Semiconductor News 20251227

Main News

1. Samsung and SK Hynix accelerate HBM4 production to early 2026
   Samsung Electronics and SK Hynix have reportedly accelerated their production schedule for their next-generation memory, HBM4, and are accelerating their competition to begin mass production in early 2026.Demand for high-bandwidth memory in the AI market To meet the explosive growth in demand for high-bandwidth memory in the AI market, the two companies are rapidly advancing technology development and capital investment. As a result, the battle for leadership in the memory field, which determines the performance of next-generation AI chips, is expected to intensify ahead of schedule.
2. SK Hynix ships HBM4 to Nvidia in partnership with TSMC
   SK Hynix has set a goal to begin full production of HBM4 in February through its partnership with TSMC. With final samples already shipped to Nvidia, SK Hynix is making concrete steps to maintain its leadership in the HBM market, a move that, combined with TSMC’s packaging technology, will ensure adoption by key customer Nvidia’s next-generation AI accelerators. Nvidia’s next-generation AI accelerators.
3. Japanese Government Quadruples Spending on AI and Semiconductors to Boost Support for Rapidus
   The Japanese government has announced plans to quadruple its spending on semiconductors and physical AI (Physical AI). In particular, the emphasis is on deepening state support for Rapidus, which aims to domestically produce cutting-edge semiconductors. This is a strategic expansion of investment to establish a domestic manufacturing base from the perspective of economic security and to increase international competitiveness in the area of AI implementation, including robots and industrial equipment.
4. Nvidia may not adopt Intel’s 18A process
   Nvidia has reportedly tested Intel’s most advanced manufacturing process, 18A, but has made no progress toward adoption. 18A is Intel’s trump card in rebuilding its foundry business, but However, it may not have been able to win orders from Nvidia, the king of AI chips. While this is a headwind for Intel’s contract manufacturing strategy, it also suggests that TSMC’s dominance remains unassailable.
5. AMD Receives MI308 Chip Order from Alibaba to Rival Nvidia
   AMD has reportedly won a major order for its MI308 AI chip from China’s Alibaba Group. This is a move to compete with Nvidia’s H200 and shows AMD’s growing presence in the huge Chinese market. With the constraints of U.S. export restrictions, it can be seen that Chinese tech giants are diversifying their AI infrastructure by adopting AMD’s solutions as an alternative to Nvidia.
6. SoftBank and Fujitsu collaborate to develop power-saving memory for AI
   Fujitsu will participate in the development of next-generation memory for AI led by SoftBank. With power consumption in AI data centers becoming an issue, the two companies aim to establish memory technology with superior power-saving performance. Combining the technological capabilities of Japanese companies that once led the semiconductor industry, the two companies aim to regain a foothold in the AI infrastructure market through domestically produced technology and to improve energy efficiency at the hardware level.
7. Nvidia and AI Chip Startup Groq Sign 29 Trillion W on Technology Deal
   Nvidia has reportedly signed a 29 trillion won (about $3 trillion) technology deal with AI inference chip startup Groq, which has strengths in high-speed inference processing technology. Nvidia’s dominance in the AI chip market may be further strengthened.
8. Google Fires Procurement Executive Responsible for HBM Supply Shortage
   Google has reportedly fired its executive in charge of procurement, holding him responsible for causing a severe supply shortage of HBM (high bandwidth memory), which is essential for AI servers. This is a tough move in the face of intensifying competition from Microsoft, Meta, and others.
9. Intel Unveils Giant Chiplet Package Technology Surpassing TSMC
   Intel has unveiled an extreme multi-chiplet package construction technology that reaches 12 times the size of today’s largest AI processors. This is said to surpass even TSMC’s planned largest design, demonstrating Intel’s persistence in regaining leadership not only in miniaturization but also in packaging technology. It is notable as the hardware foundation for future giant AI clusters.
10. Huawei to Sell Ascend 950 AI Chip in South Korea in 2026
    China’s Huawei reportedly plans to sell its Ascend 950 AI chip in the South Korean market in 2026. The company aims to go beyond mere chip sales to large-scale deployment on a cluster basis. This is a move by Huawei, which is under U.S. sanctions, to enter the South Korean market, where demand for AI is growing, and could have an impact on the AI hardware power structure in the Asian region.

News List