MW Micron's stock is dropping. Is Google partly to blame?

By Britney Nguyen

Google introduced an algorithm that it says improves memory usage in AI models. Whether that will actually eat into business for Micron and rivals is unclear.

Micron's stock was down about 3% on Wednesday afternoon.

The rally in memory stocks was losing further steam Wednesday, as investors are seemingly worried about the sustainability of sky-high component prices.

Micron Technology's stock $(MU)$ was down 3.3% in afternoon trading and on track to log its fifth straight session of declines. Sandisk's stock $(SNDK)$ was off 3.2%, falling for the fourth session in a row.

The latest source of pressure may have been an announcement from Google $(GOOGL)$ $(GOOG)$, which introduced its TurboQuant compression algorithm late Tuesday. Google said TurboQuant dramatically reduces memory usage while improving the speed and efficiency of AI models.

The algorithm addresses vectors, which help AI models understand and process simple to complex information. High-dimensional vectors consume a lot of memory, Google explained, therefore creating bottlenecks in key-value (KV) cache, or the technique that allows an AI model to "remember" data it has processed so it can respond to inquiries faster.

Read on: Micron's stock gains officially carry the company into an exclusive club

While high-dimensional vectors can be compressed through vector quantization, that technique comes with its own memory issues as it requires calculating and storing additional data, "partially defeating the purpose of vector quantization," Google said. TurboQuant, which Google published a research paper about last April, is its answer to that problem.

Shares of storage makers Seagate Technology $(STX)$ and Western Digital $(WDC)$ were also trading lower, with both down almost 1% in Wednesday afternoon action.

Memory and storage makers have benefited from components shortages that have allowed them to raise prices significantly on their offerings. As context windows - the amount of data processed at one time by an AI model - get larger, demand for memory has rocketed.

Some investors may be seeing Google's announcement as a sign that AI players won't need so much memory going forward, thereby crimping the massive pricing power of suppliers like Micron.

But others say it's not so simple, drawing parallels to past instances in which technological advancements drove down the cost of resources - which, in turn, increased demand because the resources were more accessible.

"Focus is on Jevons paradox," Jefferies equities-trading analyst Jeffrey Favuzza said in a Wednesday note, referring to the term for that phenomenon.

Google said its new compression algorithm reduces KV-cache memory in AI models by at least six times and increases speeds by up to eight times - essentially making inference, or the process of running AI models, cheaper. The Jevons paradox suggests that would drive up AI-model adoption and thus lead to more demand for memory.

Despite the stock reactions, Favuzza said his impression of early buy-side sentiment was that the announcement "isn't as big of a deal as it seems on the surface."

Another potential driver of the souring memory sentiment may be the fear of increased capacity, which could also lead to lower prices.

Micron is one of three producers in the world of high-bandwidth memory, along with South Korea's SK Hynix (KR:000660) and Samsung Electronics (KR:005930). The memory-chip makers have been reluctant to add capacity, out of fear that demand will slow down in what is typically a cyclical industry. The scarcity of supply has driven memory prices way higher.

But SK Hynix said in a regulatory filing Tuesday that it plans to purchase $8 billion worth of extreme-ultraviolet scanners from ASML $(ASML)$, which would be used to step up its chip-production efforts.

Meanwhile, OpenAI announced on Tuesday afternoon that it is winding down the app for its AI video generator, Sora - an example of a compute-intensive product that requires a lot of memory and storage. The move comes as OpenAI reshuffles its strategy to dedicate more compute resources toward core business products ahead of a possible initial public offering, the Wall Street Journal reported.

See more: What OpenAI scrapping its Sora video generator tells us about AI

-Britney Nguyen

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

March 25, 2026 15:00 ET (19:00 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.