VLSI KoreaTSMC GIGAFAB Facilities
TSMC identifies Fab 12 among the six GIGAFAB facilities it operates.
- Published
- Not stated
- Retrieved
- 2026-10-14
Semiconductor graph
NVIDIA25 entities and 26 documented connections shown.
Drag the map or use pan buttons to move. Tab to a node or connection and press Enter to inspect its evidence. Arrows follow the actual claim direction.
CompanyChip / productTechnologyPlace / facility
Select a node to reveal relationship labelsRelationships & evidence (26)
| Subject | Relationship | Object | Status | Evidence |
|---|---|---|---|---|
| TSMC Fab 12 | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Fab 12 as a named GIGAFAB facility; operator identity only | |
| AMD Instinct MI300X | AMD | SUPPORTEDSource-reported snapshot: 2023-12-06 | Why are these connected?Product authorship statement; not a manufacturing claim. AMD Delivers Leadership Portfolio of Data Center AI Solutions with AMD Instinct MI300 SeriesAMD presents MI300X as an AMD Instinct accelerator.
| |
| Intel | AMD | SUPPORTEDSource-reported snapshot: 2026-02-04 | Why are these connected?This is AMD's self-reported, segment-specific competition statement. AMD 2025 Annual Report on Form 10-KAMD states that it competes primarily with Intel in Data Center for its CPU, GPU, DPU, and AI NIC server products.
| |
| NVIDIA GA100 GPU | TSMC 7nm N7 | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?Evidence names the process and TSMC; it does not identify a fab location. NVIDIA Ampere Architecture In-DepthNVIDIA identifies the GA100 GPU that powers A100 as fabricated on TSMC's 7nm N7 process.
| |
| TSMC Fab 22 | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Fab 22 as a named GIGAFAB facility; operator identity only TSMC GIGAFAB FacilitiesTSMC identifies Fab 22 among the six GIGAFAB facilities it operates.
| |
| NVIDIA A100 Tensor Core GPU | HBM2 | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?This is the cited 40 GB configuration; later A100 variants use HBM2e. NVIDIA Ampere Architecture In-DepthNVIDIA specifies 40 GB of HBM2 for the A100 discussed in the architecture article.
| |
| NVIDIA GH100 GPU | NVIDIA H100 Tensor Core GPU | SUPPORTEDSource-reported snapshot: 2022-03-22 | Why are these connected?GH100 die in the H100 product family. NVIDIA Hopper Architecture In-DepthNVIDIA describes GH100 as the GPU implementation underlying H100.
| |
| TSMC Fab 20 | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Fab 20 as a named GIGAFAB facility; operator identity only TSMC GIGAFAB FacilitiesTSMC identifies Fab 20 among the six GIGAFAB facilities it operates.
| |
| NVIDIA | Samsung Electronics | SUPPORTEDValidity dates not established | Why are these connected?Portfolio-level company claim; it does not associate Samsung with Blackwell, a specific process, or a physical fab. NVIDIA 2025 Annual Report on Form 10-KNVIDIA reports using Samsung to produce its semiconductor wafers.
| |
| TSMC Fab 18 | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Fab 18 as a named GIGAFAB facility; operator identity only TSMC GIGAFAB FacilitiesTSMC identifies Fab 18 among the six GIGAFAB facilities it operates.
| |
| NVIDIA A100 Tensor Core GPU | NVIDIA third-generation NVLink | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?Interconnect availability and topology can vary by A100 form factor/system. NVIDIA Ampere Architecture In-DepthNVIDIA says third-generation NVLink is implemented in A100 GPUs.
| |
| NVIDIA | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Portfolio-level company claim; it does not identify a product, process, fab, wafer allocation, or duration. NVIDIA 2025 Annual Report on Form 10-KNVIDIA reports using TSMC to produce its semiconductor wafers.
| |
| NVIDIA GH100 GPU | TSMC 4N | SUPPORTEDSource-reported snapshot: 2022-03-22 | Why are these connected?Evidence names the process and TSMC; it does not identify a fab location. NVIDIA Hopper Architecture In-DepthNVIDIA says the GH100 GPU powering H100 is fabricated using TSMC 4N, customized for NVIDIA.
