Analysts are debating whether increased efficiency will ultimately reduce or boost memory demand, citing Jevons Paradox. Shares of companies that make computer memory and storage took a hit after Google researchers announced a new compression technique. This new method could significantly reduce the amount of memory needed for artificial intelligence (AI) workloads.
SK Hynix Inc., a major producer of memory chips crucial for AI applications, saw its shares drop by as much as 6.4% on the Korea Exchange. Kioxia Holdings Corp., a flash memory manufacturer, experienced a similar decline in Tokyo. These drops followed losses by Micron Technology Inc. and Sandisk Corp. on Wednesday in New York.
Alphabet Inc.’s Google said its new TurboQuant technology can cut the memory required to run large language models by at least six times. This would lower the overall cost of training AI. Memory is a vital component in Nvidia Corp.’s AI accelerators, and its demand has surged dramatically during the recent AI boom.
The news from Google caused some worry that memory needs might decrease. However, those optimistic about the booming global memory market argue that improved efficiency will actually increase demand, rather than reduce it. They point to a theory known as the Jevons Paradox.
This 19th-century idea, which states that more efficient technology can lead to increased demand, was mentioned in a note from JPMorgan Chase & Co.’s trading desk. Their analysts suggested that investors might take some profits after the news, but they see no immediate threat to memory consumption.
Morgan Stanley analyst Shawn Kim wrote that TurboQuant is positive for large cloud providers (hyperscalers) due to the potential return on investment. He added that it might even benefit memory makers in the long run, as “a lower cost per token can also lead to higher product adoption demand.”
Memory and storage product prices have been rising in recent months because of shortages caused by the huge demand from the AI boom. This has led to rapid stock increases for related companies, such as Kioxia’s 700% surge since late August.
Some analysts disagreed with the idea that TurboQuant would negatively affect memory demand, again referring to Jevons Paradox. This theory, from an English economist, about coal production suggests that the more efficient technology becomes, the more demand will grow. The same idea came up last year when China’s low-cost DeepSeek AI model sparked fears of a reduced need for more advanced technology.
Ortus Advisors analyst Andrew Jackson wrote in a note on Smartkarma that the Google development might make “little difference to demand given the extreme supply constraints.” For Kioxia, he added, “after such massive gains it makes sense we see a bit of profit-taking creep in.”
