NVDA may be a scam: Maybe the real AI gold rush is the junk stonk we pump and dump along the way

I've never cared much for NVIDIA because it's a mega-cap stock and completely out of my league.

But I recently read an article claiming that NVIDIA is a scam. I thought "how is this possible". After all, Enron and Wirecard were not among the largest companies in the world. I can't imagine even one of the world's largest companies being a scam.

However, I did do a research. One of the centrepieces of the NVIDIA narrative is that AI demand drove demand for graphics cards, allowing them to grow sales and profits extremely quickly without a significant increase in the number of cards they made. I've also heard that China is figuring out how to import graphics cards and that US officials are restricting NVIDIA's graphics card exports. smci has even been questioned about evading sanctions because of this.

So I expect graphics card prices to rise similarly to what they did in 2020. However, I was surprised to find that the price of Series 4 graphics cards, which are known as ones of the best NVIDIA graphics cards for training AI, not only didn't go up in the last 18 months, but instead went down by about 20 per cent. OK, this is not what you would expect an AI frenzy to look like.


Of course, it's not as if the NVIDIA narrative doesn't take this into account. Its saying that they're shifting capacity to AI-specific compute cards (not graphics cards) with much higher margins. However, given that Ethernet's shift to PoS in October 2022 increased the supply of graphics cards at the start of the charts (early 2023), this further reduction in supply should result in a significant increase in current graphics card prices compared to early 2023, not a decrease. Not to mention, the purportedly extremely high pricing of its AI-specific compute cards should lead to increased demand for gaming graphics cards that 1 can partially replace their functions, further leading to an increase in their prices.

And oddly enough, AMD's graphics cards actually had less of a price drop than NVIDIA's. But their TTM sales revenue growth is only in single digits. And SMCI, which was recently shorted by Hindenburg, had "only" 100% sales growth after possible accounting fraud, so how could NVIDIA have 200% sales growth? The numbers don't add up.

What's even stranger is that despite a great graphics card price increase in 2020 due to miner demand as a result of the ethereum price spike, NVIDIA didn't get a sales increase and even by 2021, when the supply chain issues have eased, only had less than 50% sales growth compared to 2019, which accompanied more than 100% CoGS growth. So how is it possible that sales growth of nearly 200% with only less than 100% CoGS growth is now possible without graphics card prices skyrocketing, but instead only graphics card prices dropping? Also, the m-score is very high for NVDA:

If one's book is considered more fake than Tesla's, then you need to be careful.

Of course, I have not and am not prepared to short NVIDIA at all, nor am I encouraging anyone to short it. Even given its factor rankings, I think those who are long it are making a good choice among large-cap stocks. But I just find it amazing that this is so patently ridiculous and yet hardly anyone sees it.

Edit:

The article to make me do the research.

I thought it's just another claptrap doom-monger trying to con viewers to get advertising revenue through baseless nonsense without any basis in fact. But now I'm not so sure.

The flagship H100 GPU (14,592 CUDA cores, 80GB of HBM3 capacity, 5,120-bit memory bus) is priced at $30,000 (average).

This isn't your family gaming GPU.

If it's a scan, someone call Musk and Zuckerberg. They plan on spending billions building out AI.

This is the price action for H100.

Prices are falling just as fast, or faster.

It is not what you would expect from a real demand mania

I chose the price chart for game cards simply because they are easier to understand.

After all, this isn't your familiar dot-com bubble scam with nearly zero revenue like those biopharma scam stonks. It is more like the China Hustles with cooked books like PDD, SMCI, ZTO, NIO, etc.

Edit:

Also, AMD has its flagship acceleratior cards MI300X/325X vs Nvidia's H100/H200

They are also not the gaming cards you are familiar with.

If the fabled shortage of accelerator cards due to the so-called "insane demand from big tech, startups and China-Russia" does indeed exist, then you'd expect AMD to also have at least more than 10% sales growth instead of 8% Q vs PYQ and 6% TTM vs PTM

Perhaps CUDA and first mover lock-in is the critical differentiator. Does AMD have a CUDA like solution? If it doesn't, fungibility doesn't exist. I don't know enough about that infrastructure to say.

