🧔♂️ A friendly human may check it before it goes live. More news here
Samsung shares near record high on AI chip demand
Samsung Electronics shares surged as much as 6% to 94,300 won in Seoul on October 10, on track to reach a record high after markets reopened following a weeklong holiday.
The South Korean tech giant’s stock has climbed 76% so far this year, raising its market capitalization to US$390 billion.
Investors are betting on Samsung’s prospects in high-bandwidth memory for AI, as the company negotiates supply volumes with Nvidia and benefits from a supply deal with OpenAI that also involves SK Hynix.
Analysts also expect prices for traditional DRAM and NAND chips to rise due to supply shortages, benefiting Samsung’s core business.
Morgan Stanley increased its price target for Samsung by 14% to 111,000 won, citing strong Q4 memory chip pricing and continued demand into 2026.
Samsung’s gains and a 145% rise for SK Hynix in 2025 have contributed to South Korea’s Kospi benchmark rising 50% year to date.
🔗 Source: Bloomberg
🧠 Food for thought
Implications, context, and why it matters.
Samsung’s HBM timeline leaves supply gaps ahead
- Samsung passed Nvidia’s HBM3E (latest generation of High Bandwidth Memory) tests, but large shipments likely slip to 2026. SK hynix began 12‑high HBM3E (a stack of 12 memory layers) in September 2024 1. Micron followed in early 2025 2. Samsung sits third in line, so Nvidia supply stays tight 3.
- The green light arrived about 18 months after Samsung finished development, after failed tries 3. Accounts say 12‑high HBM3E work started in February 2024, with redesigned sample sent in the first half of 2025, pushed by chairman Lee Jae‑yong (Samsung’s executive chairman) 2. The pace still trails rivals.
- Nvidia’s letter says it plans to adopt Samsung HBM3E and intends to place orders 2. Details on volume, pricing, or schedules remain in play. Near‑term revenue may stay small.
Packaging limits spur early bookings
- CoWoS, especially CoWoS‑L that pairs GPUs with HBM on a silicon interposer, stays tight through at least H1 2026 4. Hyperscalers (large cloud providers operating massive data centers) add accelerators faster than capacity grows, so AI servers wait even when GPUs ship.
- Builders of AI infrastructure need early slots with OSATs (Outsourced Semiconductor Assembly and Test providers) such as ASE, Amkor, or SJSemi 5. Non‑TSMC houses could hold 31% of CoWoS‑class capacity by Q4 2026.
- Watch expansion by advanced packaging firms and substrate makers (they make the high‑density circuit boards used in chip packaging) 6. Ibiden with AT&S lead abroad. Unimicron with Shinko scale North America to meet demand and offset tariffs that raised duties up to 25% in 2025.
Recent Samsung developments
Stay updated on the go with our mobile app.
Get latest insights with smoother, more personalized experience through TIA mobile app.




