SK hynix Inc. shipped samples of its next‑generation HBM4E DRAM to major AI customers, delivering the 12‑layer stack on schedule. The move follows the company’s established expertise in HBM3 and HBM4 production. By shipping samples now, SK hynix signals readiness to scale the new memory for high‑performance data centers.

The 12‑layer HBM4E achieves a peak data rate of 16 Gbps per pin, a boost that cuts transfer latency and raises throughput for AI training and inference. Power efficiency climbs more than 20 percent over previous models, while a 48 GB capacity in a single stack delivers unprecedented density. These gains reduce energy draw in large‑scale compute clusters.

SK hynix uses Advanced MR‑MUF technology to maintain structural stability and improve heat resistance by 17 percent versus the prior HBM4, ensuring reliable operation under heavy workloads. With a proven track record in mass production, the firm plans to partner closely with customers to move the technology into mass‑produced chips. The launch tightens the memory supply chain for AI systems and accelerate deployment today.