Here’s the moment Nvidia became the AI king and what it must do this week to keep the throne
0 4
It’s all about to happen. But until it happens, then no money should be made waiting for it. We have to see its success to invest in it. Otherwise, the heck with it. There. That had been the bear case against artificial intelligence, generative AI, accelerated computing and large language models (LLMs), all the essential gobbledygook until May 2023. That’s when Nvidia and CEO Jensen Huang “missed” by $4 billion. That’s right, they reported $4 billion more in revenue than expected. I say “missed” because Jensen likes to tell the story that way. Every time he does, he knows he gets a laugh from me, although my wife was a little baffled about how it could be a “miss” when Jensen told the same joke to her. I like the metaphor, though, because until that quarter you had to be in on the “joke,” so to speak, you had to be in on how Nvidia’s chips were being used by all the Titans, a word I like far more than “hyperscalers,” a word given to us by another total fave of mine, Jayshree Ullal. She’s the CEO of Arista Networks , the preferred networking solution Titans including Club names Amazon , Alphabet , Microsoft and Meta Platforms . No one is laughing about Jayshree’s company, either, as her white box solutions are beating Cisco Systems when it comes to the internet plumbing that connect Nvidia chips to the Titans. Nvidia was just shedding its identity as a gaming chip company, and not known as either an accelerated chip company or a company at the heart of generative AI, with chips that can be trained and then give you the power of inference. It’s the wonder of inference behind all the systems that allow you to input questions to ChatGPT (Microsoft-backed OpenAI), Claude (Amazon-backed Anthropic) and Gemini (Alphabet’s Google). It powers the AI assistants that you have right now, although when Microsoft unveils its assistant in your personal computer, it should cause the biggest refresh in PC history, a reason why we own Best Buy even as it won’t propel the stock until late fall. But let’s not get ahead of ourselves. Last year at this time, Nvidia snuck into the U.S. trillion-dollar market cap club, with that May quarter. That took 24 years. The next trillion in stock market value took 270 days. Now, it is knocking on the door of a third trillion — the likes of which have only been topped by Club name Apple and Microsoft. But ahead of that $3 trillion milestone for Nvidia is the roadblock of Wednesday evening’s earnings . So much is ahead of Nvidia because so many people still haven’t seen what the company’s chips can do. Right now, Nvidia seems to be behind what ads you get when you order something, or what movements robots make, or small but productive changes in call centers, and requests to retailers that can be answered by something far better than a human. It allows machines to be doctors and give you answers with emotions. It is, in short, the genesis behind something that’s between a parlor game — ChatGPT give me a haiku about my love of baseball — and something that might make it so people are more productive. It all seems such small potatoes, frankly, which is why Wednesday seems such a looming day. Last year was about the “beat.” This year, the expectations are so high that everyone’s worried that the bar has, at last, caught up with the analysts who have taken it so high that there’s no way it can give you the magnitude you got last year. A beat and raise seems impossible — and if it is possible, it won’t be next time. So, we have gone from the phase of “it’s about to happen” to “we better see more success” because right now not only it is not worth $3 trillion in market cap, but it might not be worth $2 trillion if all the Titans have their chips and Advanced Micro Devices (AMD) is about to burst on the scene The zeitgeist of Nvidia’s trajectory is so fought that even I, the ultimate “don’t trade it, own it” believer, dread that moment not long after Wednesday’s 4 p.m. ET closing bell rings. What can Nvidia possibly do for an encore? Isn’t that what everyone is thinking? But what if the answer is, this time, you will actually see the initial fruits of Blackwell, the master machine that is so much faster than Nvidia’s current iteration? Until this Wednesday, I think that most people believe Nvidia is like Intel in the 1990s when the 286 and 386 chip and then the 486 kept beating expectations because each new iteration was so much more powerful — and yet, at the same time, then Intel CEO Andy Grove was able to master a transition that nobody thought could be mastered. That transition included those months when we had the tail-end of the old chip and the front end of the new chip. Why buy a 386 when the 486 was around the corner? Somehow, the transition produced no gap in earnings. It remains a mystery how it was really pulled off thirty years later. What does Nvidia really have to do then for an encore? Simple. Nothing. It has to come up with a whole new shot. When you scream encore at the opera, you just hope to get one more song, one more aria. Blackwell, much more