The AI trade has entered a puzzling phase. Do we know who the winners are anymore?
HS 15/02/2026
All hail Anthropic, the destroyer of everything. What an amazing service. It can wipe out Adobe by allowing you to design something better that has a much simpler commerce system, designed personally for you. That one is child’s play. ServiceNow may call itself the “AI Control Tower” to signify who is in charge, but the cognoscenti doesn’t regard CEO Bill McDermott as anything other than a salesman – even as he has won more business than anyone else from the pre-AI school and I don’t hear many complaints. Aneel Bhusri is coming back to Workday , the software giant he co-founded, to be its CEO. That must mean that there’s not much to the old Workday, formerly run by Carl Eschenbach, who wasn’t able to articulate why you should use Workday when any fool can code with Anthropic to do better than the expensive Workday, which is best known for its HR and finance software, two deadweight areas better staffed by Anthropicans, or whatever God – oops CEO Dario Amodei – says to be the case. Atlassian ? I can use AI to code an alternative to Atlassian’s collaboration software. Anything they can do I can do better. Oh and Salesforce , with its Agentforce that has so many customers who, unlike the others, cheer for them? Ones like Stuart Miller from Lennar , Ramon Laguarta from PepsiCo , Richard Smith, the COO of FedEx International, which has pulled away from the transport pack. Ones like Michael Dell, from the eponymous hardware company, and Laura Alber, CEO of Williams-Sonoma , and David Joyner from CVS Health , where the kind agentic robots replace the unkind robots who tell you which number to press and please don’t forget the #. Those are the stocks that have been crushed and are presumed to be the Digital Equipment Corporations, or the Bings, the AltaVistas and the Prodigys. They are like shooting fish in a barrel because we know that Anthropic is the king of business-to-business AI, which is much better than business-to-consumer. They just landed a $380 billion post-money valuation , and if they came public, clamoring retail investors would pay $500 billion. They wouldn’t pay $1 trillion, because that’s reserved for OpenAI, which the public will love the way it loved the 1999 dotcoms. The people in Silicon Valley and on the syndicate desks know this and they know that OpenAI will get around its tricky governance structure issue because CEO Sam Altman, like Elon Musk, will just call you an idiot and you will be an idiot because we live in a capitalist system and he has more money than you. In fact, shame on you for making a judgment because you are too stupid to talk to him, let alone criticize him. What fools these non-OpenAI mortals must be. It’s worse, of course. The software incumbents thought they had been given a reprieve from the “AI is eating software” threat in the fall, when it looked like OpenAI was on the ropes and couldn’t pay Oracle for its ridiculous, albeit brilliant, data center build strategy – ridiculous because Oracle had no idea how costly it may be and brilliant because Larry Ellison is richer than you and therefore smarter than you. Haven’t you learned that by now? IGV .DJI 6M mountain A popular software ETF’s performance versus the blue-chip Dow Jones Industrial Average over the past six months. Now, the great hyperscalers are wrecking their balance sheets. They all raised their capital expenditure outlooks to ridiculous numbers because they fear being an also-ran like Bing and, in some cases, like Microsoft , they have nothing better. Furthermore, Microsoft executives threw in their lot with the mercurial Altman, who doesn’t care at all about them and seems to outsmart them at every turn. Plus, they have 15 million paid subscribers for Microsoft 365 Copilot – out of over 450 million M365 commercial seats – and they are proud of that? What do we know about Copilot other than we have to hide it every time we pull up a document so we don’t have to use it to summarize the document? It’s the same as its video-conferencing software Teams. We can’t get rid of Teams, but it is harder to use than Zoom so we prefer to use Zoom. No one writes about it, but Microsoft force feeds you its stuff because while it isn’t as good, you are trapped in its world because our companies aren’t going to shell out for what we use at home, which is Apple. It gets worse. Of course, it always gets worse. Amazon feels aggrieved that it beat all the Street’s metrics for Amazon Web Services, yet its stock was still punished after earnings. Did Amazon executives really think we cared about the other stuff when we learned how much more it was going to spend in capex than expected ($200 billion versus the $147 billion consensus )? We didn’t like it any more than Google , which also issued a larger-than-expected 2026 capex guide, but Google was saved by having the superior Gemini, at least for the moment. Meta Platforms ? Who knows? Worse? Who cares? Sure, the Ray-Ban AI glasses are a hit , but they don’t do jack to please the Street, so now the Street fights back with a glorious judgment of irrelevance. As a pathetic member of the general media, I don’t even get a call back from the CFO after the earnings call to ask questions. You can’t see me, but I have a thumb and forefinger up on my forehead making an L. Sure, there’s Broadcom , theoretically important because it is Google’s partner in designing the TPU, Google’s custom AI chip that powers Gemini. But did you hear Broadcom praised on Google’s earnings call? Broadcom is toe jam, a new moniker. It can’t go higher to save its life. And then there is Nvidia , essentially the founder of this AI boom thanks to its incredible GPU chips, run by Jensen Huang. What does he do? How about make a product that costs so much that these customers feel compelled make their own inferior chip or join forces with second-place player AMD. Hence, Nvidia has become a stock that goes from $196 to $186, and then $193 to $183. Jensen says don’t worry because AI is the fourth industrial revolution, but there aren’t enough non-hyperscalers willing to pay – at least not yet – for the generative AI and accelerated computing infrastructure. This is the world we find ourselves in. We have an arms race where companies with pristine