Home > Business > Industry

Samsung ships HBM4E samples ahead of schedule

Published: 29 May. 2026, 10:31 Updated: 29 May. 2026, 13:46

LEE JAE-LIM
[email protected]

Samples of Samsung Electronics' next-generation 12-layer high bandwidth memory, HBM4E [SAMSUNG ELECTRONICS]

Samsung Electronics has begun shipping samples of its next-generation 12-layer high bandwidth memory (HBM), HBM4E, to customers.

Following sample supply and optimization, the chipmaker plans to begin mass production based on customer demands.

The sample shipments began earlier than Samsung initially forecast, as it originally expected delivery to take place in the middle of this year. The accelerated timeline reflects the company's push to stay ahead in the increasingly competitive AI memory market.

“Following the successful mass production of HBM4, Samsung has once again demonstrated its distinct technological edge with HBM4E,” said Hwang Sang-joon, the executive vice president and head of memory development at Samsung Electronics. “Through our advanced manufacturing capabilities and pre-emptive infrastructure investments, we will continue to drive the growth of the global AI memory market.”

Samples of Samsung Electronics' next-generation 12-layer high bandwidth memory, HBM4E [SAMSUNG ELECTRONICS]

The product's per-pin operating speed ranges from 14 gigabits per second (Gbps) to a maximum speed of 16 Gbps. It also delivers 3.6 terabytes per second of bandwidth per single stack, helping to improve computing performance for large language models and next-generation AI systems.

The chip's capacity has also been upgraded to 48GB, 30 percent larger than its predecessor. The product lineup will be expanded to include an 8-layer variation with 32GB and a 16-layer variation with 64GB capacity to meet a range of customer needs.

The product applies 10-nanometer-class 1c dynamic random-access memory, or DRAM, and a 4-nanometer logic process, both proprietary chip manufacturing technologies of Samsung.

According to the company, the product's low-power design and optimized packaging structure have improved energy efficiency by 16 percent and thermal resistance by more than 14 percent compared to the previous generation, addressing challenges with heat dissipation and lowering energy consumption in AI data centers with intensive workloads. As AI clusters grow in scale and density, those efficiency gains are expected to translate into meaningful cost savings for hyperscalers and cloud providers.

Samsung also plans to expand its supply of HBM4, which began shipping in February. The company said that customer response has been “highly positive” in terms of performance and energy efficiency.

Samsung Electronics' next-generation 12-layer high bandwidth memory, HBM4E, right, next to a 500 won coin for scale [SAMSUNG ELECTRONICS]

Its HBM revenue is expected to triple compared to last year, driven by the AI boom, and the chipmaker is rushing to ramp up production capacity at its Pyeongtaek chip complex in Gyeonggi. Construction of its P5 factory, which had been halted since the semiconductor downturn, resumed this year, with the completion timeline accelerated by roughly six months compared to earlier plans. P6, its sixth and final factory at the Pyeongtaek cluster, is also expected to break ground later this year, six months ahead of the original schedule of early 2027.

Additionally, Samsung will spend at least 110 trillion won ($73.3 billion) on facilities and research and development this year — a record-high expenditure for the company — and specifically noted that funds will mainly go toward high-performing HBMs. The investment signals Samsung's confidence that demand for AI memory will remain robust in the coming years.

BY LEE JAE-LIM [[email protected]]