Approximately 30 years ago, the advent and proliferation of the internet began changing corporate America for the better. The internet represented a significant leap forward for businesses, opening new sales channels and providing marketing opportunities that hadn't previously existed. It also paved the way for the retail investor revolution.
For three decades, investors have watched several hyped trends with sizable addressable markets come and go; but none have come close to matching the impact of the internet... until now.
Artificial intelligence (AI) appears to be the long-awaited answer to what comes next after the internet. Empowering software and systems to make split-second decisions, and to become more efficient at their assigned tasks, all without human oversight, is a game-changing technology that the analysts at PwC believe will add $15.7 trillion to the global economy by 2030. This otherworldly growth potential is why shares of trillion-dollar AI stocks Nvidia (NVDA 1.04%) and Broadcom (AVGO 0.25%) have skyrocketed by 1,140% and 583%, respectively, since the start of 2023.
Image source: Getty Images.
But while growth estimates for Wall Street's leading AI stocks are robust, several historical headwinds point to the artificial intelligence bubble bursting in 2026.
Next-big-thing technologies have a checkered past
To preface this discussion, understand that no data point or correlated event can guarantee, with 100% accuracy, what's going to happen with a specific stock or hyped trend. Nevertheless, certain correlations are hard to ignore on Wall Street -- especially those pertaining to next-big-thing trends.
Over the last three decades, investors have been privy to a number of hyped technology trends, including the internet, genome decoding, nanotechnology, blockchain technology, the metaverse, and now AI and quantum computing. These next-big-thing trends all share two things in common:
- Pie-in-the-sky addressable opportunities that get investors excited.
- The need for the underlying technology to mature and evolve.
Unfortunately, investors have a terrible habit of focusing on the first common trait and overlook the fact that innovations require ample time to develop. For 30 years, investors have persistently overestimated the adoption, utility, and/or optimization of new technologies, resulting in lofty expectations not being met.
NASDAQ: NVDA
Key Data Points
On one hand, optimists can argue that the insatiable demand for Nvidia's graphics processing units (GPUs) -- these are the brains of AI-accelerated data centers -- and Broadcom's AI-networking solutions demonstrate that we're not in a bubble. But dig beneath these headline sales figures, and you'll discover that most businesses spending big bucks on AI infrastructure aren't close to optimizing this technology or generating a positive return on their investment.
Although it's not impossible for AI to avoid a bubble-bursting event early in its expansion, history tells us that no hyped technology for at least 30 years has sidestepped this fate.
Valuation tells the tale
Unfortunately, historical precedent comes into play more than once for the AI revolution.
While value is subjective -- what you find to be pricey might be viewed as a bargain by another investor -- history shows that arbitrary lines in the sand can be drawn that help establish the probability of a bubble forming and bursting.
For example, in the year leading up to the bursting of the dot-com bubble, several leading internet companies, including Amazon, Microsoft, and Cisco Systems, peaked at trailing-12-month price-to-sales (P/S) ratios ranging from 31 to 43, respectively. These three stocks all fell by roughly 75% to 90% on a peak-to-trough basis after the dot-com bubble popped.
A P/S ratio of 30 has historically proved to be unsustainable for megacap companies leading the charge of next-big-thing technologies. Though the P/S ratio doesn't help determine when the music might stop for a hyped technology or the stocks behind it, it does foreshadow trouble to come.
NASDAQ: AVGO
Key Data Points
Nvidia, the face of the AI revolution, once again surpassed a P/S ratio above 30 in early November. Meanwhile, Broadcom's P/S ratio peaked at nearly 33 last week, and AI-data mining specialist Palantir Technologies sports a P/S ratio of 112, as of the closing bell on Dec. 2.
Even with sustained double-digit annual sales growth rates, the valuations for these AI market leaders can't be historically justified.
To add to the above, we're set to enter 2026 with the second priciest stock market on record, when back-tested 155 years. If a significant correction ensues, growth stocks that trade at aggressive premiums, such as Nvidia, Broadcom, and Palantir, would likely be among the hardest hit.
Image source: Getty Images.
Competitive pressures can pull the plug
The final piece of the puzzle that can send this near-parabolic AI rally over the proverbial cliff is mounting competitive pressure on the businesses leading the charge.
Nvidia's operating results suggest it's had little competition to speak of. It's netting $30,000 to $40,000 for each of its high-end GPUs, which represents a decisive premium over the selling price of competing chips used in enterprise data centers. Its Hopper (H100), Blackwell, and Blackwell Ultra GPUs also have no external competition that's come close to matching their compute abilities.
The bulk of Nvidia's pricing power has resulted from the persistent scarcity of AI-GPUs. As long as demand for AI chips sizably outpaces their supply, Nvidia should enjoy strong pricing power and a gross margin above 70%.
If all Nvidia had to concern itself with was the possibility of external GPU companies trying to chip away at its market share (pun fully intended), there wouldn't be much to worry about. But there is, indeed, more.
Many of Nvidia's top customers by net sales are also internally developing AI-GPUs or AI solutions of their own to use in their data centers. Even though these chips and solutions aren't a threat to Nvidia's compute superiority, they're notably cheaper than Nvidia's hardware and will likely be more readily available, considering most of Nvidia's GPUs are backordered.
In other words, the stage is set for these alternative chips and solutions to effectively steal valuable data center real estate from Nvidia and significantly reduce the AI-GPU scarcity that's fueled its otherworldly pricing power and high gross margin.
If the lofty expectations of AI market leaders become unattainable, it can pave the way for the AI bubble to burst in 2026.
