又一个芯片巨头，要抢HBM，SK海力再创新高

Core Viewpoint - AWS is becoming a key customer for SK Hynix's HBM business, as the company invests heavily in global AI data centers and seeks to strengthen its collaboration with SK Group [1][2]. Group 1: AWS Investments - AWS plans to invest AUD 20 billion (approximately KRW 17.6 trillion) to expand its data centers in Australia from this year until 2029 [1]. - AWS has also committed to investing USD 10 billion in North Carolina, USD 20 billion in Pennsylvania, and USD 5 billion in Taiwan [1]. - This year, AWS's investment in AI infrastructure is expected to reach USD 100 billion, representing a 20% increase compared to the previous year [1]. Group 2: Collaboration with SK Group - AWS is set to collaborate with SK Group to build a large AI data center in the Ulsan National Industrial Complex, targeting an operational capacity of 103 MW by 2029, with an investment of USD 4 billion [2]. - SK Hynix is preparing to supply 12-layer HBM3E products to AWS, responding to the growing interest in HBM as AWS expands its AI semiconductor production [4]. Group 3: HBM Demand and Technology - The demand for HBM is expected to rise due to AWS's investments in AI data centers, as HBM significantly enhances data processing performance compared to existing DRAM [2]. - AWS is projected to account for 7% of the total demand for Nvidia's GB200 and GB300 chips this year [2]. - AWS is designing its own machine learning chip, "Trainium," which incorporates HBM, with the latest version, "Trainium 2," featuring HBM3 and HBM3E products [2]. Group 4: Upcoming Chip Developments - AWS plans to release the next-generation chip "Trainium 3" by the end of this year or next year, which will double the computing performance and improve energy efficiency by approximately 40%, with a total memory capacity of 144 GB [3].