00:00It's Benzinga bringing Wall Street to Main Street.
00:02Google's research on improving AI efficiency is pressuring memory stocks
00:06as investors worry it could reduce future demand for chips, according to CNBC.
00:11Google unveiled TurboQuant, a compression method that it said can reduce the memory
00:15needed to run large language models by six times by shrinking the key value cash that
00:20stores past calculations. The technique aims to improve AI model efficiency,
00:25a key focus for leading labs. Shares of SK Hynix fell six percent, Samsung dropped nearly five
00:31percent, and Kioxia declined nearly six percent in Asia, while SanDisk and Micron also fell in the
00:38U.S. In a post on X, Matthew Prince called the development Google's Deep Seek, citing efficiency
00:44gains. Analyst Ray Wang said the breakthrough target's a key bottleneck and may still drive
00:49higher memory usage as models improve. Analysts said recent declines were also driven by profit-taking
00:55after strong gains, with long-term demand supported by high prices and supply shortages.
01:00For all things money, visit Benzinga.com.
Comments