According to a report released by KB Securities, as China expands its AI data‑center investments, Samsung Electronics’ revenue is expected to grow significantly. In particular, the company’s memory business will benefit from development of AI accelerators and improved computational efficiency, along with strengthening supply chains.
Kim Dong‑won, head of research at Samsung Electronics, said that China’s semiconductor industrialization policy is not meant to reduce Samsung’s memory sales but rather is expected to drive additional revenue. He explained that Chinese AI accelerators will have lower compute efficiency than Nvidia GPUs, so more accelerators and memory are needed to achieve the same AI compute volume. He also emphasized that the technical superiority of CxMT in high‑performance server DRAM remains, and that the four major cloud service providers in China – Baidu, Alibaba, Tencent, and ByteDance – have requested cooperation from Samsung Electronics not only for long‑term supply agreements (LTAs) but also to produce their own AI accelerators (ASICs).
Global data‑center investment is expected to be led by the United States while China will accelerate its growth in a coordinated manner. Accordingly, the AI infrastructure investment cycle remains at an early stage. As of 2026, the U.S. AI data‑center capacity is 55 GW, projected to reach 121 GW by 2030; China’s capacity will grow from 40 GW to 80 GW – a doubling.
Kim noted that the total AI data‑center investment in the United States and China is estimated at about $5 trillion, with both countries’ capacities expanding from 95 GW to 201 GW (2.1×) during the same period. He further projected an annual investment of roughly ₩1800 trillion, and that as AI model sophistication and demand rise, along with increased adoption of in‑house accelerators, server compute performance and memory capacity will both increase.
(Article written by Lee Su·Hankyung.com reporter 2su@hankyung.com)