Nvidia (NASDAQ:NVDA), a company that designs semiconductors, needs no introduction. It has probably been the most talked about stock on Wall Street and recently surpassed Apple (NASDAQ:AAPL) in market cap. I believe the AI bubble, which extends to Nvidia’s cloud computing customers and a plethora of unprofitable large language models, could soon pop. As competition becomes viable and demand inevitably cools, Nvidia’s profitability will likely collapse. Thus, I am bearish on NVDA stock, and so is a well-known billionaire investor.
Don't Miss Our Christmas Offers:
- Discover the latest stocks recommended by top Wall Street analysts, all in one place with Analyst Top Stocks
- Make smarter investments with weekly expert stock picks from the Smart Investor Newsletter
Why Nvidia’s Profitability Can Collapse
On July 9, 2022, I concluded publicly that Tesla (NASDAQ:TSLA) was the “last bubble standing” and warned that margins would likely fall amid increasing competition and a cooling economy. Since then, Tesla’s operating income has collapsed along with its stock.
What if I told you that the same thing could happen to Nvidia? Investors are confusing a cyclical business for an extremely deep moat, perpetual growth story. But in semiconductor design, you’re only as good as your next product. You have to constantly race against competitors and compete for the best employees. Just look at what happened to the once-dominant Intel (NASDAQ:INTC).
Nvidia’s sales have soared due to the introduction of large language models (LLMs), which require a lot of computing power. The problem is, like dot com companies in the early 2000’s, most LLM’s are not profitable.
Meanwhile, Nvidia’s primary customers, cloud computing companies, are excessively profitable and have to compete on offerings and price. AWS, Microsoft Azure, and Google Cloud look like telecom companies in the late 1990s. Their margins and returns on assets will likely suffer as the industry matures, which could hurt their semiconductor suppliers (Nvidia).
Competition is rushing into the semiconductor space. I’m seeing excessive investment by Intel and other players, viable and cheaper AI chips from AMD (NASDAQ:AMD), and new entrants in China and big tech. AMD recently said its chips are being used to power ChatGPT. The firm stated, “The AMD Instinct MI300X and ROCm software stack is powering the Azure OpenAI Chat GPT 3.5 and 4 services.”
Meanwhile, both Nvidia and its big tech customers seem to be over-earning with excessive margins and returns on assets. Nvidia currently has a silly high return on assets of 55% and a silly high operating margin of 60%. Capitalism simply does not support this level of profitability. Profitability like this attracts competition, which will, in turn, cut these returns and margins in half, or worse, in my opinion.
I also expect that demand will cool in the generative AI space. Over the past year, companies have been absolutely throwing cash at everything AI, often having no concrete visibility into the returns on these investments. The LLM losers will eventually drop out of the space as cash burn becomes too much or as they lose market share. Additionally, Nvidia’s orders can be canceled quickly if there is a recession or tech downturn like there was from 2000-2003.
Why Nvidia’s Blackwell Is Overpriced
Nvidia recently released Blackwell, a GPU that is expected to sell for $30,000 to $40,000. However, Blackwell seems to primarily increase the speed of training and delivery for large language models, with little to no increase in the quality of answers produced. In my opinion, LLMs like Perplexity (an AI search engine) don’t need to be any faster than they already are. Therefore, I don’t see a use case strong enough to justify Blackwell’s massive price tag.
Moreover, Nvidia is currently engaging in price-gouging, driven by its first-mover advantage and robust demand. To return to my previous analogy, Tesla was doing the exact same thing in 2020, 2021, and 2022.
Mark my words on this one (in full disclosure, this is an informed guess): the price of AI GPUs will fall over the next three, five, and 10 years. In the context of history, the price of DRAM chips collapsed at a rate of 48 percent per annum from 1994 to 2001, and the price of MPU chips collapsed at a rate of 63 percent per annum over the same period. This was despite increasing demand.
Is NVDA Stock a Buy, According to Analysts?
Currently, 37 out of 40 analysts covering NVDA give it a Buy rating, three rate it a Hold, and zero analysts rate it a Sell, resulting in a Strong Buy consensus rating. The average Nvidia stock price target is $126.82, implying upside potential of 4.9%. Analyst price targets range from a low of $90 per share to a high of $150 per share.
Why Stanley Druckenmiller Sold Nvidia and Thinks AI Is “Overhyped”
In my opinion, the best “analyst” on Nvidia is legendary investor Stanley Druckenmiller and his team. Druckenmiller, who famously got burned in the dot-com bubble and learned a lesson or two, recently detailed on CNBC why he sold his Nvidia shares after a substantial profit. He said much of what he originally saw has now been realized by the market.
Advocating caution on AI, Druckenmiller said the following: “If you bought the Nasdaq in ’99, it went down 80% before that all came to fruition. That’s not going to happen with AI, but it could rhyme. AI could rhyme with the internet. As we go through all this capital spending we need to do, the payoff, while it’s incrementally coming in by the day… the big payoff might be 4 to 5 years from now. So, AI might be a little overhyped now, but underhyped long-term.”
Given Druckenmiller’s actions in selling Nvidia but also praising Perplexity, I believe Druckenmiller thinks the big opportunity is in AI search, not AI chips. If one of the large language models can displace Google Search in four to five years, that could be huge. Guessing which one, and whether or not it will belong to Google itself, could be more difficult. There will be a lot of losers.
The Bottom Line on Nvidia Stock
I believe Nvidia’s return on assets and margins will likely collapse alongside the prices of its AI semiconductors. One needs to look no further than history to see this coming. Semiconductor prices collapsed dramatically in the 1990s despite increasing demand. Moreover, Blackwell is likely overpriced, and AMD’s chips are being used for ChatGPT. Semiconductor companies are cyclical and only as good as their next product and their employees; they should be valued as such.
