On May 9th, it was reported that DRAM memory chips, memory modules, NAND flash, and SSD hard drives are accelerating in a new round of price hikes. In its latest report, Jibang Consulting has significantly raised its price increase expectations for the second quarter, especially for memory.
In the first quarter of 2024, contract prices for memory chips surged by as much as 20%. Originally expected to increase slightly by 3-8% in the second quarter, Jibang Consulting has now revised this to 13-18%.
The trend for NAND flash is similar, with a surge of 23-28% in the first quarter. The second quarter was initially expected to increase by 13-18%, but it has now been adjusted to 15-20%.
The only silver lining is for eMMC/UFS flash memory, which is expected to increase by only about 10% in the second quarter—not a small increase either.
Fubon Consulting stated that the Hualien earthquake on April 3 was a turning point.
Before the quake, both the spot prices and trading volumes of memory and flash memory had been declining, lacking upward momentum. This was mainly due to weak demand in the smartphone and laptop markets outside of AI, with PC manufacturers gradually increasing their inventories and buyers unwilling to accept continuous price hikes.
However, after the quake, PC manufacturers, for various reasons, began to accept significant increases in contract prices for memory and flash memory. By the end of April, a new round of contract price negotiations had been successively completed, with increases exceeding previous expectations. On one hand, buyers showed willingness to increase their inventories, and on the other hand, AI demand remained strong.
Of particular note is the strong demand for HBM high-bandwidth memory in AI applications, leading to original manufacturers expanding production capacity one after another.
For example, Samsung's HBM3E, which utilizes advanced 1α technology, is set to significantly increase production in the third quarter, with this segment expected to account for approximately 60% of total capacity by the end of this year. This will naturally displace DDR5 production capacity, further exacerbating price hikes.
At the same time, AI inference servers are increasingly focusing on energy efficiency, giving priority to enterprise-grade QLC SSDs, which will also occupy a considerable amount of production capacity, leading to a shortage of consumer-grade NAND.
