00:02Samsung and AMD just took aim at AI's memory bottleneck.
00:07The fix is capacity, not just speed.
00:11And SK Hynix joined the race.
00:13It's Wednesday, October 14th.
00:16Let's take a look at a key development in Korea's industry sector.
00:20Samsung and AMD jointly demonstrated CXL-based memory expansion
00:25at the OCP Global Summit in California.
00:28AI agents are multiplying inference data, making capacity as critical as speed.
00:35SK Hynix showed its own CXL-based rack there.
00:38Here's what the numbers show.
00:40Eighth-generation HBM5 leads Samsung's speed roadmap.
00:45CXL memory pooling lets servers share memory,
00:48trimming data center investment and operating costs.
00:52Samsung has supplied AMD with HBM3e since 2025
00:56and added priority HBM4 supply this year.
01:01Separately, HD Hyundai and Siemens opened a Busan center,
01:05targeting 30% shorter shipbuilding times by 2030.
01:09So, what does this mean for supply chain managers?
01:13Earlier, we said the fix is capacity, not just speed.
01:17Here's what that actually means for you.
01:20Memory makers now enter at the data center design stage, so sourcing decisions move earlier.
01:26Watch whether cloud giants adopt CXL pooling at scale.
01:31Watch HBM4 volumes for AMD's Instinct MI455X.
01:36And watch SK Hynix's CXL rack.
01:40That's today's AI Prism, Industry Daily.
01:43This episode was produced with AI assistance based on sole economic daily reporting
01:49and reviewed by a human editor.
01:51AI Prism is a Juan Ifra award-winning series.
01:55We'll be back tomorrow.
01:57We'll see you next time.
01:59We'll see you next time.
