SK Hynix shipped samples of its next-generation HBM4E high-bandwidth memory chips to major customers on Thursday, June 13, 2026 [1, 2, 5].

The move is critical as the global race to supply AI accelerators intensifies. By delivering these samples, the company aims to secure its footprint in the fast-growing AI semiconductor market and meet the escalating demand for high-performance computing [2, 3].

The new HBM4E chips feature a 12-layer memory structure [4]. According to technical specifications, the chip capacity reaches 48 GB [4]. These hardware improvements are designed to support the massive data processing requirements of modern artificial intelligence models.

Efficiency is a primary focus of the new architecture. The company said that energy efficiency improves by over 20% compared to previous iterations [4]. This reduction in power consumption is a key metric for data centers managing the heat, and electricity costs of large-scale AI deployments.

Financial markets responded positively to the announcement. SK Hynix shares surged to a record high following the news that samples had been delivered [1].

While SK Hynix is moving aggressively, the competitive landscape remains tight. Some reports indicate that Samsung Electronics has also begun shipping HBM4E samples to its own customers, suggesting a simultaneous push by South Korean giants to lead the industry.

“Energy efficiency improves by over 20%”

The delivery of HBM4E samples signals a shift toward higher-density memory requirements for AI accelerators. As capacity reaches 48 GB and energy efficiency improves, the industry is moving away from raw power toward sustainable performance. The simultaneous push by both SK Hynix and Samsung indicates that the window for establishing a dominant standard in the next generation of AI memory is closing rapidly.