of a supercomputer, isn’t an aria. It’s a new way to compute, something that happens when it is so much faster than the accelerated computing that Nvidia currently provides. It doesn’t obviate the old — unlike Intel’s then-386, it doesn’t go out of style when the 486 arrives. The old is still good when stacked together with more old. But the new is so revolutionary and so expensive that if you project Nvidia’s earnings it might be worthy of $3 trillion status. If each Blackwell costs, $40,000 and Nvidia can produce a huge number of Blackwells and Blackwells can be so much smarter than, say, the H100, then the analysts will be baffled but will recognize still one more gigantic new beat and raise scenario from this not-so-small company. I think it is a much taller problem than what Intel faced, not because of a difficult transition—Jensen has made sure that its H100s are still valuable — but because the mortals who decide stock prices won’t be able to get their brains around what Blackwell can do. What does it mean, for instance, if a Blackwell-led platform — not a chip but a platform — can ingest video? What does it mean if it can speed research from years to weeks? Will it be able to help cure ALS, as the team I introduced Friday on “Mad Money,” EverythingALS , hopes to do? When I say help cure I mean allow all of the genetic material that can be harvested from ALS victims to be explored and changed so people, at first, don’t die from it and then make it just chronic, and then cure it — the path that Merck is on with Keytruda, its cancer solver? Is that what speed does? Is that what this machine can produce? The task then is two-fold on Wednesday: We have to see the results from the current iteration and it has to be dramatically special and we have to be able to perceive the results of what Blackwell can do. That’s a one-two punch equivalent to Apple having the iPhone challenged if not dwarfed by the Vision Pro. Yes, it may be that hard. With that in mind let me give you the “own it don’t trade it” case for Nvidia. First, the next iteration has to make it so anything that currently moves, including moves made by humans, can be mimicked by robots. For shorthand, the best example is a robotic bartender that can take a drink order, make it better than a human and virtually sympathize with a patron because of training and inference. Second, we have to think that such a creation is not a trick but something that can be put to work right now to flip burgers or sell an annuity. Third, it has to distance itself from anything AMD makes or that Amazon or Google or Microsoft makes so we realize that those are not rivals, just augmenters of their own systems that allow Nvidia to put its chips to work within the ecosystem of those customers. Too much to ask? Now we are getting into what Jensen does to beat expectations. No, he doesn’t care as much about his stock as Grove did, as “Only the Paranoid Survive,” the best business book ever written that makes the others look like silly, soporific name-dropping tomes. But he does care about beating or he wouldn’t have been so proud of what happened last May. If Jensen can’t pull it off? Here’s the virtue of Wednesday. He may be able to tell the story, through Nvidia CFO Collette Kress, of what Blackwell can do over the next several years as the chips that will be known as supercomputers going forward can do. Maybe we can imagine what they can do after Nvidia explains what they can do that’s so different from the current iterations. Tall order, but Nvidia is a company that’s beaten tall orders routinely we just didn’t really know it until last May. So, we have to hope for the realistic and the ethereal in one earnings report, something far more than a simple beat-and-raise l, and I am betting that we get that, or at least we get the future picture of such an animal. If we do, then the stock can go higher. Anything less than that? We get a sell-off and perhaps a sell-off of large-scale proportions that could drive us as far away from tech as possible, maybe to General Mills or Union Pacific or Club stock Eli Lilly , and GE Vernova , or anything else because everything else will be a safe haven from a tech sector that is inflated by the worth of Nvidia and really nothing else including Microsoft, Alphabet and Apple (my other “own it, don’t trade it stock). A tall order indeed from the sales, the profits, and the forecast lines that will come over the tape within the half hour after Wednesday’s close on Wall Street. (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.
Nvidia founder and CEO Jensen Huang displays products on stage during the annual Nvidia GTC Conference at the SAP Center in San Jose, California, on March 18, 2024.
Josh Edelson | Afp | Getty Images
It’s all about to happen. But until it happens, then no money should be made waiting for it. We have to see its success to invest in it. Otherwise, the heck with it.
There. That had been the bear case against artificial intelligence, generative AI, accelerated computing and large language models (LLMs), all the essential gobbledygook until May 2023.
Read the full article here
0 4