SK hynix, Samsung, and Micron fighting for NVIDIA supply contracts for new 16-Hi HBM4 orders

Samsung, SK hynix, and Micron are all fighting each other in developing new 16-Hi HBM, because NVIDIA requested supply of the new memory chips for the second half of 2026.

2

16-Hi HBM hasn't been commercialized previously with many technological hurdles to overcome, including DRAM stacking, as things get far more complicated with more HBM stacks. In a new report from the Electronic Times, NVIDIA reportedly requested that domestic and foreign memory manufacturers deliver 16-Hi HBM memory chips by Q4 2026.

SK hynix and Samsung Electronics in South Korea, as well as US-based Micron, have all begun full-scale development work for the mass production supply of 16-Hi HBM memory chips to NVIDIA. The outlet reports that concrete contracts haven't been signed yet; it reports that discussions are happening internally regarding the initial production volumes of 16-Hi HBM chips.

An industry official said: "Following the supply of 12-Hi HBM4, a request for 16-HI supply has also been made, so we are establishing a very fast development schedule. Performance evaluation could begin as early as before the third quarter of next year".

• Read more: SK hynix shows off 16-Hi HBM3E memory chips for new AI GPUs at CES 2025
• Read more: SK hynix showcases world's first HBM4: 16-Hi, 2TB/sec bandwidth, TSMC logic die
• Read more: Future-gen HBM4, HBM5, HBM6, HBM7, and HBM8 teased with 15,000W AI GPUs by 2038

12-Hi HBM4 will enter mass supply in early 2026, but development of the denser 16-Hi HBM4 memory chips has accelerated. Another high-ranking industry official suggested that we might see 16-Hi HBM called HBM4E, adding: "It is likely that the product will be 16-HI HBM4, but the generation or name could change depending on performance and mass production timing".

16-Hi HBM stacks are extremely complicated to fab, being the densest amount of HBM stacks so far (12-Hi is here, 16-Hi is accelerating development) with even next-gen HBM5 memory maxing out at 16-Hi. We won't get to 20-Hi and 24-Hi HBM memory stacks until HBM7, which is expected sometime in 2035, while future-gen HBM8 is also expected to max out at 24-Hi HBM stacks.

The new 16-Hi HBM memory chips being worked on in 2026 are an industry-first for the semiconductor world, with observations suggesting technical challenges will be "formidable", with concerns around the thin wafer thickness. Existing 12-HI HBM has a wafer thickness of about 50 micrometers (㎛), with the report suggesting this must be reduced to around 30㎛ to stack 16 layers.

• Read more: Micron ships 11Gbps HBM4 and is working on even faster HBM4E
• Read more: Samsung accelerates HBM4E process: aims for 3.25TB/sec bandwidth for NVIDIA Rubin AI GPUs

The HBM4 thickness standard defined by JEDEC (Joint Electron Device Engineering Council) is 775㎛. It is about 50㎛ thicker than the existing HBM3E (725㎛). This leeway was given because implementing thin HBM becomes more difficult as the number of layers increases.

However, the prevailing opinion is that JEDEC will not further increase the thickness for this 16-HI HBM. This means more DRAMs must be stacked within 775㎛.

Currently, Samsung Electronics and Micron are proceeding with the bonding process using Thermal Compression Non-Conductive Film (TC-NCF), while SK Hynix uses Mass Reflow Molded Underfill (MR-MUF).

An industry official said: "The current bonding material thickness is around 10㎛, and to stack more DRAMs, this must be reduced to secure competitiveness. Dissipating the heat generated after bonding is also a challenge".