{
 "id": "2.01",
 "slug": "hbm-hbm3e-hbm4",
 "url": "/parts/memory-and-storage/hbm-hbm3e-hbm4",
 "name": "HBM",
 "qualifiers": "HBM3E, HBM4",
 "layer": {
  "id": 2,
  "name": "Memory & storage"
 },
 "parent_id": null,
 "children": [],
 "summary": "High-bandwidth memory stacked directly on the same package as an AI accelerator, giving the GPU or ASIC very fast, close-range access to the memory it needs to stay fed.",
 "description_md": "HBM sits on the accelerator package itself, next to the compute die, rather than on a separate module across the board — the physical packaging that puts it there belongs to Layer 1. It is the first thing a new accelerator generation is designed around, since the number of stacks and their capacity set a hard ceiling on how large a model an accelerator can hold in memory before it has to reach further out to DRAM. AI training and inference workloads scale primarily by adding parameters and context length, both of which are bound by per-accelerator memory capacity, so HBM content has risen with every accelerator generation. The taxonomy flags HBM as **the most supply-constrained component in the stack**, and also notes that HBM4 base die is fabricated at a foundry partner for some suppliers or produced in-house by others — a qualification detail worth tracking, since a supplier's base-die sourcing shapes its cost position and lead times. For an investor, the things to watch are supplier qualification status on each new accelerator design win, and how stacking-tool and base-die capacity trail the memory makers' own output.",
 "ai_delta": "Larger models and longer context windows push accelerator designs toward more HBM stacks and higher per-stack capacity with each generation, keeping HBM the tightest link in the memory chain.",
 "bottleneck_status": "constrained",
 "scale_tiers": [
  "edge",
  "enterprise",
  "hyperscale_building",
  "ai_campus",
  "gw_class"
 ],
 "density_modes": [
  "any"
 ],
 "scale_notes": "Content is fixed per accelerator package; total HBM demand scales with the number of accelerators deployed rather than with facility size or cooling design.",
 "flow": [
  "compute",
  "data"
 ],
 "companies": [
  {
   "company_id": "besi",
   "name": "BESI",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "BESI.AS",
   "product_note": null,
   "note": null,
   "source_url": null
  },
  {
   "company_id": "disco",
   "name": "Disco",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "6146.T",
   "product_note": null,
   "note": null,
   "source_url": null
  },
  {
   "company_id": "hanmi-semiconductor",
   "name": "Hanmi Semiconductor",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "042700.KS",
   "product_note": null,
   "note": "stacking tools",
   "source_url": null
  },
  {
   "company_id": "micron",
   "name": "Micron",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "MU",
   "product_note": null,
   "note": null,
   "source_url": null
  },
  {
   "company_id": "samsung-electronics",
   "name": "Samsung Electronics",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "005930.KS",
   "product_note": null,
   "note": null,
   "source_url": null
  },
  {
   "company_id": "sk-hynix",
   "name": "SK Hynix",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "000660.KS",
   "product_note": null,
   "note": "leader",
   "source_url": null
  }
 ]
}