers. Sure, they are good at call centers and document creation. The agents do a magnificent job as force multipliers. But with the exception of a small handful of companies that were overstaffed or have pivoted to new businesses — who the heck knows what Oracle is really doing — it just hasn’t happened yet. What we learned from the SpaceX IPO Let’s put these three points together in a way that can be investable, or uninvestable, for that matter: the geese produce eggs that aren’t that special, and we don’t want to feed them because the returns are simply subpar versus other stocks. The companies are pilloried by everyone, and there is no combined narrative about what the heck we are doing. The narrative has gotten so corrupted that we are just betting now that Elon Musk, CEO of Tesla and SpaceX , can throw up one of the largest buildings ever built terrestrially, flick a switch, and outmode hundreds of billions of dollars’ worth of data centers. I love Elon Musk. But the day that SpaceX went public, June 12, and the day after, now look like some sort of fantasy that is not going to end this nightmare. That was the moment we should have realized several things, and, if you please, I want to put it in numbered form. It’s the perfect analog for what has happened. First, we needed to recognize that the money that went to SpaceX was not the traditional golden goose money from institutions and a smattering of individuals. It was a very 1999-type deal with heavy retail presence. It was an incredibly well-done deal. That it went up too much was not the underwriters’ fault. Once public, it got hijacked by memesters who did their best to ramp it up overnight. That’s been a game plan that has worked even for stocks as sizable as Palantir . It failed here. That’s because the size of this deal and the amount of stock that is sloshing around and will be spawned is too great. Many institutions were thrilled to ring the register and were free to do so. The memesters were thrown back after two days’ worth of nighttime hijinks. Unlike with smaller stocks, they didn’t control the opening with their buys. So there were just immediate losses. Lightning struck that gang. They’ve gone back to manipulating penny stocks, knowing that law enforcement won’t bother with them because this market is really all about caveat emptor. What we have come to realize, though, is that it is Musk-dedicated money. There is no golden goose with this man. There are just promises, many of which he made good on, and institutional buyers who regard him as the Warren Buffett of tech. Patience will get you everywhere. These institutions remind me of the firms that stick by Michael Saylor and Strategy . True believers cannot be dissuaded. The difference is that Saylor’s playing nothing but defense now. I don’t expect a defrocking. This guy has nine lives. And I don’t want to say that Saylor, who is a total alchemist, is anything like the greatest inventor of our time. The intersection is the rabid fan base. You just need to know that these pools of money, the Musk money pool and the Saylor money pool, are dedicated only to their practitioners. So the actual events, whether it was the breakdown of crypto or the sale of SpaceX from $200 to roughly $153, did not raise more money for anything else. You don’t see the golden geese getting anything. Why is that? Second, the companies building the data centers are not alchemists. They are real, they are good. But we forget how they got anointed. Their rally started in March of 2023, with the interest-rate shock and the subsequent mini-crisis at banks. As money fled from financials, it did something you rarely see: it went to tech. Specific tech that had figured out how to use the internet to make the most money. Each had something going on that made that happen. Microsoft had Azure, Google had YouTube, the largest communication medium on Earth, and a turbocharged Google Cloud Partners. Amazon had Amazon Web Services. Meta owned billions of eyeballs. Tesla had self-driving and robots, or at least the promise of them, which, because it is Musk, is enough. Then there was the oddity, Nvidia , which had a product so much faster and more durable that it became the basis for the data center. There was enough business for all of these companies. They were golden geese laying eggs all over the place. Their profitability was insane, and owning them became nirvana. And lastly, a couple of things happened that we never expected. ChatGPT was born, and it wasn’t born by one of these elite companies. The artificial intelligence chatbot was developed by OpenAI, and had the market to itself. And it was a market like Apple’s, which I have so far deliberately left out for later. Then Anthropic released Claude, a business-to-business product. The Street loves the latter model, the public loves the former, and is