Nvidia Stock: It's Time To Take Profits (NASDAQ:NVDA) (2024)

In my last article in this space on March 15th, I recommended that investors sell Super Micro Computer (SMCI) and buy Nvidia Corporation (NASDAQ:NVDA) for those bullish on the AI investment theme. At the time, SMCI was trading at a higher multiple than NVDA, something that was rare historically. I did not believe SMCI deserved a higher multiple because NVDA has a far deeper moat. That call worked out well, and anyone who sold $1000 worth of SMCI at the time now has $1376 of NVDA instead of $720 worth of SMCI (nearly a double.)

Here's my next call: you should take profits on NVDA right now. I think the current situation has all the hallmarks of a blow-off top, and I think shares could trade down 25% over the next month (which would bring us all the way back from a valuation of 3 weeks ago) and more by year-end.

I think this current rally is being driven primarily by the 10:1 stock split, rather than fundamentals, and this week could turn into a "sell the news" event.

History doesn't repeat, but it often rhymes

The year is 2000, the Internet is going to change the world, and Cisco (CSCO) is the quintessential "picks and shovels" play with a deep technological moat.

Passing Microsoft in net worth on March 28th, 2000, after tripling over the past year to a valuation of over $500 billion, with predictions of it becoming the first trillion-dollar company.

Now replace all of the above with Nvidia today but with an additional trailing zero on the valuation target. Tripled in under a year, check. Wildly bullish targets, check. A non-recurring hardware business model prone to boom/bust cycles, check.

There are absolutely differences. Cisco traded at nearly a 200x P/E at its peak, versus 45x for Nvidia now. But Cisco grew revenue steadily over a long period, rather than having an incredible 4 quarter parabolic ramp like Nvidia has had.

Cisco actually kept growing well for the next 5 years after the peak, but once the hyper growth period ended and more competition emerged, Cisco's P/E declined from its 200x peak to below 20 by 2005.

I believe something similar could happen here, with Nvidia Data Center revenue peaking and then declining into a steady state far below current levels.

Nvidia Revenue Durability and Valuation

Nvidia is now trading at 45x 2025 EPS, compared to tech peers Microsoft (MSFT) at 35x, Apple (AAPL) at 30x, and Google (GOOG) at 23x. Is this justified by Nvidia's growth? That's the bull case, although Nvidia has the least predictable revenue of the major technology players in my view.

Nvidia has a history of changing revenue direction rapidly, like in 2022 when the crypto bubble went into reverse, and Nvidia saw revenue decline 29% in just 2 quarters. (Nvidia's FY23=CY22.)

Now look at the last 12 months of revenue data for Nvidia, with Data Center revenue being up 6x. A few things I think when I look at this data.

• ChatGPT was released to the public on November 30th, 2022. AI has been a thing for a while, but a publicly released product spurred a ton of interest in it. "AI Leaders" are given premium valuations. Who wouldn't be investing in AI here?
• The Q1-25 print for NVDA and Q2-25 guide was largely in line with most expectations. In fact, the stock was flat/up 1-2% for a while after the earnings were released. I believe most of the 33% post-earnings run was due to the stock split and not the earnings themselves.
• The decrease in Gaming and Professional Visualization revenue. Do not let the titles fool you, the GeForce Gaming cards and Quadro Professional Visualization cards are used for a lot more than gaming and rendering. There is a lot of AI work/execution that gets done on these cards, although memory limitations prevent training on massive data sets. The cloud GPU providers even sell time on the more expensive Quadro cards. While it's only a single data point, I believe stagnation or declines in these cards may provide an early read on true end user demand, versus the larger Tech companies building out AI Cloud capacity in hopes to resell it later.

I think investors who are taking the recent revenue trends and extrapolating many years into the future are making a mistake.

Reseller Demand

Nvidia's largest customers are driving the majority of the Data Center revenue. Amazon, Meta, Microsoft and Alphabet are estimated to account for 40% of Nvidia's revenue. There is also a significant number of sales from newer resellers like CoreWeave, Lambda Labs, and others. This demand isn't (primarily) for their own use, but to resell to other companies.

Currently, the cloud providers are overbuilding capacity, and for good reason. To carry the premium pricing over the smaller players, they need to guarantee availability of GPU resources when a customer wants it, so this is a bit of a land grab for the cloud providers. But this overbuilding will not last forever.

