Arm's Hermann Hauser warns investors AI prices have run ahead of fundamentals

Hermann Hauser, who co-founded Acorn Computers and helped spin the chip designer Arm out of it, told CNBC's Tech Download that investors are conflating two questions they should hold apart. One is whether the AI revolution is real. The other is whether today's prices are. His answer to the first is emphatic, and to the second sceptical.

That framing does something most commentators avoid. It separates the technology from its valuation and answers the two in opposite directions. Hauser called AI a revolution larger than any before it, then warned that some share prices have moved far ahead of what the underlying businesses can support. The ride, in his own framing, will resemble a rollercoaster, real value underneath and violent swings on top.

His track record gives the warning weight. Hauser co-founded Acorn in 1978 with Chris Curry, helped spin Arm out of it in 1990, and now backs European deep-tech companies through Amadeus Capital Partners. Few investors have watched as many technology waves rise and break at close range.

Two questions investors keep collapsing into one

Most AI bulls and bears argue over a single point, whether the boom is justified. Hauser splits that into two. On the reality of the revolution he is emphatic, calling it larger than any technology wave before it. On the reality of current prices he is skeptical. Answering the two together, he suggested, is how investors talk themselves into paying any price for a genuine breakthrough. The gains behind that temptation have been extraordinary.

Those returns are why research houses have grown cautious. Morningstar's Lorraine Tan said in July that the highest-flying AI and memory chip stocks could give back 20 to 30 percent of their value before they looked attractive again, and that the extraordinary revenue driving the rally would level off as supply caught up. The names that led the market higher are the ones most exposed. So far they have gone the other way: the memory stocks rallied hard again this week.

Micron illustrated the tension. It reported record revenue of $41.46 billion for the quarter to May 28 and raised its dividend by 30 percent, to 15 cents a share, yet Morningstar reads those results as a high point rather than a floor. Micron spent nothing on buybacks that quarter, after $650 million over the previous two, and told investors repurchases would be the main way it returns capital from here.

On the hardware itself, Hauser described an industry straining against physical limits. AI is expensive to run, chips are difficult to cool, memory is costly, and bottlenecks sit across the supply chain. Those constraints, he argued, are forcing a rethink of how computers are built, because the current design spends too much energy simply moving data around.

Where the hardware goes next

Two emerging approaches drew Hauser's attention. In-memory computing and photonic computing both aim to cut the energy spent shuttling data between processors and memory, the step that makes today's systems so power hungry. He argued the payoff from those shifts could rival earlier breakthroughs in chip architecture, changes large enough to reset which companies lead the industry.

His caution extended beyond the current cycle. Hauser said quantum computing could reach a point he calls Q Day, when a machine can break the encryption that protects online communication, as early as 2029. That timeline matters to any business whose value rests on data staying private.

The technology will still pay off, in his telling. The open question is who collects that payoff, and at what price today's buyers are entering.

The fragile part is not the top

Hauser's most specific advice concerns where a correction would land. The largest labs, he said, are not the vulnerable ones. He named OpenAI and Anthropic as holding sizable capital reserves, enough to withstand turbulence even if expectations reset around them. Size and funding, in his read, are what let a company outlast a period when investors stop paying for promises.

This is a revolution that will create more value than probably any other technology revolution that we've ever seen.

Hauser does not say outright who absorbs a reset. What he points to is the circular financing running through the sector, deals in which the same money moves between suppliers and customers. On that reading the exposure sits in the tiers below the leading labs, among the AI-adjacent businesses and infrastructure plays whose valuations depend on growth staying exponential, rather than at the top.

That stance puts Hauser at odds with some of AI's loudest champions. Masayoshi Son of SoftBank told SoftBank's annual meeting on June 24, 2026 that calling AI a bubble is an insult, and at the same meeting said it would be blasphemy against AI to call it one. Hauser is explicitly more cautious, without joining the outright bears who expect the whole structure to fall. His middle position is harder to caricature and, for that reason, harder to dismiss.

He carried the argument to Europe as well. European technology risks becoming what he called a technology colony of the United States, Hauser said, unless the region supplies its startups with growth capital and experienced managers able to scale them. The warning ties his investing to a broader worry about who owns the next platform.

His position reduces to three claims held at once. The revolution is real. The prices are not all real. And the biggest names are not the fragile part. Each cuts against a different piece of the consensus that carried markets to records, and together they describe a boom that survives without every ticker inside it surviving.

A chorus of cautious voices

Hauser is not speaking into a vacuum. Through 2026 a string of investors and researchers with long records have told clients and audiences to reexamine how much of a portfolio now rides on a single theme. Their warnings differ in tone and timing, but they converge on one point, the danger of concentration in a market where a handful of stocks set the direction for everyone else.

What other investors are telling clients

Jeremy Grantham, co-founder of GMO, told CNBC's Squawk Box on June 26, 2026 that AI enthusiasm had carried U.S. equities to the most expensive valuations in American history. He pointed to the ratio of total market value to gross domestic product, the so-called Buffett indicator, which had climbed to 235%, roughly two and a half times the size of the economy.

Scott Galloway, the NYU professor, puts about 40% of the S&P 500 in AI-focused businesses and argues that leaves investors with real concentration risk. He has said founders fear an economic crash within 12 to 24 months from overinvestment, and has flagged cheap Chinese models undercutting US firms, a practice he calls AI dumping. Ray Dalio, who founded Bridgewater Associates, posted on January 8, 2026 that the AI boom was in the early stages of a bubble, and urged investors to brace for what follows.

The market itself has started sorting. Memory makers surged while some earlier favorites, including Oracle, faced sharper scrutiny as investors demanded returns they could see now. One worry that has spread is circular financing, where investors help fund the very revenue they later count, an arrangement critics call structurally unstable. Nvidia, AI's defining stock, is up only about 4% this year, a sign that even the biggest name has cooled from earlier highs.

Dimon draws a dot-com parallel

Jamie Dimon added the loudest voice from banking. The JPMorgan Chase chief executive said on August 12, 2026, on the Master Investor Podcast, that the stock market felt a little too complacent about risks, and he compared the AI boom directly to the dot-com cycle of the late 1990s. His bank is itself a heavy investor in and user of the technology, which sharpened the warning.

When I look at AI itself, the amount of money being spent is huge. Will it pay off in total? Probably, just like the internet did. Will it pay off the way you expect and the timetable you expect? Definitely not.

Dimon noted that before the internet's long-term winners emerged, darlings such as Yahoo and Netscape went bankrupt or faded. The lesson he drew for AI is that a transformative technology and a surviving stock are not the same thing, and many richly valued companies today will not clear the shakeout.

A separate warning came from inside the research itself. Yoshua Bengio, a Turing Award winner and one of deep learning's founding figures, told Bloomberg in June 2026 that the field is building systems it does not yet know how to control, and that heavy financial incentives are pushing companies to move faster than safety allows. He has since launched LawZero, a non-profit lab, to develop AI designed to understand the world without pursuing its own goals.

For investors, the practical response most of these voices reach for is the same. Spread exposure rather than concentrate it. Some analysts point to broad market funds such as the Vanguard Total Stock Market ETF as a way to hold the theme without betting the portfolio on a handful of names. Max Wasserman of Miramar Capital put the discipline plainly to Yahoo Finance, saying investors need to understand what they own and what happens if they are wrong.

Hauser's contribution is the sorting logic beneath that advice. A correction, if it arrives, need not fall evenly. The labs with cash and the firms building real value should endure, while the stretched middle absorbs the reset. Whether investors can tell those two groups apart before the next earnings season settles the debate is the test his warning leaves behind.