balance sheets destroyed them in an instant, in part, because OpenAI and Anthropic don’t have to show how much they are making – or losing. We have companies known to be wealthy who are now debasing themselves. We have enterprise software worth nothing. We have whole private-equity firms like Thoma Bravo and Vista Equity Partners filled with enterprise software that went down in value after they were purchased with debt. If only Thoma Bravo and Vista were publicly traded. If they were, maybe some of the other software stocks we’ve talked about would be holding up a little better because everyone’s attention would be on those two software-focused PE firms. So, now you have read the obituary. What is the truth? It’s so brutal you won’t want to hear it. First, I speak to a huge number of CEOs in whole clusters of industries. On air, I ask them about AI and how they embrace it, and they say it is remarkable. They then mouth what you just read. When I ask them afterward, off air, they say the same thing. I challenge them, though, by saying that sounds like a smart person doing something they should be doing, not a machine doing what a smart person should be doing. When I press them with that, they come back to their goals with AI, which often includes getting AI to help them do things that are dull, dirty or dangerous. Right now, AI is really helping in dull. There is a lot of drudgery at every level of management that you make young people do. AI can replace that. But can AI replace the judgment they learn, the skills they learn from being in the office? The only problem is that AI models make mistakes, so even though the models can essentially “look up” everything in their training data, a poorly timed mistake could make you lose a client. The risk of these crucial errors make it so that I don’t care how many or how few tokens are used, or how little energy it might conserve. Right now, it is just … how do I say it? How about worthless? How about those dirty jobs? I have seen robots clean hospital rooms. That’s positive. I hear there are robots that can make many drinks at a time. That would be good for Bar San Miguel, my once-owned small plate tavern that we sold because my wife opened Fosforo, an agave spirit company, and you can’t own a liquor brand and serve liquor in our great country. Now, I am sure this fellow Dario from Anthropic is a genius because everyone says he is a genius. I have read his predictions . They are about 100% wrong. That doesn’t make him less of a genius. In fact, he’s more of a genius because he has credibility despite that record. Now that’s a genius. Not a con man. A genius. This is the world we find ourselves in. We have an arms race where companies with pristine balance sheets destroyed them in an instant Jim Cramer Last fall, Anthropic said it disrupted an espionage campaign being carried out by AI agents. But call me skeptical that it can really protect you from cyberattacks in the same way that Club holding CrowdStrike can. Maybe it doesn’t patch holes in your AI agents as well as Anthropic, but then again, it never offered that service. CrowdStrike is hard to penetrate and if you are breached, there is much that CrowdStrike does that Anthropic never dreamed of. It might take four quarters for people to understand that, but by that time, it will be too late to buy CrowdStrike. That’s why we continue to own the stock, even though it’s also been swept up in the software sell-off. We bought more earlier this month into the weakness. I am not minimizing Anthropic. I am simply saying that some of its claims – particularly Dario’s late January essay on the “risks of powerful AI” while the startup was in the midst of that massive fund-raising round – were awfully bold. And yet, they were totally lapped up by the investors and the press. I found it naïve, but I am just a stock guy who knows how to move stocks and knows how stocks move. Maybe Anthropic obviates me, too. Here’s what I do know. Anthropic is real. So is OpenAI. So are the efforts of the hyperscalers. But so far, the only stock-market winners in the last six months are the companies that produce products that are in shortage: Micron , Sandisk , Seagate and Western Digital . The fact that Anthropic can’t make a disk drive or a floppy drive is too bad. Applied Materials , Lam Research and KLA Corp. make the machines used to make the hard drives and memory chips and their order books are filled up. I wish Anthropic knew how to make those machines. They were always too complex to knock off. Then there are the companies that cool the data center and provide pumps to do so. Thank you Vertiv , for your huge orders. No, thank you Anthropic. And then there are the companies that supply equipment to the electric grid – Eaton and GE Vernova – and the ones that turn fuel into electricity – again, GE Vernova, hence why that’s a good company to own. We should have owned Vertiv. We knew it would have good orders, but I was so hung up on not having more companies on our sheets and more tech, in particular, so I didn’t buy it. My bad. So, now we are in the situation where we presume Anthropic will destroy everything enterprise software, even as we have customers that swear by their products. Investors are worried that if companies get paid by the seat, the number of seats must be declining because AI will enable companies to cut workforces. We know that Thoma Bravo and Vista own lots of software companies with pre-Anthropic valuations. We know that the boring old memory products are in short supply. And we know the hyperscalers can’t afford to keep building data centers without borrowing huge amounts of money, so no more buybacks and no hope of material dividends and tons of dilutive stock-based compensation. We know that if you utter a word of this piece, you will be regarded as a luddite. You know if you think that Nvidia’s cost is the problem, you know I think you are wrong. Oh, and you know Apple is the winner because it has nothing to do with any of this massive spending, except it is a free rider because it is loved by so many and its customers are all anyone of these companies wants anyway. And there you have it. (Jim Cramer’s Charitable Trust is long NVDA, MSFT, AAPL, GOOGL, META, AVGO, ETN and GEV. 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.
