{
 "id": "1.09",
 "slug": "leading-edge-foundry",
 "url": "/parts/compute-silicon-and-packaging/leading-edge-foundry",
 "name": "Leading-edge foundry",
 "qualifiers": null,
 "layer": {
  "id": 1,
  "name": "Compute silicon & packaging"
 },
 "parent_id": null,
 "children": [],
 "summary": "The advanced-node semiconductor fabs that manufacture the compute dies for essentially every AI accelerator on the market.",
 "description_md": "Foundry is where a chip design (from Merchant GPUs / accelerators, Hyperscaler custom ASICs, or Startup accelerators) becomes a physical wafer of dies. It sits upstream of silicon wafers as a raw material input and feeds directly into advanced packaging, where the finished die is combined with memory stacks and mounted onto a substrate. Almost every AI accelerator shipping today, regardless of which company designed it, is manufactured at the most advanced available process nodes, which means the finished-chip supply chain funnels through a small number of qualified fabs regardless of how many chip designers there are.\nThat concentration is the central fact investors need to hold onto for this part: demand growth for AI accelerators is, almost one for one, demand growth for advanced-node wafer capacity, and capacity additions at the most advanced process nodes take years and enormous capital, so the foundry side of the chain moves on a much slower clock than chip design announcements do. Qualification of a new process node is a multi-year undertaking for both the foundry and each customer's chip design, which makes switching suppliers between the handful of qualified providers a rare and consequential event rather than a routine one.",
 "ai_delta": "Every additional AI accelerator design and unit sold is another wafer order at the most advanced process nodes, concentrating incremental demand at a small number of qualified fabs.",
 "bottleneck_status": null,
 "scale_tiers": [
  "enterprise",
  "hyperscale_building",
  "ai_campus",
  "gw_class"
 ],
 "density_modes": [
  "any"
 ],
 "scale_notes": "Wafer volume ordered scales with total accelerator unit count across the industry, not with the size of any single data center buildout.",
 "flow": [
  "compute"
 ],
 "companies": [
  {
   "company_id": "intel",
   "name": "Intel",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "INTC",
   "product_note": "Foundry",
   "note": null,
   "source_url": null
  },
  {
   "company_id": "samsung-electronics",
   "name": "Samsung Electronics",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "005930.KS",
   "product_note": "Foundry",
   "note": null,
   "source_url": null
  },
  {
   "company_id": "tsmc",
   "name": "TSMC",
   "role": "manufacturer",
   "confidence": "inferred",
   "ownership_type": "public",
   "ticker": "2330.TW",
   "product_note": null,
   "note": "for essentially every AI accelerator shipping today",
   "source_url": null
  }
 ]
}