Samsung leaders push for AI focus as competition rises

Samsung Electronics’ co-CEOs have called for greater focus on AI as competition in the sector increases globally.

Vice Chairman and CEO Jun Young-hyun, who leads Samsung’s chip division, said the company must respond to rising demand for AI chips and develop AI technologies for all aspects of semiconductor production.

He noted that Samsung is the only chipmaker offering logic chips, memory, foundry services, and advanced packaging as a single provider.

President and co-CEO Roh Tae-moon, who heads Samsung’s device division, urged tighter integration of AI across Samsung’s products, including mobile devices, TVs, and appliances.

Roh said that AI is changing how the company works and that Samsung must strengthen its ability to compete and adapt to external challenges.

Separately, Samsung SDI CEO Choi Joo-sun encouraged staff to recognize risks and prepare for future opportunities in 2026.

🔗 Source: Yonhap

🧠 Food for thought

Implications, context, and why it matters.

Samsung’s AI chip bet rests on High Bandwidth Memory (HBM) execution and packaging scale that remain unproven

Samsung cleared Nvidia’s HBM3E in late 2025 after 18 months with failed tries, but large Nvidia volumes wait until 2026 as SK hynix and Micron handle orders, plus 2026 HBM capacity is sold out 1 2.
Through Broadcom, a chip and networking supplier, Samsung supplies over 60% of Google Tensor Processing Unit (TPU) HBM after a 1a-class dynamic random-access memory (DRAM) redesign fixed heat issues, where “1a” denotes a DRAM process generation 3.
HBM4 timing stays uncertain as Samsung reached 11 Gb/s and 36 GB using 1c DRAM, a subsequent DRAM process generation, plus 4-nm logic dies, while SK hynix plans volume HBM4 in late Q2 2026 3. Leadership says Samsung alone offers logic, memory, foundry, and advanced packaging, while advantage depends on Chip-on-Wafer-on-Substrate (CoWoS)-style advanced packaging capacity, which provides 2.5D integration of logic and memory that determines whether it can absorb AI accelerator demand shifting from Taiwan Semiconductor Manufacturing Company (TSMC).

Device AI developers should prioritize Samsung’s Software Development Kit (SDK) and policies ahead of possible 2026 Galaxy AI monetization

Galaxy AI stays free through 2025, with purchase possible after, as Samsung adds AI across phones, TVs, and appliances 4.
Samsung Neural SDK no longer ships to third-party developers. It runs only on Samsung devices, and TensorFlow models do not run on the Exynos Neural Processing Unit (NPU), where Exynos is Samsung’s in-house mobile processor brand 5. Exynos AI Studio converts models from PyTorch, Open Neural Network Exchange (ONNX), TensorFlow, as well as TensorFlow Lite (TFLite) for NPU execution, then applies quantization plus graph optimization for efficient on-device AI; for enterprise deployments, the Knox Service Plugin lets administrators restrict cloud-dependent AI and enforce on-device processing for work profiles 6 7.