00:02You're listening to AI PRISM from Seoul Economic Daily.
00:06GPU rental costs just jumped nearly 25% for a Korean AI cloud provider.
00:11NVIDIA's CEO is now openly siding with the White House against AI slowdown calls.
00:17And industry groups are warning Korea's biggest chip fabs could run without finished roads.
00:22It's Tuesday, September 22nd. Let's look at a business story shaping corporate Korea.
00:27Korean AI cloud provider Vessel AI raised its NVIDIA H100 rental rate 24.7% this month,
00:34following similar hikes at U.S. neocloud firms.
00:38Compute scarcity is now hitting older GPU models, not just the newest chips.
00:43At the same time, industry groups are pressing government to fix infrastructure gaps before major fabs come online.
00:50Here's what the numbers show.
00:53VESSLAI's H100 hourly rate rose from $2.39 to $2.98, while the broader market rate for H100 rentals
01:03climbed 21.9% in a single month to $3.28 per trading platform, ORN.
01:12VESSLAI operates roughly 7,000 GPUs today and plans to expand to 10,000 by year-end.
01:19NVIDIA
01:19CEO Jensen Huang publicly aligned with President Trump against AI slowdown calls during a live phone call at the All
01:28-In Summit on September 14th.
01:30The Federation of Korean Industries submitted 41 policy requests, including a demand to cut water facility approval timelines
01:37from 42 months to 18 months ahead of SK Hynix's Yongin fab, set to start operations in February.
01:45Korea's memory chip exports rose 350.7% year-on-year for September 1st to 20th, reaching $21.5 billion.
01:53Some Korean firms have already dropped out of government AI transition project bids due to GPU shortages.
02:00So what does this mean for corporate Korea?
02:02Earlier, we said GPU rental costs just jumped nearly 25%.
02:06Here's what that actually means for you.
02:09Compute access, not just algorithm quality, is becoming a board-level competitiveness issue for any company running AI workloads at
02:16scale.
02:17Meanwhile, infrastructure bottlenecks at Yongin and elsewhere could delay production timelines,
02:22even as chip demand and export volumes keep climbing.
02:25What to watch?
02:26Whether more Korean AI cloud providers raise GPU rental rates.
02:30SK Hynix's Yongin fab startup in February.
02:33And whether access roads are ready.
02:35And any government response to the 41 industry policy requests submitted this week.
02:40That's today's AI Prism, CEO.
02:43This episode was produced with AI assistance based on Seoul Economic Daily reporting,
02:48and reviewed by a human editor.
02:50AI Prism is a Juan Afro award-winning series.
02:53We'll be back tomorrow.
02:54You've been listening to AI Prism from Seoul Economic Daily.
댓글