Lambert here: I don’t play the ponies, so Richter’s charts, as such, have little charm for me. That said, the money sloshing into AI is so ginormous that most of it must be stupid. And at some point, you have to wonder — although we don’t seem to, even after the dot.com bubble, the Great Financial Crash, and whatever it was that fastened facehuggers like Meta and Uber onto the body politic — whether turning over society-wide capital allocation to stupid people driven by greed, fear, and the stories they tell themselves is optimal.
By Wolf Richter, editor of Wolf Street. Originally published at Wolf Street.
Nvidia is special because the dollars are suddenly so huge – a hair over $2 trillion at the open on Friday, and just a little below $2 trillion at the close. Over the past 12 months, market cap has spiked by $1.46 trillion, including the biggest-ever-for-any-stock one-day spike of $277 billion on Thursday, on exceeding revenue expectations by $2 billion.
There are other charts that look even wilder because these are crazy times, and all kinds of stuff has been spiking and in crazy ways. And Tesla’s chart used to look like this too, before the shares plunged 70% in 2022. Crazy spikes like this generally don’t lead to a permanently high plateau. But Nvidia is now huge, and it’s floating on top of a mindboggling AI-mania.
Nvidia has always been a volatile stock because something always derails the latest GPU-to-the-moon narrative. For example, between November 2021 and October 2022, Nvidia shares plunged by 65%, barely dodging our Imploded Stocks pantheon (minimum requirement: -70% from high). And now the numbers are much bigger. We show the market cap in the chart because the dollars are now so huge that they really matter for the overall market:
The utter mania around generative AI that has suddenly gripped corporate America led to an explosion in sales of high-dollar and high-margin systems of GPUs for Nvidia.
Just about every major company is touting its progress with AI – for now, mostly just spending money on it – and periodically there are hilarious stories about generative AI’s not so intelligent work. I mean, can you fire AI if it screws up badly enough? Or is that suddenly a human’s fault?
The amounts of money thrown at generative AI from all directions are just vast. And Nvidia is getting its share of the pie. Nvidia’s revenues might hit $100 billion in 2024, knock on wood, up from $27 billion in 2022. And it just might, or might not, surpass Tesla’s revenues, which were already $97 billion in 2023.
These Price Spikes Are a State of Mind: See Tesla.
So we just accidently remembered that Tesla’s stock too went through these WTF spikes until it hit $414.50 on November 4, 2021, and then plunged 70% over the following year and was inducted into our pantheon of Imploded Stocks in December 2022, when its shares hit $123, after which they dropped some more. These spikes unwind brutally.
Then came the big rally through mid-July 2023, and then the shares spiraled down again, and today they’re down 54% from the peak in November 2021, and they’re still overvalued though Tesla has become a much bigger and a much more successful automaker since the all-time high in November 2021.
Since that all-time high, Tesla has become very profitable. Its Model Y has become the #1 best-selling model in the world, blowing away Toyota’s Corolla, and it has become the #2 best-selling model in the US, just behind Ford’s F-series truck. And since the all-time high, Tesla opened two new factories, one in Germany and one in Texas, and production and deliveries have soared, and it’s eating market share of the legacy automakers in big gulps, and despite its already substantial size, its global deliveries jumped by 38% in 2023.
And yet, over those two years when Tesla’s business continued to boom, Tesla’s market cap plunged by over $600 billion. Turns out, these WTF spikes are a state of mind. And when that state of mind changes, the spikes unwind.
How Long will the Nvidia Narrative Last?
Will Nvidia have a monopoly on AI-suitable GPU systems? Will no one figure out how to make something competitive and sell it a lower price? And will Meta, a gigantic customer of Nvidia’s GPU systems, suddenly discover that it has all the GPUs it’ll need, just like it had suddenly discovered that it had way too much office space and way too many employees? Companies are kind of weird in what they suddenly discover. All kinds of things can change that might rattle that overstretched state of mind – same thing that Tesla went through to get off its WTF spike.
32 comments
Comments are closed.
“the money sloshing into AI is so enormous….”
Dot.com 2 ?
Dot-com -> subprime -> crypto -> AI…
Not sure what will come next, but the basic problem is that in order to maintain the compound growth of yesteryear VCs and like has to find way to get a profit for over a trillion USDs pr year.
I’ve heard some call this the “everything bubble.” I don’t think that is inaccurate. I’ve given up on this market long ago as I think it is too dangerous for my old money. I’m going with the safe plays like short term T-bills for now.
