# What Is HBM (high-bandwidth memory)?

> HBM (high-bandwidth memory) is the stacked, high-bandwidth memory used on AI accelerators, in other words GPU memory. It is extremely fast but limited in capacity and expensive, and model weights and KV cache compete for it.

URL: https://vuke.top/en/glossary/hbm · Updated November 12, 2026

HBM (high-bandwidth memory) is the stacked, high-bandwidth memory used on AI accelerators, in other words GPU memory. It is extremely fast but limited in capacity and expensive, and model weights and KV cache compete for it.

## In the measurements

Each MI308X card in the test platform on this site has 192 GB of HBM. Report R5 argues that an external KV tier can stand in for 128 GB-class HBM residency (MetaX N260 platform) [R5](https://vuke.top/en/evidence/R5).

Source: [R5](https://vuke.top/en/evidence/R5)
