The first 12-layer HBM4 samples shipped early, offering 2TB/s speed, 36GB capacity, improved heat control, and faster AI memory performance for future use.
SK hynix Inc. have launched the world’s first samples of 12-layer HBM4, a high-performance DRAM designed for AI applications, to major customers. These early deliveries were made ahead of schedule, supported by the company’s experience and leadership in the HBM market. Certification processes with customers are now underway.
The 12-layer HBM4 offers industry-leading capacity and speed, with a data bandwidth exceeding 2TB per second—enough to handle over 400 full-HD movies in just one second. This marks a 60% speed increase over the previous HBM3E generation.
Using its Advanced MR-MUF process, the company claims to achieve a 36GB capacity while enhancing heat dissipation and chip stability, minimizing warpage. This technology was previously proven in earlier HBM3E production.
“We have enhanced our position as a front-runner in the AI ecosystem following years of consistent efforts to overcome technological challenges in accordance with customer demands,” said Justin Kim, President & Head of AI Infra at SK hynix. “We are now ready to smoothly proceed with the performance certification and preparatory works for mass production, taking advantage of the experience we have built as the industry’s largest HBM provider.”
Some of the unique features of the 12-layer HBM4 include:
- First in the world to ship 12-layer HBM4 samples
- Over 2TB per second data speed (60% faster than HBM3E)
- Can handle over 400 full-HD movies in one second
- 36GB capacity – highest for any 12-layer HBM
- Uses Advanced MR-MUF tech to reduce chip bending
- Better heat control for more stable performance
- Built on SK hynix’s proven HBM experience
SK hynix Inc. is a company from Korea. It makes memory chips called DRAM and NAND flash, used in devices. The company sells these chips to customers around the world. Its stock is traded in Korea and listed in Luxembourg.
For more information, click here.