I have always been a fan of charts and technical analysis, but I don’t consider it the end all guide for investing. But I will say, if you spin up a chart that goes back to say, 1995 before the dot-com bubble and look from there to today – it sure doesn’t look normal. Indexes don’t go straight up forever, especially with some of the underlying data and names we see not doing as well.
It looks like a big ball of air looking for a pin. To put my tin foil hat on, this is an election year. Biden is the preferred candidate. Will they try to keep the plates spinning until the election? A stock market event will not help Joe, but if they can keep the plates spinning, and Joe loses, they can blame Trump if there is a market event – to the delight of many, no doubt. Insiders and Wall Street make money no matter if it goes up or down. Too tin foily? Perhaps. But nothing would surprise me given the people in charge.
I just got a warning from fidelity 401k:
Based on what we know about your current holdings, you appear to be invested too conservatively for your age.1
I am 61.
Things that make you go hmmm
Too conservatively at age 61? hmmmm. Because your retirement horizon is 40 years from now? Me thinks Fidelity is looking out for themselves. / ;)
Keynes: ” When a country’s investment decisions are made in a casino, they are likely to be made badly.”
In the span of about 20 yrs, no less. When a system that is sold as the most efficient way to allocate scare resources, in fact does the complete opposite, you can be sure said system is totally corrupted and eating itself from the inside out.
To paraphrase Trotsky: The power is lying in the street. We just need to pick it up.
His post on reverse stock splits, revealing the rot in the system, is also quite an interesting point.
https://wolfstreet.com/2024/02/23/this-stuff-is-just-funny-we-have-a-boom-in-mega-reverse-stock-splits-to-keep-the-ballooning-number-of-imploded-stocks-listed-a-while-longer/
It’s an interesting time, if one looks where we just were on broad indices as of mid to late October. Holy crap on a cracker, the bulls have been running. I like to think that perhaps these periods begin to flatten or even out, and when they do watch out below. One quiver to note for the Nvidia market action, they do have positive earnings and they do have growth. Growth won’t last forever…
Minor quibble. Everything went wrong in 2022, and I do mean that in every sense of the word, when it came to investing approaches. Long equities, long fixed income…nothing really worked as inflation was certainly becoming the Thing to reckon with.
Ed Dowd points out $300 trillion (with a ‘t’) of global debt has a margin call coming in May. If the Fed and Treasury can’t keep it together, if the debt can’t roll over for some reason, look out below.
The frothy, breathless AI side of the market and business media has many other market pundits warning of an impending downdraft. Given that many rats are already on record cashing out (see tech and other billionaires in the US and EU) you have to think that the AI publicity is a distraction.
Living in interest times.
https://www.washingtonpost.com/doonesbury/
Doonesbury today
That one hearkens back to the Golden Age of BD and Phred! I have long maintained that somewhere on the outskirts of Nashville is a dedicated computer that writes 99% of the commercial Country Music played on generic Country stations. Fifteen minutes stuck in traffic is all you need to understand that. The final five minutes will sell you a new, loaded F-150 for $15K off “factory invoice.”
Computers aren’t writing them yet, but modern country music is now so formulaic that it could easily happen at some point. Computers are definitely used to record the music, and they are also used to force musicians to keep time (also known as a click track).
Modern country = 1970s Eagles lite musically + twangy vocal + formula lyrics (I love you OR they look down on us but we’re gonna keep doing our traditional country values OR let’s go raise some hell).
Any relation to Jimmie Rodgers, Ernest Tubb, Hank Williams, or Patsy Cline is extremely tenuous.
We are at the top of the AI hype cycle and every startup wearing the AI invincibility cloak automatically becomes a magnet for capital. Nine figure seed rounds abound and this money is promptly redirected to fill Nvidia’s coffers as the demand for GPUs continues to be “up and to the right”. In the past one would have been forgiven for thinking the gross tonnage of market cap dollars added to Nvidia suggests that AI is having its Cambrian moment but having lived through the recent hype about NFTs, Web3 et al one has to keep one’s own counsel and wait for the “trough of disillusionment” ala Gartner Hype Cycle to administer the inevitable shakeout (with its attendant pullback of capital as it retreats to the sidelines to wait for the next big thing) and see where the dominoes fall, who’s left standing etc before formulating a clear eyed view about the future of this market. One thing that’s becoming clear is that the LLMs are becoming/will become commoditised so it’s not immediately apparent how the largess will migrate from the tooling layer (where Nvidia operates) to the application layer where most of the startup activity is happening (and frankly most of these startups are building wrapper apps that put them in the cross hairs of the incumbents whose LLMs they depend on, and who will squash them the moment they migrate up the stack into the application layer to claw back revenue lost from model commoditization).
