Did Nvidia Reach Out First? The Decisive Reason Samsung Electronics Will Flip the HBM4 Game | Hanbando CEO Park Jun-young

엔비디아가 먼저 손 내밀었다? 삼성전자 HBM4 판 뒤집는 결정적 이유ㅣ한반도 박준영 대표
Watch on YouTube ↗  |  February 13, 2026 at 08:30  |  21:45  |  815 Money Talk (815머니톡)
Speakers
Park Jun-young — CEO, Hanbando
Hwang Ui-won — Team Lead, PB Team, Shinhan Securities Sales Division

Summary

In this first part of an interview, Park Jun-young, CEO of Hanbando, explains why Samsung Electronics may be better positioned in HBM4 than expected, citing D1C DRAM, an in-house 4nm base die, over-spec performance, and decent yield. He expects SK hynix to stay strong in the first half on sold-out HBM3 but lose HBM4 share, while Micron's reported exclusion is not confirmed. The discussion also covers a longer memory supercycle, rising DRAM and NAND prices, and HBF as a future Nvidia-driven watch item.

  • Host Hwang Ui-won interviews Park Jun-young, CEO of Hanbando, about HBM4 and memory markets.
  • Samsung Electronics' HBM4 chances improve on D1C DRAM, 4nm base die, and over-spec performance.
  • SK hynix remains the first-half leader on sold-out HBM3, but HBM4 share may normalize.
  • Micron's HBM4 exclusion report is not confirmed, and its HBM3 business remains strong.
  • DRAM and NAND forecasts show strong revenue and price growth into 2027.
  • NAND/SSD demand is rising with AI inference needs.
  • HBF is an early Nvidia-requested future growth theme with SK hynix and SanDisk involvement.
  • The segment ends with a preview of physical AI and Samsung foundry topics.
Ideas
Park Jun-young CEO, Hanbando 1:36
HBM market stays tight and critical.
The HBM market remains tight: HBM3 is sold out this year across suppliers, and HBM4 is a new product that will ramp in the second half. Because GPU performance is constrained by HBM read speed and capacity, memory has become the main bottleneck in AI hardware, giving HBM suppliers strong pricing and demand visibility. Any supplier qualification or capacity expansion is therefore a positive catalyst.
Park Jun-young CEO, Hanbando 1:44
Micron HBM4 exclusion not confirmed.
The report that Micron will be excluded from Nvidia's HBM4 supply chain is not official and should not be treated as confirmed; it traces to a SemiAnalysis report. Micron's HBM3 revenue remains strong because HBM3 is sold out, and its single-chip quality is decent, so the door to HBM4 is not closed. However, Micron still uses an older D1B-class die and may enter HBM4 late, so investors need more confirmation on qualification and roadmap.
Park Jun-young CEO, Hanbando 3:55
Samsung HBM4 tech can reverse share.
Samsung Electronics is better positioned in HBM4 than it was in HBM3 because its D1C DRAM die improves power efficiency, it can make the base die in-house on a 4nm process versus SK hynix's outsourced 8/12nm, and it has achieved 11.7Gbps, above the JEDEC spec. Yield concerns are less severe because D1C has been in development and DRAM yields are easier than foundry yields. This over-spec strategy resembles SK hynix's successful HBM3 certification push, and since GPU performance depends on HBM speed, Samsung can close the gap and potentially reverse the HBM4 share game, with share possibly rising from around 22% toward 35% and breaking SK hynix's near-monopoly.
Park Jun-young CEO, Hanbando 10:54
HBM3 keeps SK hynix strong near term.
SK hynix remains attractive in the first half because HBM3 is sold out industry-wide and customers are satisfied with its D1B process, keeping its sales leadership. HBM4 share will likely fall from the current near-monopoly to around 50% as Samsung gains, but SK hynix should remain the largest HBM supplier by revenue, with Samsung possibly catching up around 2027. The key inflection point comes when Nvidia's Vera Rubin and HBM4 volumes ramp in the second half.
Park Jun-young CEO, Hanbando 13:38
Memory supercycle has room to run.
The memory cycle is likely longer and stronger than past cycles. DRAM market revenue is forecast to double from about $250bn to $500bn in one year and rise to about $750bn in 2027; DRAM contract prices are expected to rise around 95% year over year and NAND prices by more than 50%. Hyperscalers can absorb higher memory costs because HBM is a small part of an expensive GPU/server bill, while PC and mobile demand suffers, so memory makers still have pricing room.
Park Jun-young CEO, Hanbando 17:50
HBF is early, Nvidia-driven watch item.
HBF has become a larger future growth theme because Nvidia specifically requested it, pulling development earlier than the prior post-2028 timeline; SK hynix and SanDisk are working together on it. However, HBF is technically harder and more expensive than high-speed SSD, and some engineers question whether it is a must-have. A prototype is due by year-end, and if Nvidia shows that HBF materially improves its chips, it could become a major new momentum driver.
Park Jun-young CEO, Hanbando 18:21
Inference makes high-speed NAND/SSD essential.
As AI moves from training to inference, data must be stored close to the accelerator, making high-speed SSD/NAND increasingly important. NAND prices have already risen sharply because of this, and HBF is harder and more expensive than high-speed SSD, so the near-term beneficiary is the SSD/NAND market rather than HBF. HBF may create another catalyst later, but current momentum is in SSD/NAND.
Up Next

This 815 Money Talk (815머니톡) video, published February 13, 2026, features Park Jun-young discussing HBM memory, MU, 005930.KS, 000660.KS, DRAM, SNDK, NAND Flash. 7 trade ideas extracted by AI with direction and confidence scoring.

Speakers: Park Jun-young  · Tickers: HBM memory, MU, 005930.KS, 000660.KS, DRAM, SNDK, NAND Flash