The Real Reason NVIDIA Entrusted Grok-3 to Samsung Electronics... Earnings Will Explode 'At This Time' / The Decisive Reason for Saying 'Industry Conditions Are Unwavering' Amid Semiconductor Cycle Debate | Hanbando CEO Park Jun-young

Watch on YouTube ↗  |  August 27, 2026 at 06:00  |  27:13  |  815 Money Talk (815머니톡)
Speakers
Park Jun-young — CEO, Hanbando

Summary

Park Jun-young, CEO of Hanbando, discusses Korean semiconductor stocks ahead of Nvidia earnings. He argues Nvidia's Grok 3 LPU and Samsung's 4nm win support Samsung Electronics foundry earnings, while a Q3 earnings rebound and foreign inflows could lift Samsung and SK hynix. He rejects the view that the memory cycle is turning, citing no wafer supply growth until late 2027 and HBM-driven DRAM tightness.

  • Nvidia's Grok 3 LPU uses SRAM and Samsung's 4nm, giving Samsung a near-exclusive foundry role.
  • Samsung 4nm is roughly 30 percent cheaper than TSMC and is already running near full capacity.
  • HBM4 logic dies on Samsung 4nm made their first monthly profit in June.
  • Foreign net buying in Korean semis likely requires AI earnings confirmation and Q3 results.
  • Memory wafer supply does not increase until late 2027; next year may see the worst shortage.
  • HBM consumes about 2.5-3 wafers, tightening plain DRAM and supporting Samsung and SK hynix.
Ideas
Park Jun-young CEO, Hanbando 2:25
Korean semis rebound around Q3 earnings.
The memory industry is not entering a cyclical downturn. Wafer production does not increase until late 2027, AI demand is still expanding, next year is expected to have the worst shortage, and rising HBM consumes roughly 2.5-3 wafers, tightening plain DRAM supply further. Memory is now embedded in Nvidia GPUs rather than sold directly to price-cutting end customers, so memory prices are unlikely to fall. Plain DRAM orders are strong and Samsung has the supply advantage, keeping Samsung and SK hynix supported through 2028.
000660.KS HIGH
Park Jun-young CEO, Hanbando 2:25
Korean semis rebound around Q3 earnings.
Nvidia's Grok 3 LPU uses SRAM, a memory-type chip made in Samsung's system semiconductor/foundry business where Samsung is nearly monopolistic, and this is additive rather than a replacement for HBM. Nvidia selected Samsung's 4nm instead of TSMC because Samsung's 4nm yields are solid, it is about 30 percent cheaper than TSMC, and TSMC is already full; Samsung's 4nm is near full utilization, HBM4 logic dies on 4nm made their first monthly profit in June, and capacity is expected to grow about 50 percent next year. Advanced packaging is also short, allowing Samsung to bundle memory with foundry capacity to win AMD, Google, Meta and Amazon custom orders.
005930.KS HIGH
Up Next
3:48
Nvidia's Upbeat Sales Outlook Eases Concerns About AI Economy
Bloomberg Markets · Aug 27
1:10:57
Samjeon-Nix Shareholder Returns, Won-Dollar Exchange Rate Decline Effect | Chesley Investment Advisory Executive Director Park Se-ik [Womae Shinbak / 26.08.27.Thu]
Chesley Investment Advisory (체슬리투자자문) · Aug 27
1:32:41
Z.AI Claims New Model Taking on Anthropic, OpenAI | The China Show | 8/27/2026
Bloomberg Markets · Aug 27
41:07
Can Tokenized Stocks Beat Your Broker?
Unchained (Chopping Block) · Aug 27
1:32:26
Tech Stocks Rise on Nvidia's Outlook; BOK Hikes Rates | The Asia Trade 8/27/2026
Bloomberg Markets · Aug 27

This 815 Money Talk (815머니톡) video, published August 27, 2026, features Park Jun-young discussing 000660.KS, 005930.KS. 2 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Park Jun-young  · Tickers: 000660.KS, 005930.KS