The decline of Wall Street’s artificial intelligence (AI) darling may only be just beginning.
For the better part of the last 30 years, investors have focused on the next big thing: technology, innovation, or trend.
Since the internet completely changed the growth trajectory of companies in the mid-1990s, we have witnessed huge market allocations in genome sequencing, business-to-business commerce, Chinese stocks, US housing, nanotechnology, 3D printing, cryptocurrency, blockchain technology, cannabis, augmented reality/virtual reality, and the metaverse. But the next big trend that could be the best of the bunch is Artificial Intelligence (AI).
According to an in-depth report published by analysts at PwC, “Sizing the Prize,” AI could benefit the global economy by $15.7 trillion by the end of the decade. These gains are expected to come in the form of increased productivity and a range of consumer benefits.
The $15.7 trillion opportunity leaves plenty of room for multiple companies to emerge as winners, perhaps none more so than semiconductor giant NVIDIA. (NVDA 8.15%).
Image source: Getty Images.
AI revolution drives Nvidia’s historic breakthrough
As 2022 ended, NVIDIA was a $360 billion company and a marginal player in the technology industry. Less than 18 months later, fresh off a historic 1-for-10 stock split, the company’s valuation peaked at $3.46 trillion, briefly making it the most valuable public company in the world.
Never before in history has a market-leading company created over $3 trillion in value in less than 18 months, and this unprecedented movement has undoubtedly been driven by the AI revolution.
It didn’t take long for NVIDIA’s AI graphic processing units (GPUs) to become the first choice for companies wanting to run generative AI solutions and build/train large language models (LLMs). Orders are backlogged for the company’s H100 GPUs and the next generation of Blackwell chips, scheduled to debut in Q1 2025.
In addition to this, Nvidia’s CUDA platform plays a key role in keeping enterprise clients loyal to its ecosystem. CUDA is the toolkit that developers use to build LLMs and get the most out of Nvidia GPUs.
Controlling an estimated 98% of the GPU market for AI-accelerated data centers over the past two years has given Nvidia incredible pricing power: Companies are paying $10,000 to $15,000 for Advanced Micro Devices’ MI300X AI-GPU, compared to $30,000 to $40,000 for Nvidia’s H100. Premiums of 100 to 300% over competing chips have helped the company increase its gross margins by more than 10 percentage points over the past six quarters.
Despite this seemingly ideal execution, Nvidia shares are down 27% from their all-time high hit in June and could be headed for a steeper decline.
Nvidia may eventually lose three-quarters of its value
An eventual 75% drop from an all-time high of $140.76 probably sounds absurd given the high growth and profit projections Wall Street expects Nvidia to achieve over the next few years, but when it comes to the next big innovation, history has yet to be beaten.
While quite a few breakthrough technologies and trends have been huge successes for patient investors (e.g. the internet and business-to-business commerce), one thing that all new technologies, innovations and trends have in common is an early stage bubble.
While it’s impossible to predict exactly when the excitement around a potentially game-changing innovation will die down, one thing that remains consistent when it comes to Wall Street’s “next big thing” is that investors always overestimate how quickly a technology or trend will go mainstream.
Judging by the amount of money companies are spending on AI data centers, you might think artificial intelligence is on track for rapid adoption. But digging deeper reveals that most companies have no clear plan for how to adopt AI or monetize the technology for their own benefit. This is a clear sign that we are witnessing the next bubble forming. History shows that market leaders in the next big bubble have lost 75% to 99% of their value from peak to trough.
The only saving grace for Nvidia is that it had other fundamental business units in place before the craze for AI took hold: Even if the technology takes a few years to mature and sales fall off dramatically, Nvidia can rely on revenue from selling GPUs for gaming and cryptocurrency mining, as well as from its virtualization software division, to cushion some of its revenue and profits.
Image source: Getty Images.
Nvidia will also face fundamental pressures.
Beyond historical precedent, there are also fundamental reasons to believe that Nvidia’s luster will fade as external and internal competitive pressures grow.
As mentioned earlier, AMD’s MI300X is priced at a massive 50% to 75% lower than the selling price of the H100. With Nvidia’s massive backlog and the appeal of fairly cheap AI-GPUs, AMD and other Nvidia rivals shouldn’t have a hard time finding demand for their datacenter hardware.
What most investors fail to realize is the potential for internal competition.
The four “Magnificent Seven” companies account for roughly 40% of Nvidia’s net sales. While it’s great to see Nvidia as the go-to hardware supplier for some of the most influential companies in the U.S., investors should also understand that all four of these companies are also developing AI-GPUs in-house to bring to their own data centers. Even if Nvidia’s H100 and Blackwell chips maintain their computing dominance, the Magnificent Seven companies are incentivized to use these cheaper, more readily available chips developed in-house as a complement to their own data centers.
The real problem with increased internal competition is that it reduces the scarcity of AI-GPUs, which has been the driving force behind Nvidia’s extraordinary pricing power. As pricing power declines over time, the company’s gross margins are expected to decline.
Moreover, internal competition will reduce Nvidia’s chances of securing valuable data center real estate from major U.S. companies, suggesting that orders from the Magnificent Seven have likely peaked.
NVIDIA is also historically overpriced. Only a few times in the past 30 years have a company on the cusp of the next big innovation traded at a trailing 12-month price-to-sales multiple (P/S) of 40. Historically, these stocks have fallen about 90% from peak to trough.
Whether it’s historical precedent or the underlying roadmap, it seems likely that Nvidia will ultimately lose many of the gains it has made since the start of 2023.
