SemiAnalysis: Memory as a proportion of cloud vendors' capital expenditure will exceed 40%, and the market has not yet fully priced this in.

robot
Abstract generation in progress
Mars Finance News, July 4 – Independent semiconductor and AI research firm SemiAnalysis noted that the proportion of memory spending in hyperscale cloud providers' capital expenditure is becoming a hot topic in the market. Particularly after Micron's earnings report last week, some market participants were shocked by the high proportion of memory spending expected next year. SemiAnalysis first proposed back in February that memory spending would account for about 30% of cloud providers' capital expenditure. At that time, most clients questioned this figure, citing that memory typically only accounts for a dozen or so percent of a server's bill of materials. In May, as memory prices rose faster than expected, SemiAnalysis further responded: if DRAM, NAND, and HBM are combined, memory spending in Nvidia systems will exceed 30% by the end of 2026 and further rise to over 40% by 2027. SemiAnalysis expects that market understanding of this structural shift will deepen over the coming months. This means that against the backdrop of continuous expansion in AI capital expenditure, the memory segment is upgrading from a "minor item" in server costs to the "single largest cost item." Its impact on memory manufacturers' pricing power and profit elasticity has not yet been fully priced in.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned