It’s time for some of my reflections on a week in San Francisco talking to dozens of executives and workers in tech. It was a grueling trip, but a necessary one, because we are at a pivotal time when it comes to the revolution of artificial intelligence and accelerated computing. AI gets talked about a lot as it’s considered something that all companies have to be thinking about. Accelerated computing makes generative AI possible — and yet, it gets very little attention. There is no debate about AI. It must be implemented, in some form, by every company. But there is a debate about how it should be done. Do you use Microsoft ‘s Copilot AI assistant to interrogate your data? Do you have to build your own model on premises? Or can you just settle for using a cloud version and hire different vendors to interpret your data to make your company better at what it does — make it more productive and valuable to other companies? Club portfolio company Microsoft, right now, is considered the dominant AI force — and when combined with ChatGPT, it is the default choice for companies to work with in some way to be sure they are taking advantage of what AI has to offer. What is it that AI really has to offer? I have to tell you that after my trip I come back with more questions than answers about what it can do for anyone and anything. I know from Nvidia CEO Jensen Huang, who had a terrific fireside chat with Marc Benioff, CEO of Club holding Salesforce , that many companies still don’t know what to do with the lightning speed of accelerated computing or the platform that Club name Nvidia is developing for any company that can afford it. Here are the six themes that I see at play. 1. Biggest bets Companies need to train their devices, giving them the data they need and the ability to access it correctly in some coherent form. Then it can be used to answer questions without the help of computer scientists. I have yet to hear anyone give a specific use case that they are using right now that’s seems really helpful except Amazon , Alphabet , Microsoft, and Meta Platforms (all portfolio stocks), and now Oracle . Almost everyone else seems to want to deploy AI in some meaningful way but don’t seem to know how. That’s why you can still own Nvidia, which is the brain behind pretty much everything, and the hyperscalers, or tech titans, which know how to use it. Meta Platforms can get you the right ads, designed by them, that will make it hard to consider advertising with anyone else — except maybe Google and Amazon. Microsoft’s Copilot lets you access data easily. It’s available on HP ‘s AI personal computers right now but hasn’t really attracted a lot of interest yet — good for Club company Best Buy ‘s forward earnings. Advanced Micro Devices , also a portfolio stock, makes a lot of the chips for the AI PC, again a terrific reason to own that stock of AMD. Alphabet uses it for Waymo, which finally seems to be ready for primetime . Waymo is teaming up with Uber to go into Atlanta and Austin, Texas. The tie-up is great for both. But make no mistake about it, Waymo is the brains of the outfit. We took Waymo self-driving cars everywhere in San Francisco — and after our second one, I didn’t want to go back to a human drive. The future is now when it comes to driverless vehicles, and I have seen the use case for Alphabet that makes me want to hold on to it — at last. Amazon’s always been a user of AI and the company is brilliant and inferring what you need. Amazon is a pure winner. 2. Bots and robots Everyone else? It’s just not clear, until what I saw at Dreamforce, Salesforce’s annual showcase event, something that makes me really excited about AI. It will be practical and very big. At Dreamforce, Benioff revealed the first truly useful initiative that will make it so companies can interact with customers in a seamless and intelligence way using artificial intelligence: Agentforce. Benioff’s brilliant keynote was all about how disappointing AI has been and how companies have spent billions on AI and have nothing to show for it. We keep hearing that the obvious use case, the replacement of the call center will be tremendous. That said, no one seems to have done anything with it that is different from what’s currently being done. Salesforce is going to change all that. Benioff unveiled Agentforce, a program takes all the data a company has, trains the chips with them, and then unleashes them as agents who know pretty much everything about every account and what questions that account’s customers might ask. The use case we saw was how Saks Fifth Avenue uses Agentforce to handle any question you might have about the luxury retailer and what it has for you. That’s a good start. But what Benioff envisions is a wholesale takeover of the entire selling and information process a company may want. The supposition is simple: We all know that an agent, which is really a bot, can be programmed to answer everything that you might need when you call. The agent learns more every day as the data is refreshed and just keeps getting smarter and smarter. From what I have seen and learned last week, in a very short period of time, is the customer will want an agent over a human. We are about to be blown away about how much smarter, more patient and more intuitive agents are then people. There was a time when you could call a company and speak to a human. But humans got too expensive so we dispensed with them and companies seem to have gotten to the point where they don’t want you to call at all. So they give you the press one, press two runaround. That all ends with Agentforce. We are going to go back to the way we were, and you can speak to someone better than a human. It is the true use case of AI because it will make your company so much better and more lucrative. I can see every retailer of any size, any company that has interaction with clients or potential clients, will be at a huge advantage. I’d say unless you were at the keynote you might not realize how potent this will be. For some companies, companies like CVS , this is a no brainer. It will work for any company that has regular questions from customers, like insurance and banks. I found all of this quite exciting, and the first real commercial homerun that will come from current AI. That’s the good news. The bad news? I didn’t see anyone else with a startling use case. Instead, everything seemed incremental in its opportunity. You did get the sense that lots of money is, indeed being thrown away on AI baubles, which may or may not payoff. I