There has been a lot of talk about an artificial intelligence (AI) stock bubble, but I'm not buying it. Arguments about a bubble often go back to the Dotcom Bubble, claiming that there are similarities between that event and the current AI buzz.
Bears also point out that most of the spending comes from a few hyperscalers, but even that assertion requires a deeper look. Here's why I believe the AI rally still has room to run and isn't a bubble that's about to burst.
Image source: Getty Images
The AI rally is fueled by solid fundamentals and multi-year deals
The Dotcom era captured people's imaginations just as artificial intelligence does now. However, the dotcom bubble was also a time when a company could add ".com" to its name and immediately see a spike in its stock price, regardless of its fundamentals.
It's impossible to deny that some of that is going on in the AI industry. Sustainable sneaker company Allbirds rebranded to Smartbird, which is now a company that buys AI hardware and leases it to customers. The stock gained almost 600% in a single day upon the pivot but has since shed all of those gains.
However, Smartbird is a small player that's easy to forget in a burgeoning industry. Micron is a much larger AI hardware company that is valued above $1 trillion. Its revenue more than quadrupled year-over-year in its fiscal 2026 third quarter. Investors will be hard-pressed to find that type of growth anywhere else, but that's not the most exciting part about Micron's results.
Micron has been securing multi-year customer deals that "significantly enhance the durability and predictability of Micron's strong financial performance." It's not just Micron. Sandisk also announced that it has been securing multi-year commitments that include financial guarantees from customers.
These long-term deals minimize the risk of a cyclical bust, which is normal for chipmakers. AI demand can break that cycle, especially as demand heats up over multiple years.
Physical AI can be bigger than current AI tools
AI models and agents are the main focus right now. Models like ChatGPT, Claude, and Gemini require tremendous amounts of compute to process every query. That bodes well for AI data center operators and constructors. It also means that components that are required for each data center will gain demand.
However, the push to physical AI is just starting, and it can accelerate rapidly once it gains momentum. The two main physical AI catalysts right now are humanoid robots and self-driving vehicles. Autonomous cars have become more common in cities. Alphabet's Waymo is the market leader in the industry, and it has served more than 20 million rides. Waymo also claims that its vehicles are more than 10 times safer than an average human driver.
Tesla is still working on its Optimus humanoid robots, which can perform various tasks. China hosted its second annual World Humanoid Robot Games, which included a robot that beat Usain Bolt's 100-meter sprint world record.
The technology hasn't been perfected yet, but these competitions hint at what is possible within the next 5-10 years. AI data centers, memory chips, and other key components power these robots. The more robots countries produce, and consumers purchase, the stronger AI demand will become.
It's not just the hyperscalers
A major bearish argument is that hyperscalers do most of the spending. Analysts look carefully when tech giants like Microsoft and Amazon report earnings because of the capital expenditures figure. If these companies say they will cut back on AI spending, the theory is that AI stocks will crash.
First, that's not happening anytime soon. Amazon raised its annual capital expenditures forecast from $200 billion to $220 billion, citing higher memory costs. The company doesn't care that Micron and others are charging more for their chips. Amazon is still committed to making those orders. That type of demand doesn't go away quickly.
AI data center providers like Iren aren't just working with hyperscalers. The company recently emphasized its deals with leading AI developers rather than hyperscalers. Tech giants still do most of the spending, but the technology is becoming more accessible and affordable for smaller start-ups. That will push the demand for compute higher, resulting in all of the AI inputs becoming more valuable.
That doesn't sound like a bubble to me.