Do people seriously think that this is a scam?

Maybe I should signup for p123 AI so that I can have my AI tell me when to buy AI.

2 Likes

It's true that some AI models recommend NVDA to me. but they're not very good at recommending stocks in general because, as many papers have said, machine learning model's are much better in a microcap universe.

If rationing due to shortages and sanctions did exist, you'd imagine people would have to buy AMD's accelerator cards to meet their needs. But it's interesting that AMD's revenues in 2023 are even lower than in 2022, even though the H100 cost >$30,000 in 2023.

Edit: At least, NVDA's accelerator cards are said to have extremely insane profit margins like 1000%. So why can't AMD just drop the price to attract more sales?

They may be better w/ micro-cap, but from my experience they can be very good with large-cap, too. Maybe, when I return to p123 AI, I'll try a micro-cap exclusive universe.

I don't know how peeps program AMD GPUs but it's not with CUDA. NVDA has had that technology forever, it's mature, and developers know how to use it.

Prediction: if the US brings anti-trust action against NVDA, I think they'll be compelled to open CUDA to other GPUs; no more NVDA lock-in.

After all, you don't need CUDA to use Pytorch even before this toolkit, just some hassle. But if the fabled shortage of accelerator cards and the resulting insane profit margins do exist, I'm sure people won't care about these minor inconveniences.

What's more interesting is that TSMC's sales in 2023 compared to 2022 have decreased rather than increased. And the TTM sales growth is only in single digits.

So is this because TSMC is saturated with capacity and NVIDIA is moving to other partners? But UMC, the second largest chipmaker, even saw a double-digit drop in sales - even Intel, which is almost reduced to a meme stock, didn't grow its sales at such a low rate, with its TTM sales growth rate just a bit lower than TSMC's (1.99% vs. 5.61%).

As the old joke goes, "You can see the computer age everywhere but in productivity statistics", you can see AI revolution everywhere but in financial statements - except for SMCI's financial statements, which were suspected by short sellers of being faked and had to be delayed, and NVIDIA's several times more impressive financial statements. Maybe we can see the super AI revolution in OpenAi/CoreWeave/xAI/Anthropic's financial reports after its IPO. We can just dream bigger.

In conclusion, I think this means Nvidia will continue to rise. Because the scam involved is so big that its likely to be pushed up in a combined effort to favour all parties in the industry like Wirecard and Bitcoin. However, just as I would advise people not to consider going long or short Bitcoin, I would also advise people not to go long or short Nvidia.

Edit:

There doesn't seem to be a serious shortage of accelerator cards at the moment. Rather, there seems to be a bit of a surplus of accelerator cards already available in data centres at the moment. More than half of the instances are "available".



Are we talking about the stock or the company here?

Is Oracle during the dot-com bubble a relevant example? Oracle (the company) did fine while the price of the stock was a bit of a bubble. Or that is my present understanding.

" Oracle’s stock hit a peak of around $40.10 in March 2000. After the bubble burst, its stock dropped by 59.8%, similar to the broader NASDAQ’s decline. Oracle’s stock continued to decline, reaching a low of $6.26 in early 2002. It took Oracle until late 2014 to fully recover to its dot-com bubble peak. Despite this, Oracle remained a fundamentally strong company and eventually rebounded over time."

I am inclined to think Nvidia the company has been producing real product for a while and continues to produce GPU's. Whether the stock price is a bubble and whether there is hype rising to the level of a "scam" on the public is a separate question.

Some holding Oracle during the dot-com bubble might have wanted to call it a scam, I would guess.

The question I would be asking: Is this time different (assuming Oracle is a relevant example)? And note that if it is not different, timing is everything. The market can stay irrational longer than I can hold out (or stay solvent if I am shorting).

