Server DDR5 RDIMM Prices Surge Past $2,000
The spot price of DDR5 RDIMM (Registered DIMM) used in artificial intelligence (AI) servers has surpassed $2,000. The price of the 32GB product, which was in the high $300 range in November of last year, has jumped more than fivefold in about 10 months.As AI data center investments continue but DRAM
The cost of DDR5 RDIMM, a type of server memory used in artificial intelligence (AI) servers, has skyrocketed past $2,000. This price jump occurred over just 10 months, more than quintupling from the high $300 range seen in November of the previous year. Market research firm DRAMeXchange reported that the average spot price for the 32GB DDR5 RDIMM products in the second week of September was $2,050, marking a 7.9% increase from the previous week. The highest price recorded in that week was $2,380, while the lowest price was $1,950.
DDR5 RDIMM, which consists of multiple DDR5 DRAM chips mounted on a single board, is crucial for servers handling large-capacity memory required for AI data centers. The price of DDR5 RDIMM has been surging again over the past two weeks, with the average price rising from $1,800 on August 31 to $2,050 on September 14, a 13.9% increase in just two weeks.
Looking at the broader trend, the price has risen significantly over the past year, from around $380 in mid-November last year to over $2,000 in the second week of September. This price increase is primarily due to a tight supply and demand situation in server DRAM. As global cloud service providers (CSPs) continue to invest in AI data centers, demand for DDR5 mounted on general servers and high bandwidth memory (HBM) has grown.
However, suppliers' inventories remain low, and the margin of price increase is expected to be lower than in the first half of the year due to long-term agreements (LTAs) between major memory companies like Samsung Electronics, SK hynix, and Micron and CSPs.
Written by urgent.news from BusinessKorea's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.