English
Français
Deutsch
Italiano
Português
Русский
Español
简体中文
हिन्दी
日本語
Samsung, SKHynix, and Micron are set to benefit because generative artificial intelligence required massive computing power, says Citi analyst Peter Lee.
Dreamstime
One thing about developing and applying large language models for generative artificial intelligence is that it requires massive computing power.
That explains this year’s roughly 175% rally in
Nvidia
stock (ticker: NVDA). Less appreciated is that all that computing power is going to drive a surge in demand for DRAM memory chips.
In a research note on Monday, Citi analyst Peter Lee provided some stunning estimates on how important AI-relating computing will be for the memory sector in the years ahead.
Lee notes that server’s based on Nvidia’s DGX100 graphics processor requires about six times the memory of a more traditional server.
Lee estimates that AI-related memory will account for 16% of global DRAM revenue this year, spiking 31% next year, 41% in 2205, and 56% by 2027. Lee, who is based in South Korea, reiterated his Buy ratings for both
Samsung Electronics
and
SK Hynix.
The trend should also benefit
Micron Technology
(MU), the only major U.S.-based memory-chip manufacturer. Micron reports earnings after the close of trading on Wednesday, and the results are likely to reflect recent demand softness, reflecting a sharp postpandemic slide in PC sales, and an accompanying buildup of customer inventory. For the May quarter, the Street sees sales down 58% from a year ago. But any commentary from Micron on the potential benefits to demand from AI could help boost the stock.
Micron stock on Monday are 0.7% higher at $65.75. The stock has gained about 32% year to date.
Write to Eric J. Savitz at [email protected]
Read the full article here