Google's new algorithm sparks "storage heat cooling panic," but Wall Street analysts remain unfazed

Since Google research team announced a breakthrough new algorithm for AI memory compression, global memory stocks have suffered heavy losses.

This decline further accelerated on Thursday and Friday. However, Wall Street analysts remain optimistic.

Global chip stocks continue to “bleed”

In South Korea’s stock market, memory giant Samsung Electronics closed down 4.71% on Thursday and continued to fall 3.16% in early Friday trading; SK Hynix closed down 6.23% on Thursday and fell nearly 5% in early Friday trading.

In the Japanese stock market, Kioxia closed down 5.7% on Thursday and dropped 7.18% in early Friday trading.

In the U.S. stock market, Micron Technology, Western Digital, and SanDisk all further declined sharply after closing lower on Wednesday, with Thursday’s drops of 6.9%, 7.7%, and 11%, respectively.

In recent months, storage chip stocks soared as major tech giants heavily invested in AI infrastructure, leading to a shortage of memory supply and a significant rise in memory prices and profits.

As of this Wednesday, SK Hynix and Samsung’s stock prices have surged over 50% this year, while Japan’s long-struggling Kioxia has more than doubled.

However, the emergence of Google’s technology has shattered this thriving scene.

Google’s new compression algorithm, TurboQuant, reportedly reduces runtime cache memory usage of large language models by at least 6 times and boosts performance by 8 times. Industry insiders expect that Google’s new technology could alleviate the memory supply shortage, potentially lowering memory prices.

Wall Street collective optimism

Unlike the sharp sell-off in the stock market, Wall Street analysts have collectively expressed optimism in their reports.

Morgan Stanley analyst Shawn Ki wrote that Google’s research should have a more positive impact on the industry because it addresses a key bottleneck—improving the efficiency of the so-called key value cache used for reasoning (i.e., running AI models).

He stated, “If models can significantly reduce memory requirements without sacrificing performance, the cost of serving each query will be substantially lowered, enabling more profitable AI deployment.”

Ki also noted that, given the return on investment opportunities, the release of TurboQuant is beneficial for large enterprises. In the long run, this could also benefit memory manufacturers, as “reducing per-unit costs can lead to higher product adoption demand.”

Like many “optimists” in the AI industry and among analysts, he cited the so-called “Jevons Paradox,” an economic theory describing an counterintuitive relationship between technological progress and resource consumption. It states that when technological efficiency improves, resource consumption not only fails to decrease but may actually surge.

JPMorgan Chase and Citigroup also referenced this theory. JPMorgan analysts said that although some investors might take profits from this news in the short term, memory demand will not be threatened in the near future.

Andrew Jackson, an analyst at Ortus Advisors, bluntly stated, “Given the current extreme tightness of memory supply, Google’s new algorithm may have little impact on memory demand.”

(Source: Cailian Press)