These companies are also developing their own hardware platforms, and while initially targeted to handle internal workloads, eventually these will be resold to other cost-conscious customers. Nvidia's margins are just too great right now to be maintained for more than a few years, in my opinion, and large customers will adapt their code to ASIC's or other specialized products like Google TPU's for savings. Nvidia has a moat with CUDA, its proprietary API, but this moat is far from impregnable. It just becomes a math problem eventually: pay the premium from Nvidia GPU processing, or port the code to another platform.

Even before this, any macroeconomic pullback or a shift in investor sentiment around spending money on AI without tangible results, and we will see companies "right sizing" their AI investments and conserving cash.

Data Center revenue moving from $4.3 billion to $22.5 billion in 4 quarters is simply too much too fast, and it could go in reverse quickly as well.

My Views on AI

I've mentioned this in articles occasionally, but my day job is in technology, not finance. Specifically, I've worked with Machine Learning and Deep Learning/AI projects within the medical industry for a very long time, which colors a lot of my thinking on the topic.

AI has been a big topic in medicine for a while now, and GPU's are certainly driving innovation there. But results have been a very mixed bag, and there are very real, practical challenges in getting AI to actually produce something useful.

IBM's (IBM) much heralded and hyped Watson AI super project has been a disappointment and has quietly been shut down over the past few years before being sold off, after failing in many real-world implementations.

Another excellent example of the challenges in real-world AI can be seen in publicly traded AI pure-play iCAD (ICAD). iCAD makes a product, ProfoundAI, that uses AI on mammograms to help Radiologists detect Breast Cancer. Using AI to locate breast cancer is a hugely attractive application:

• Large screening population
• Many images to look at (especially with Tomosynthesis/3D Mammography)
• Single organ
• Binary question to answer (breast cancer present: Yes/No.)

Yet, while it is a clear improvement over past machine learning-based techniques, it still struggles to outperform radiologists. At its core, the problem is one of data: it's only going to be as good as the "ground truth" data set, which generally means that anything a human missed, AI will miss, since it wasn't "trained" on it.

This data limitation is what I see with services like ChatGPT and Google Gemini. I've found these services good at summarizing data I give them, but nearly useless for asking questions, as it often gives very confident yet wrong answers, and that's before considering any ideological filters/biases present.

Nvidia 2024 versus Cisco 2000

Of course, we know Nvidia will report higher sales and income for the next few quarters, probably the next 3-4 at least. That doesn't mean the stock will be higher. Let's see what happened with Cisco in 2000, which to me feels very much like Nvidia today.

• Cisco shares peaked in March 2000.
• Cisco May 2000 earnings, revenue rose 53%, net income up 42%.
• Cisco August 2000 earnings, revenue rose 61%, net income up 69%year over year.
• Cisco November 2000 earnings, revenue rose 66%, net income rose 67%. It looks like earnings are still accelerating here!
• Cisco's first true "miss" didn't happen until February 2001, with net income still up 50% year over year. Revenue missed and Guidance was poor (only predicting 40% growth, versus the 55% the street was expecting.)

By the time the actual miss came, shares had already been cut in half, and then got cut in half again. Cisco started from a far higher multiple, near 200x at its peak; so I'm not predicting Nvidia declines 75%. But I absolutely think that we could see Nvidia peak revenue in 3–4 quarters, and decline in quarters beyond that, and that shares could be 50% lower from here if that happens.

Conclusion

AI is hot right now, undoubtedly. Other than Nvidia and other companies selling the hardware and cloud services, I have a tough time finding companies that are really winning through AI, especially those that have been investing in it before the recent hype.

• Lemonade (LMND) the AI-powered insurance company is still down 90% from its 2021 high and still losing money. There is no evidence that AI is doing anything positive for them.
• Meta said it would "expect to see a multi-year investment cycle before we fully scaled Meta AI, business AIs and more into the profitable services"
• Previously mentioned medical companies, including IBM's failures in this space.

Ultimately, I think companies are overinvesting in AI right now largely because they believe that will drive their stock prices and because that's what their competitors are doing. Any company that outright said, "we believe AI is a lot of hype, and we're choosing to invest in other things" would likely see their share price punished.

I still do not see enough AI-enabled products that are truly transformative to believe this level of investment is sustainable, and believe we are just a few quarters away from this topping. In short, I just do not believe Nvidia's sales and margins at these levels are sustainable.

I think there is an excellent chance that this week marks the peak for Nvidia, and now is a good time to take some profits.

Editor's Note: This article covers one or more microcap stocks. Please be aware of the risks associated with these stocks.

Fishtown Capital