Dan Niles: We’re in an AI Bubble, But It's Not Over Yet
Legendary tech investor Dan Niles explains why the AI boom is still early
By Zack Guzman
May 29, 2026
The AI trade is starting to look eerily similar to the dot-com boom and bust for some investors old enough to remember.
Semiconductor stocks are ripping higher. Capital expenditures are exploding. Wall Street analysts are racing to raise targets. And even as oil prices surge, bond yields climb, and inflation fears return, the market continues to push higher anyway.
For veteran tech investor Dan Niles, that’s not necessarily a contradiction. In fact, he thinks it’s exactly what bubbles look like.
“We’re in a bubble,” Niles told Coinage in a new interview. “But it doesn’t mean it can’t keep going up further.”
Niles, the founder of Niles Investment Management, believes the current AI boom mirrors the late 1990s internet infrastructure buildout more closely than most investors realize. And despite growing concerns that the rally has become overheated, he argues the market may still be early in the cycle.
“I think we have at least a year to go,” Niles said.
The comparison matters because the setup today looks surprisingly similar to the years before the dot-com crash. Back in 1997 and 1998, macroeconomic shocks repeatedly rattled markets. But underneath the volatility, internet infrastructure spending continued accelerating.
“The S&P was down 11% intra-year in 1997,” Niles said. “It was down 19% intra-year in 1998. But underneath that, you had this internet infrastructure buildout going on.”
By year-end, the market had recovered both times, finishing up 31% in 1997 and 27% in 1998. Niles believes AI is following the same playbook. The difference, according to him, is that the technology itself may be even more transformative this time around.
Read More from Coinage:
What Strategy Promoting An 11.5% Dividend Means for Bitcoin (and Michael Saylor)
He stole $200 million. He gave it back. Now, he’s ready to explain why
“What you saw in January of this year is you had OpenClaw come out and be formalized, and that kicked off agentic AI,” he said.
Niles argues the shift from simple chatbot interactions to autonomous AI agents fundamentally changes the economics of compute demand. Instead of asking ChatGPT a single question, users are increasingly tasking AI systems with coordinating workflows, analyzing documents, gathering research, building spreadsheets, and generating presentations simultaneously.
“All of a sudden you’re using 10 to 100 times more tokens,” Niles said.
The result has been a historic rally in semiconductor stocks. As the Wall Street Journal recently highlighted, the PHLX Semiconductor Index is posting its best start to a year on record, up 82% so far in 2026. Only 1995 came close, in what was the early indicator of the tech shift heading into the dot-com boom.
But even as Niles remains bullish near-term, he’s explicit that the current environment has all the hallmarks of a bubble. The problem, he argues, is that transformative technologies inevitably attract too much capital.
“You have 100 companies that each think they’re going to get 50% market share,” Niles said. “That’s not going to work.” Eventually, winners and losers emerge. And when that happens, the market’s enthusiasm can unwind violently.
“I fully expect sometime next year we’re going to be talking about why certain sectors like semiconductors are down 30% to 50%,” he said.
But for now, Niles thinks the underlying demand story remains too powerful to ignore.
One of the more important shifts he sees happening beneath the surface is the changing architecture of AI infrastructure itself. Earlier AI systems relied overwhelmingly on GPUs. But agentic AI increasingly requires orchestration — coordinating multiple tasks and processes simultaneously — which boosts demand for CPUs as well.
“If you go from 8-to-1 GPUs-to-CPUs to 1-to-1 GPUs-to-CPUs,” Niles said, “some of these historical microprocessor vendors should do pretty well.”
That’s part of why Niles remains constructive not just on Nvidia, but on broader semiconductor infrastructure, including optical networking and names like Intel and Micron.
Still, that doesn't mean volatility can't hit. Niles has recently advocated holding more cash and becoming more tactical as the rally grows increasingly speculative.
“There’s nothing wrong investing in a bubble,” he said. “But make sure you understand you’re in one.”
For now, though, he believes the momentum behind agentic AI remains strong enough to keep the trade alive — even if the ending eventually looks painfully familiar.
