Steering Clear of Memory Stocks: Cathie Wood’s Unconventional Take on AI Boom
In a recent podcast appearance, Cathie Wood, CEO of Ark Invest, detailed why her firm is avoiding memory chip giants like SK Hynix and Micron Technology. Contrary to popular belief, she argues that the soaring prices of high-bandwidth memory driven by AI demand are a warning sign, not an investment opportunity.
The Red Flag in Price Spikes: Cyclicality and Commoditization
Wood describes high-bandwidth memory as the “most cyclical and commoditizable” segment of the semiconductor supply chain. She views extreme price surges—sometimes tripling or increasing tenfold—as evidence of a distorted market. Such volatility underscores the segment’s inherent vulnerability: it’s hard to maintain product differentiation, leaving prices at the mercy of supply and demand swings.
“When a commodity experiences that kind of explosive price move, it often signals an impending inflection point or a severe supply chain bottleneck,” Wood noted. “For investors, that should be read as a cautionary signal, not a green light to buy.”
Betting on the Next Wave: AI Chips That Bypass Memory Dependency
Instead, Ark Invest is placing bets on companies redesigning AI hardware from the ground up. Startups like Cerebras Systems and Groq are pioneering chip architectures that minimize or eliminate the need for vast amounts of external high-bandwidth memory.
Wood drew a parallel to another industry: “It’s similar to Tesla’s efforts to design cobalt out of its batteries. The goal isn’t just cost savings—it’s to bypass the most volatile and constrained links in the supply chain.” She believes the AI chip industry is undergoing a similar shift, using architectural innovation to reduce reliance on expensive, cyclical memory components.
The Bigger Picture: Investing Beyond the Cycle
This investment thesis stems from Wood’s long-held conviction that sustainable value is created by breakthrough innovation, not by trading cyclical commodities. The current memory shortage and price hikes, in her view, are merely accelerating the search for alternatives.
For market participants, the implication is clear: focus may need to shift from “who sells the scarce resource” to “who rewrites the rules of computation.” The real battleground in the AI hardware revolution might be moving from memory manufacturing to fundamental chip architecture and system design.