According to the latest analysis report published by KB Securities, as AI technology fully enters the inference stage, the central axis of the memory industry is changing rapidly. There are expectations that Samsung Electronics' corporate value will be re-evaluated once more, given its top priority in securing memory production capacity, and the investment opinion for the stock is maintained as 'Buy', with a target price set at 360,000 won. This suggests that it is not merely a short-term positive factor but rather a long-term investment logic that points to the core of the AI industry structure change.
Industry experts have pointed out that in the past era when AI learning models had to process large volumes of data at once, the combination of graphics processing units and high-bandwidth memory played an absolute role. However, with the rise of AI inference and agent technologies, methods of processing information and data sequentially have become essential. Consequently, it is expected that the importance of central processing units, server dynamic random access memory, low-power mobile DDR 5X, NAND flash, and others will expand rapidly. The landscape of CPU performance competition is very likely to be determined by the amount of general-purpose memory installed, and this change will bring huge synergy effects not only to CPU companies such as Intel but also to memory companies including Samsung Electronics.
In addition, the ratio of CPU and GPU installation is also undergoing rapid fluctuations. In the past AI learning stage, GPUs were overwhelmingly dominant, but the ratio has been adjusted from 1:8 to 1:3. As the industry moves into a multi-agent AI environment, the proportion of CPUs is expanding further. There is analysis suggesting that CPUs are likely to become the mainstream at a level of 2:1, surpassing 1:1 in the future. This aligns with the forecast that global data center capacity will expand to twice its current level by adding about 100 gigawatts by 2030, providing strong grounds for memory demand to continue a strong upward trend in the long term.
In particular, the fact that major AI platform companies still have significant room for data center expansion supports future market growth. While Anthropic's Claude, which has established a strong foothold in the agent AI field, has a 2030 data center expansion plan at the level of 10 gigawatts, OpenAI's plan is at 30 gigawatts, which is much larger in scale. This means that major AI platforms have sufficient room to lead the market through additional infrastructure investments in the future. Considering the flow of the AI market evolving from generative AI to agent AI and then to physical AI starting after 2022, the investment attractiveness is expected to continue to increase in the near future. Ultimately, in the long-term strength of memory demand that will continue until 2030, the conclusion is drawn that the ultimate market winner will be Samsung Electronics, which has mastered massive memory production capacity and supply capability.