The Nvidia stock chart does arguably track the AI hype cycle very closely (except for the latest spike which seems divorced from any actual news) so I don’t think it’s difficult to see what’s driving it. Nvidia was early to the game with AI and LLM training and has by far the best support for GPU acceleration. So instead of simply powering high end gaming and digital media processing, their hardware is now increasingly essential to business functions in practically every sector imaginable.
I agree that differentiation among LLMs is less than it used to be but there’s still a very clear link between output quality and GPU cycles for all of them, so I don’t see hardware demand dropping off before the AI hype cycle itself does. In the end something out there is generating predictions for you on hardware, and more than likely it’s Nvidia hardware right now.
All of which is to say I would not be in a hurry to short Nvidia even at 2k.
I believe a if not the predominant use of GPUs these days is bitcoin mining, whatever that is.
That faddish abuse of GPUs ended years ago.
OpenAI is supposedly soliciting capital to build their own AI chips. If that really happens, Nvidia’s stock price will naturally drop, but the winner will still be TSMC.
Crowds of humans do suffer mania and a stock market is a very large crowd of humans and…bots. (Can’t really describe the modern stock market without mentioning the algos/bots.)
When people look under the hood…
‘Sell Nvidia’
Sanctions risks, self-funded demand, customers becoming competitors . . . . Just don’t tell the shareholders
https://www.ft.com/content/e1beb7a5-6c91-4d7f-bc90-79689774881d/
And this guy saw that article and felt vindicated.
https://www.youtube.com/watch?v=DBT63k6FaK0/
Something Fishy About Nvidia’s Earnings & Coreweave!
https://www.youtube.com/watch?v=j93wlN-Tf_s/
Mainstream Press Covers Nvidia-Coreweave Story! #Randos FTW!
Time will tell.
Thanks for these links.
The most succinct description of the current AI hype I have seen. Bravo Lambert.
er… um… is ‘Magnetard’ a typo spelling for the firm named Magnetar? Inquiring minds would like to know. / heh ( See the 2008 GFC subprimes collaterized debt obligation ( CDO) trades by Magnetar. / ;)
adding per the utube vid: “And yes, there is a quiet a bit of finesse there…..”
No doubt. / ;
To some extent it is a fad and will fade, but Nvidia actually has one of the few practically worthwhile implementations of ‘AI’ with its upscaling and, especially, frame generation algorithms. AMD is playing desperate catchup, and Intel is even further behind. ‘AI’ for real-time graphics is a massive leap forward, is the future, and is here to stay. Especially with real-time ray tracing also being the new standard; it sucks up a lot of processing power and algorithms allow precious performance to be clawed back. And Nvidia has a huge headstart on both aspects.
J K Galbraith said it best:
“There can be few fields of human endeavour in which history counts for so little as in the world of finance. Past experience, to the extent that it is part of memory at all, is dismissed as the primitive refuge of those who do not have the insight to appreciate the incredible wonders of the present.”
An individual stock like Nvidia is well outside my risk tolerance, so my primary concern is the distortion it’s causing in index funds.
NVDA has a 3 year growth rate for EBITA of 84.3% ., They are projected to grow earnings without NRI by 26.2% over the next three years, and revenues by 30.83% over the next 3 years. The PEG ration is 1.44. Based on projections, the current valuation with a P/E at 66 is rich, GPU’s are a pretty complex design design so there is a decent case for NVDA having a reasonable moat, so if the bull case is correct, they could easily double earnings in 3 years, meaning the future P/E would be in the low 30’s, and if growth rates drop by half by then (doubling earnings every 6 years instead of 3.0 years), you would do pretty well. My guess is the management is probably being super conservative on their earnings estimates so that there is a positive earnings surprise every quarter for the next couple years for publicity reasons.
Its unclear by the numbers that NVDA is in bubble territory at this point, the valuation reflects extraordinary business results. On the other hand, if you start getting a lot of momentum trades doubling the price, it would just be a speculative bubble. I’m not recommending anyone buy NVDA, only that the bull case is not irrational. The multiples on TSLA during its peak were a lot more crazy.
If I have owned NVDA, I would be concerned with:
1. If some of the future sales are based on “vendor financing”. Past investors of Nortel, Lucent would know what I am talking about.
2. How much of code base ( both hardware & software ) is shared with the technology “partners” in China? IIRC, Cisco and Huawei were partners before they became competitors.
Is it fair to say if your AI depends on a robust internet, you do not have intelligence?
Asking ATT cell customers…