US memory chip stocks have lost almost $100bn in market value this week, hit by new Google research that pointed to an easing of the AI-driven hardware shortage that had sent chipmakers’ shares to record highs.
Shares in US memory maker Micron have shed more than $70bn in market capitalisation since last Friday’s close, down 15 per cent, amid a broad sell-off on Wall Street.
Sandisk, the maker of flash memory devices that was the best-performing stock in the S&P 500 last year, lost around $15bn in value during the week, while storage companies Western Digital and Seagate each lost billions.
“These stocks have had tremendous runs so it’s rational for any marginal news” to dent their shares, said Travis Prentice, chief investment officer at Informed Momentum Company, a California-based asset manager.
The memory stocks rally “doesn’t look like it’s over yet but expectations are high, so it makes sense to take some profits, especially in a troubled market environment”, he added.
Investors had been betting that a shortage of memory chips, which are vital components in the data centres that run advanced AI models, would continue well into next year.
This had made memory chip and storage providers the biggest Wall Street beneficiaries of AI this year, as Big Tech companies including Nvidia, Microsoft and Alphabet lost ground due to concerns about excessive capital expenditures.
However, the Google Research paper published this week has shaken investors’ confidence that AI will continue to demand so much storage capacity.
The search group’s TurboQuant algorithm promises to radically compress AI models without compromising the accuracy of their outputs, meaning they can be run on computers with much less memory.
Analysts at Morgan Stanley said on Thursday that efficiency improvements such as TurboQuant could reduce the infrastructure needed to run AI models.
“If models can run with materially lower memory requirements without losing performance, the cost of serving each query drops meaningfully, resulting in more profitable AI deployment,” Morgan Stanley analysts wrote. “Thus, models that need cloud clusters can fit on local hardware, effectively lowering the barrier to deploying AI at scale.”
However, analysts were not convinced this week’s sell-off was warranted. The implications for memory and computing were “neutral near term”, Morgan Stanley added, as lower AI costs would likely increase overall demand.
After falling throughout the week, Micron and other US memory providers saw their stocks rally slightly on Friday morning.
Until this week, Micron had seen its value more than quadruple in value over the past year, thanks to its position as the leading US provider of high-bandwidth memory, a key component alongside Nvidia’s chips in the vast AI data centres needed to power services like Google’s Gemini and OpenAI’s ChatGPT.
Soaring demand for computing power for AI and the growing complexity of AI systems such as Anthropic’s Claude Code have absorbed much of the available supply of memory, causing shortages elsewhere in the electronics supply chain. On Friday, Sony said it would raise PlayStation 5 prices by as much as 20 per cent, in part due to higher memory component costs.
Developments in AI have buffeted Wall Street this year, as investors worry the technology will disrupt large parts of the economy. New tools and model launches have hammered stocks in sectors from wealth managers to insurance brokers and property services.
On Friday, cyber security stocks fell sharply following reports that Anthropic’s forthcoming model has much greater capabilities and could render existing cyber defences obsolete. CrowdStrike and Palo Alto Networks fell more than 6 per cent, while Cloudflare dropped over 4 per cent.