To muddy the water a little bit: Oracle was sued for allegedly misleading stock investors. The case was dismissed in 2009.

To be clear: I hold no Nvidia and do not recall making a recommendation either way. I am glad I did not buy any Oracle during the dot-com bubble. I am not arguing anyone should buy Nvidia (or should have in the case of Oracle). I am personally interested in whether this time is different regarding any possible bubbles. I wish to increase my holdings of individual stocks and I am not sure whether to do it now or after the election. The Oracle story suggests no stocks are immune and that much of the tech sector might be at risk if there is a bubble. This is my only personal interest in this thread. I learned something by studying the history of Oracle—prompted by this thread. I don't claim to be an expert on Oracle, the dot-com bubble, bubbles in general or how to recognize accounting fraud.

I'm not actually interested in Nvidia. I would only buy it if it were recommended in one of my ports, which I could say of any stock.

We're not talking about just the widespread and even unsurprising, but still not in quotes, stock price speculations, but the far less common Enron- and WireCard-like scams where the financial reportings are grossly fraudulent and thus scams like most China Hustles.

SMCI, Enron, WireCard, Luckin Coffee, GigaCloud, Adani, Tesla, Clover, PDD, ZTO, NIO, Riot, Marathon, Icahn Enterprise, Block, and even ilearningEngine ALL share the same "the company has been producing real product for a while and continues to produce" characteristic that has nothing to do with whether or not they are seriously financially fraudulent and thus a scam. It's like the common false and absurd narrative of claiming that current internet giants are profitable "so this time is different than 2000" while ignoring the fact that the large growth stocks that were the main hype at the time were also mostly profitable like Cisco, Intel, America Online, Oracle, Enron, Microsoft and many telecom companies.

That answers my question above. Then to be clear, you believe that about Nvidia or you are just considering that possibility?

I would like to see evidence to the contrary, but I don't.

1 Like

Indeed I suspect it's the kind of China Hustles that Muddy Waters often debunks.

images (2)

But I'm not just talking about "misleading statements" that are so easy to make as to be simply dismissed by the courts, but rather financial fraud at the level of Luckin Coffee and SMCI and beyond.

Also, Oracle's stock is now trading much above its previous highs, and I'd be surprised if Nvidia can do that later.

And I think you've given an utterly ridiculous example of a comparison to whitewash Nvidia's problems and cited no evidence to try to imply that the problem is, at best, over-hyping of the stock price, not gross accounting fraud, and that even the charge of over-hyping doesn't exist because it wouldn't hold up in a court of law and would go up even more later, as it did with Oracle, even though I pointed out the difference very clearly earlier. You are pretending that I am talking about something completely different and need me to state it clearly for it to exist.

My advice: reduce your bag and acknowledge the facts.

Considering your previous attempts to at least move towards rationality on the machine learning issue, I'm honestly shocked that you would so vehemently reject and deny the facts and the truth so badly that you would actually equate the serious financial fraud issues that have been repeatedly raised and criticised as "no problem' issues" that wouldn't even hold up in a court of law as "at best the stock may be a little overpriced but then it'll still be well above the highs". "that there's nothing wrong with it" and probably even that "NVIDIA is just a great, great, great stock with high long-term expected returns and a very good stock, not the best stock at best".

Bags make people dumber. But if you don't even have a bag, then I feel sorry for your preconceptions about the stock market.

Obviously, I don't think a normally-reading adult would ignore the strong allegations of financial fraud above and see the problem as "just a bit overpriced at best at the moment" simply because of a misreading. This can only be a strong denial of facts and evidence and a paranoid expression of one's own prejudices wrapped in the language of modesty.

Edit:

Of course, if you just continue to make excuses and cling to your stubborn prejudices, then I don't really want you to continue to make replies here that very well may include personal attacks, thank you.




Interestingly, while all independent sources agree that AMD's gaming graphics sales in Q1 2024 are not lower or even higher than in Q1 2023, its earnings report states that its gaming graphics sales are down by nearly half and "coincide" with a similar increase in sales in its data centre division to keep the total sales nearly the same.