oblivious to the dichotomy. Two players change the game The hyperscalers didn’t see this coming. At all. Sure, Microsoft bought a big stake in OpenAI, but OpenAI has always been a freeform effort. A bunch of companies backed Anthropic, but Anthropic never lost control, and these companies just turned their stakes into investments that they don’t want to take profits in. Lots of people described the deals as circular, but that’s because they were in diapers during the dot-com period. They have big mouths, though. What matters is that there were two new players with instantly recognizable products that made their companies magnets for ever higher fund raises. They reached trillion-dollar status with just a handful of backers, but the backers are venture capital, institutional money, hedge fund money, and Nvidia’s money. All of these investments were geared toward the Anthropic IPO and, to a lesser extent, the OpenAI IPO. The latter, to a lesser extent, because it has a consumer-based model and an executive who has adopted a picaresque management tradition. These two companies changed the game, disrupted the Magnificent Seven, and made investing in the hyperscalers, except Apple, fraught. The eggs weren’t golden anymore. They were more like the byproduct, copper: still needed but only at a price. Amazon, Microsoft, Meta, and Alphabet were threatened by no one until these two companies splashed onto the scene. They felt like the splash was more of a riptide. They decided to adopt a defensive stance of keeping up with the two Joneses, Anthropic and OpenAI, rather than the aggressive gambit of investing with them. Was it protection money? Was it a hedged bet? Was it a belief that if they built data centers, they would get their business? Was it all of those? I think the latter. In retrospect, Meta dropped out. It is building out power, but for whom? We don’t know. Maybe just its own advertising model? That’s an ugh, and that’s why its stock is going down. A coherent statement from Zuckerberg right now, one that says, “We aren’t going to spend this data center money and wreck our balance sheet,” or, even better, “We are going to monetize the power by building a web services system,” —either one would get the stock out of the doldrums, which makes it a tantalizing investment still. Tesla went all in, building out data centers in a remarkable Musk way — buy a lot of Nvidia chips, string them together, and profit from them by renting them to hungry players like Anthropic and Google, both rather amazing because Google spends a huge amount of time bashing Nvidia and saying its chips are much, much better. How ironic. Google has a 15% stake in Anthropic and a gigantic semiconductor business, but it uses a Nvidia-based data center. You might suspect that Google’s chips, as great as they may be, aren’t as popular with users. They want Nvidia, which is a principal reason why Nvidia is the world’s largest company for now. Microsoft released an AI product, Copilot, but it is widely regarded as the second fiddle to other AI products. That’s bad, much worse than we all thought. Because while Microsoft has Azure, it also has a huge software business that can be disrupted, and that business looks more like the carrion of software-as-a-service (SaaS), to be quite pejorative about it. And it is building out data centers to support its AI and Azure businesses, but it might be overbuilding even as it says it throttled back spending. Amazon is just in it to win it. After stalling out, Amazon Web Services is going great guns. It has a $50 billion semiconductor business. It has Prime, which is very profitable, and advertising, which is extremely profitable. But along the way, due to its commitment to the data center build-out, it lost its pristine balance sheet. The company now feels like it did before it ever got profitable. Crucially, is it supposed to make money with its investment in AI next year? That’s why people stay in it. That’s why we stay in it. We trust CEO Andrew Jassy. We trust CFO Brian Olsavsky. Why not? Don’t they deserve our trust? Alphabet, after initially stumbling on its AI strategy, has hit the jackpot with Gemini, its family of multimodal AI models, and has struck a deal with Apple that makes it relevant in AI, no matter what, by making it, as well as Google Search, the official Apple AI product. Apple proves that it is better to be lucky than good. It wasn’t a matter of luck to have the best consumer devices; that’s a given. It was lucky that the company didn’t spend too much on AI because Google’s doing that and then paying Apple to use its AI (we don’t know the real terms of the deal but we know that when you combined the amount that Apple gets paid to be the destination for Google Search with how much it gets paid by Alphabet to use Gemini, it comes out ahead). Until last week, we thought Apple was the biggest winner, hence why its stock was the last man standing. Makes sense: if you have 2.5 billion devices out there and they have a “free” AI without the encumbrance of data centers, that’s a fabulous business. Its balance sheet remains pristine. Alphabet, correctly sensing that there may not be enough money to finance SpaceX, Anthropic, OpenAI, as well as Amazon or Microsoft if they wanted to tap the markets, ingeniously raised money in a spot fashion. Wow, in retrospect, that was genius because it kept the balance sheet from being ruined. That’s what kept people in that one. Nvidia was still a winner; you raised money to pay the piper. When you go into a data center, it is what you see. So, these companies managed to catch up to OpenAI but not Anthropic, which, while wildly promotional, has now captured the hearts and minds of Wall Street with its AI assistant Claude. It looked like everything could work out. If Amazon were right, it could start making money next year with these investments, and the spending could stop. Alphabet had its money and a winning AI combination because of Apple. SpaceX got its money and its two classes of stock, so it can now buy Tesla and remain true to its flock. Meta had its advertising business and a potential move into the data center business, which would likely boost its stock by 200 points. Microsoft had Azure and hoped it would dazzle us with something better than Copilot. OpenAI and Anthropic had their trillion-dollar valuations and Musk-like status, at least to the institutions that participated. It always seemed like OpenAI’s private stock was walked up by ever-hopeful owners. That’s right, they already owned it and kept buying. Anthropic? Still real-deal status: a company that could be profitable but was going for growth. An oligopolistic market And then? Storage happened. The most basic of chips — even as they are just different enough that you can’t pit them together to get better prices — went from being tight in price to being stratospheric. We couldn’t own SK Hynix or Samsung here. They are the biggest makers of storage. We can buy some real also-ran companies: Western Digital , Sandisk , and Seagate , plus a tremendous manufacturer in Micron . Every one of these was needed to go full-out. But it was harder than you think. Sandisk, Seagate, and Western Digital had been burned so many times over the last three decades that they were prudent in putting up plants. At last, in 2024, Micron told me I was being too aggressive in my demand forecast, which I had made because Nvidia CEO Jenson Huang expressed concern that there might not be another storage method. It wasn’t out of nowhere, but these companies now have a legitimate oligopoly on storage. It looks like it might stay that way indefinitely because they need Applied Materials , Lam Research , and KLA to make the machines that produce all the chips they can sell. Good luck. These machines are next to impossible to procure. I like them more than I like Micron here and certainly more than Sandisk, Seagate, or Western Digital. Now this oligopolistic pricing has begun to really sting every customer, except Nvidia. Apple is complaining. The others are keeping quiet, but they are trying to pass on the price where they can. It’s business-to-business, so we won’t know the price. But we do know that any plans to become uber-profitable next year from AI may be dashed by the lack of storage. Not only that, but we realized that they weren’t the dross turned into gold in the data center. There is Corning , because fiber is replacing copper. There are Coherent , Lumentum , and Cisco , all for optical. And, most importantly, there’s the birth of agents, the part of the AI process that holds out the promise of some profitability for the hyperscalers. Agents, though, aren’t powered by GPUs (graphics processing units), which are still needed. They are powered by CPUs (central processing units) made by AMD and Intel . As AMD makes both GPUs and CPUs, it is in the catbird seat. Intel? By dint of its CEO, Lip-Bu Tan, it is truly in the mix, using profits from otherwise commodity CPUs to make everything else, including packaging and foundries. Foundries. Keep that in mind. It’s why Intel remains my favorite stock. It is the solution to so many problems. It can produce more chips, which will only play an ever larger role in the data center, perhaps dwarfing Nvidia’s role. At one point, we thought that you would need four GPUs for every one CPU. But Intel CEO Lip- Bu Tan says that soon enough it might be reversed, four CPUs for every one GPU. Given that it seems to have unlimited power to produce CPUs — as does AMD — it is an intermittent Goose producer, making deals, making profits, and perhaps making the foundries that can save the industry from the stranglehold of the memory oligopolists. Last week, with that Micron quarter, it all came home to roost. The oligopolists have won. They have hurt Apple’s profitability. They may have made it so that building data centers is just too prohibitive for everyone, and that the cost of using a data center, if you are Anthropic or OpenAI, is too high. Anthropic now has to rush its IPO, but into very unfriendly hands, given all the negatives I just traced. OpenAI? It must go public now. Risk it. Get the money that might have gone to Anthropic — now. We are now to the point that the unthinkable might have to occur. If OpenAI doesn’t go public, it will run out of money, and Microsoft will buy it for much less than it is going for now. Can any of the others merge? The egos seem way too big. Apple is merging with no one. But what will Meta do if it doesn’t build a data center? Will Amazon not become profitable in AI next year? Then what will it do? The market doesn’t have the appetite it did for that sneaky Google and its spot offering. SpaceX is fine; there are enough true believers for the balance sheet to be irrelevant. Apple could miss the quarter because of the oligopoly and be revealed to be too expensive a stock. Nvidia is no longer the king of the hill, and it is moving quickly into sovereign AI, perhaps because Huang says the day is coming when these giants all run out of money. Because the public has lost its appetite and moved on to other things. You saw much more than just Micron’s ascent last week. You saw the other growth stocks, the drugmakers, get premium multiples. That could last through earnings, but not if bond prices fall and yields rise. Some industrials are going higher, chiefly non-data-center ones. If the war in Iran is over, that means aerospace can rise. Ditto the banks. They should be AI winners, but have you seen any sign of that occurring? Just press releases. And last week, you saw the return of software stocks, as they aren’t about to report and have gotten too cheap. I don’t trust them, but the move seems more than just a couple of days’ worth of points. It’s all very unsettling. We stay in Amazon because of the promise of profitability in AI next year. We stay in Alphabet because it has the money. We stay with Meta because it seems they are ready to unleash the world’s best data center cohort. We stay with Microsoft because it can develop better AI, and it might be able to buy OpenAI for a reasonable price if OpenAI fails to go public soon. We stay in Nvidia because it is ridiculously cheap, and it might cash in its winnings and just buy its stock, just like Apple did, taking out a quarter or even a third of its float. Why not? It worked for Apple; it will work for Nvidia. All ifs but doable ifs. It does make sense to trim Microsoft. Maybe even Amazon. I have said own it, don’t trade it for Apple and Nvidia. But the day may be coming when I can’t say it anymore. Now, remember, all of these ifs depend on a government that remains feckless and people who are too disparate to stop the data center build-out. But that will happen on its own soon enough, which is why everything tied to the data center is now suspect. Except one company: Intel. It is the only one with the optionality. It can bust the oligopoly. It can coin money on CPUs as the oligopolists do in storage. It can go into the lucrative packaging market because that’s the world Lip-Bu Tan is from. I am not deriding the opticals. I know that Arm Holdings has its asset-light adherents. We are one of them. Qnity is going to go from a chemical company to a tech company and get its multiple ever higher. Same perhaps with Arm. To count Broadcom out seems silly, as it is the only answer to Nvidia. I still like Eaton and GE Vernova because they are long-term builders, but, like Vertiv , the multiples are shrinking. But I keep coming back to Intel. We need a sell-off to buy more. We could get one if OpenAI is listening — they aren’t — or Anthropic makes a move to go public. We might have a sell-all-tech, meat axe call, which would allow us in. Or maybe one of the savior investors cashes out. Can you imagine if Nvidia did so and announced an accelerated share repurchase? I am now willing to call the golden geese brass geese until proven otherwise. The only certainty is Intel. The rest? They may not be as good as Johnson & Johnson , Eli Lilly , or Goldman Sachs . I repeat, every one of these hyperscalers has hope. If they simply read this piece and do what I say, they could realize that hope. Not hubris. The people know tech much better than I do. But by dint of my now 45 years in the business, I know stocks much better than they do. Ah, but my hole card is a three of clubs. Sadly, they won’t even read this, let alone do it. So let’s just stay tuned. (See here for a full list of the stocks in Jim Cramer’s Charitable Trust.) 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.