did feel that artificial intelligence is best left to the pros — and that, again, means the hyperscalers. The only admitted oddity that worked: the digital twin that can help Siemens build huge factories with the changes made on a small-scale model to save time and trim waste. I don’t mean to say that nothing will come of AI. Lots of smart people are trying to figure it out. Anything done by rote in any white-collar job will be eventually taken over by AI. But it’s a little like driverless cars: People aren’t sure yet if it can be trusted. We all have had too many bad experiences with Gemini or Claude or ChatGPT or Meta AI to trust it yet. That said, if I had an opportunity to speak to an agent in the health care chain I would gladly take that opportunity as these places just can’t staff all the people needed. What people really want is a robot that can do things specifically for them. One that can build things or get things or just give you a cup of your coffee. We want robots to handle almost all interaction with customers at restaurants or places like Club name Starbucks , wherever individuals are overwhelmed and miserable. I think these things are close by. As are bartenders who make great drinks but don’t talk much to you. We want robots to man warehouses. We want robots to handle logistics. I don’t know what will happen to people who are doing this now. I do know that they should be looking soon for other jobs as there are too many companies working on these use cases to think that there won’t be success. 3. Revered and reviled Who will make the robots? The presumption is that they will be made by Tesla — and that’s one of the big reasons why CEO Elon Musk might be buying so many Nvidia chips. Musk, by the way, is almost universally hated by everyone I talked to, a kind of virulent Donald Trump that can’t be avoided — a necessary evil. Who else is considered to be contemptible? Intel CEO Pat Gelsinger, who seems to be a loose cannon par excellence. He is considered to be the principal risk to the U.S. government from the CHIPS and Science Act, someone who might not be able to do half of the things he says he can do. People were incredulous about the idea that Qualcomm might want to buy Intel. Qualcomm is, historically, very anti-fab, that is against the in-house fabrication or manufacturing of chips. Gelsinger has be the biggest bettor on fabs imaginable. Intel is a very distant third to AMD in graphics processing units (GPU), and Arm Holdings is gunning to replace any Intel central processing units (CPUs) that AMD or Nvidia can’t replace. Intel’s Mobileye business has been run into the ground — and, despite attempts to make it seem otherwise, there are no takers for the starved Altera, Intel’s company that makes integrated circuits that can be programed after they’re make. I just don’t see what Qualcomm gets from an Intel deal. On the other hand, AMD chief Lisa Su is revered as a competitor and has produced CPU chips that can take over in personal computers and servers as well as GPU chips that can power artificial intelligence. But AMD is nowhere near the power or speed of those made by Jensen who like Su is widely respected. 4. Enterprise software I did hear, as I expected, a revulsion toward enterprise software where there are just too many companies trying to offer products that will make your company more productive and accountable and more efficient. I can’t tell if people are just tired of hearing about these companies or that they feel they have spent too much on them or if they can be done by in-house AI? There are so many multimillion-dollar programs helped by AI, which are being sold right now. I think that companies are too scared not to take them. It seemed almost bogus to me, but I don’t run an IT department or a finance department or an on-boarding department or a sales department. So maybe it is just lost on me. 5. Power problems What wasn’t lost on me was the omnipresence of Jensen and Nvidia’s Blackwell chip platform, the one that burns so hot that you need a huge amount of power to cool off the data centers, hence the love for electric companies Constellation Energy and Vistra , hence the hope for anything nuclear and for large scale natural gas turbines, the bailiwick of GE Vernova, one I wish we had but I underestimated the need for so much power. That was not very smart as I heard it again and again from Jensen. At least, we have industrials Eaton and Dover , two companies integral in the process of cooling. Those are good stocks. The long knives are out for Jensen after he did not blow away the last quarter. I think that was entirely a problem with the making of Blackwell — their complexity makes it so the yield is too low. There are lots of grumblings about how much Jensen charges — but if you can find something that does more for less, I say knock yourself out. Nvidia’s no lock. It’s attracted way too much hot money, and I sense that it still hasn’t been shaken out. 6. What slowdown Compare that, say, to Apple , where it seems like there’s no hot money in it at all so anything good you hear seems to still get reflected in the stock, even as people were fooled once again by analysts who tell you things aren’t great. I know that I pressed T-Mobile CEO Mike Sievert, which might end up being the largest seller of the new iPhone, and he could have easily said that things were as good as last year. He didn’t. He said that things were much better . Bottom line The biggest winners I heard were AMD, Nvidia, Apple and Salesforce. I didn’t get much good from other enterprise software. While this is a pivotal time for tech because of AI, I think that we must all take a step back and kind of agree with Benioff’s view that AI isn’t being used effectively to streamline or advance the ball at least right now. Maybe one day it will. Until then, be skeptical, you won’t be left behind. (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.
An aerial view of the city of San Francisco skyline and the Golden Gate Bridge in California, October 28, 2021.
Carlos Barria | Reuters
It’s time for some of my reflections on a week in San Francisco talking to dozens of executives and workers in tech. It was a grueling trip, but a necessary one, because we are at a pivotal time when it comes to the revolution of artificial intelligence and accelerated computing. AI gets talked about a lot as it’s considered something that all companies have to be thinking about. Accelerated computing makes generative AI possible — and yet, it gets very little attention.