Considering the more stable price of AMD gaming graphics cards shown above, it is hard to believe their sales of gaming graphics dropping so hard. In fact, the better estimation is that they have a modest increasing in sales of gaming graphics but a slight decreasing in sales of data center accelerators this year.

As a result, it's reasonable to assume that AMD simply reclassified half of their gaming graphics sales as data centre sales to fit their AI narrative to maximise shareholder value (higher valuation -> lower future equity return -> lower equity capital cost -> lower WACC -> higher shareholder value) because there is minus increase in the total sales considering inflation.

While this is not accounting fraud, it is a clear lie to mislead investors in order to inflate the share price. Maybe that's one of the reasons why its insiders only want to sell their stonks:

Considering that AMD's market cap is currently ten times that of SMCI and exceeds the $200bn megacap threshold, it's absurd that a megacap company would engage in such blatant fakery and go almost totally undetected.

Since the US market is already the market with the most rigorously enforced accounting system, the most sophisticated investors and the most developed short-selling mechanism in the world, and megacaps are the part of the market with the most scrutinised accounting, the most informed investors and the easiest short-selling, I can't imagine how untrustworthy anything from unlisted/private businesses, small and medium-sized public companies, or from any institution/organization or even individual, could be.

Edit: However, I don't think that means the stock market will fall or even doom and gloom. Even Bitcoin, which requires massive energy costs to maintain, and is extremely rife with customer money abuse, manipulation and fakery, has managed to maintain its highs with long term speculation, and common stock, as an inferior claim on the assets of publicly traded companies, backed by the profits of more profitable, higher growth and lower valued publicly traded companies, is much more sustainable than Bitcoin.

Edit2: Yes, this time is different. Like the allegory of Bitcoin, this time there could be worse accounting irregularities, wilder speculations and bigger crash(es). I hope they would be true because it would only benefit factor investors a lot like the mania and its crash in 1990s and early-2000s. Just be careful of a possibile momentum crash. For example, LLY as a megacaps may be sus as well

I did a lot of work on accounting fraud last year and wrote a blog post about it here: Detecting Financial Fraud: A Close Look at the Beneish M-Score - Portfolio123 Blog. There are now three ways to measure financial fraud using Portfolio123.

First, you could just look up the M-Score. It's in the snapshot page.
image
-0.75 is very high, meaning a high likelihood of manipulation.

Second, I designed a screen that's tailored to close evaluation of Beneish's metrics and my own revised metrics. It's here: https://www.portfolio123.com/app/screen/summary/284349?st=1&mt=1. The first half uses Beneish's actual methods instead of the simplified versions that most people (including P123) use. The second half uses my revision of Beneish's score, as detailed in my article. NVDA scores very high on both measures, indicating a high likelihood of manipulation.

So the financial data definitely supports your case. I also think you made a very good argument about their sales figures.

1 Like

Yes, as many have said, it is true that this time is different, and it is true that AI is different from the Internet, but in the exact opposite sense: there was a real telecoms infrastructure investment frenzy in the last Internet bubble, but Cisco still only became the world's number one stock by market capitalisation for the first time in March 2000, after the Internet bubble was heading towards bursting. And Nvidia has already briefly been the #1 market cap stock in the absence of a real AI infrastructure investment frenzy (and probably never will be one).

As with Bitcoin, which gained a market cap of over a trillion dollars as a value/productivity-destroying machine, "this time is different." Yes, the current and possibly upcoming AI bubble is even more false, crazy, value/productivity-destorying and divorced from fundamentals than it was during the dot-com bubble, e.g., many now even consider Nvidia to be an undervalued stock and the P123 system's picks of deep value stocks to be overvalued. I'm very happy to see this happening as it can only be very beneficial to factor investors. Thanks the market God for giving us more uninformed investors and dumb money!

Edit: