플레이어로 건너뛰기본문으로 건너뛰기
A new AI-era memory called HBF has arrived. Its manufacturing cost per capacity is roughly one-tenth that of HBM4 — but analysis shows it requires 6.5 times more wafers than conventional NAND to produce the same capacity. We look at why this new memory could actually reduce existing NAND supply.

#Banjinza #HBF #HBM #NANDFlash #SKHynix #Samsung #SanDisk #AIMemory #NewsClip

카테고리

🗞
뉴스
트랜스크립트
00:00A memory is being developed with a manufacturing cost per capacity, just one-tenth that of HBM.
00:07But analysts say this technology could actually reduce supply of conventional NAND.
00:12It's called high-bandwidth flash, HBF.
00:16When AI generates a response, it constantly pulls from massive model data and conversation history.
00:23Storing all of that in expensive HBM is a burden, and SSDs are too slow.
00:29HBF is a new type of memory that stacks multiple NAND chips to handle large-scale data between HBM and
00:36SSDs.
00:37According to Bernstein, its manufacturing cost per capacity is roughly one-tenth that of HBM4.
00:44But here's the twist.
00:45It uses 6.5 times more wafers than conventional NAND to produce the same capacity.
00:51If HBF takes off commercially, chip makers may shift production capacity toward it,
00:57leaving less supply for conventional SSD-grade NAND.
01:01SK Hynix and SanDisk are leading standardization efforts, while Samsung and Micron have also joined the race.
01:08Will HBF become NAND's next growth engine, or the next source of a supply crunch?
01:14This has been Benginza.

추천