| |
| NVIDIA GH100 GPU | NVIDIA | SUPPORTEDSource-reported snapshot: 2022-03-22 | Why are these connected?GH100 die. NVIDIA Hopper Architecture In-DepthNVIDIA describes GH100 as its Hopper GPU implementation.
| |
| TSMC Fab 15 | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Fab 15 as a named GIGAFAB facility; operator identity only TSMC GIGAFAB FacilitiesTSMC identifies Fab 15 among the six GIGAFAB facilities it operates.
| |
| NVIDIA Blackwell architecture | NVIDIA | SUPPORTEDSource-reported snapshot: 2024-03-18 | Why are these connected?Architecture statement; not a claim about every Blackwell SKU. NVIDIA Blackwell Platform Arrives to Power a New Era of ComputingNVIDIA presents Blackwell as its new GPU architecture.
| |
| NVIDIA GA100 GPU | NVIDIA A100 Tensor Core GPU | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?GA100 die and A100 product identity. NVIDIA Ampere Architecture In-DepthNVIDIA states that GA100 powers the A100 GPU product.
| |
| NVIDIA A100 Tensor Core GPU | NVIDIA | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?Product authorship statement; not a fabrication or supply-chain claim. NVIDIA Ampere Architecture In-DepthNVIDIA presents A100 as its GPU product based on its Ampere GPU architecture.
| |
| NVIDIA Blackwell wafers | TSMC | SUPPORTEDSource-reported snapshot: 2026-08-05 | Why are these connected?This is a Blackwell-wafer claim; it must not be generalized to NVIDIA's entire wafer portfolio. NVIDIA and Partners Build in America, for AmericaNVIDIA directly attributes volume Blackwell-wafer manufacturing at the named TSMC factory to TSMC.
| |
| NVIDIA H100 Tensor Core GPU | NVIDIA | SUPPORTEDSource-reported snapshot: 2022-03-22 | Why are these connected?Product authorship statement; not a manufacturing claim. NVIDIA Announces Hopper Architecture, the Next Generation of Accelerated ComputingNVIDIA announces H100 as an NVIDIA GPU product.
| |
| AMD | NVIDIA | SUPPORTEDSource-reported snapshot: 2026-02-04 | Why are these connected?This is AMD's self-reported, segment-specific competition statement; it is not a claim that every AMD product competes with every NVIDIA product. AMD 2025 Annual Report on Form 10-KAMD states that it competes primarily with NVIDIA in Data Center for its CPU, GPU, DPU, and AI NIC server products.
| |
| TSMC Fab 14 | TSMC | SUPPORTEDValidity dates not established | Why are these connected?Fab 14 as a named GIGAFAB facility; operator identity only TSMC GIGAFAB FacilitiesTSMC identifies Fab 14 among the six GIGAFAB facilities it operates.
| |
| NVIDIA GA100 GPU | NVIDIA | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?GA100 die. NVIDIA Ampere Architecture In-DepthNVIDIA identifies GA100 as its Ampere-architecture GPU.
| |
| NVIDIA A100 Tensor Core GPU | NVIDIA Ampere architecture | SUPPORTEDSource-reported snapshot: 2020-05-14 | Why are these connected?Applies to A100, not every Ampere product. NVIDIA Ampere Architecture In-DepthNVIDIA states that A100 is based on the Ampere GPU architecture.
| |
| NVIDIA Blackwell wafer produced in the U.S. (2025 milestone) | TSMC | SUPPORTEDSource-reported snapshot: 2025-10-17 | Why are these connected?This establishes the documented U.S. Blackwell-wafer milestone, not every Blackwell wafer or SKU. The Engines of American-Made Intelligence: NVIDIA and TSMC Celebrate First NVIDIA Blackwell Wafer Produced in the USNVIDIA reports the first NVIDIA Blackwell wafer was produced on U.S. soil at TSMC's Phoenix manufacturing facility.
| |
| NVIDIA H100 Tensor Core GPU | NVIDIA Hopper architecture | SUPPORTEDValidity dates not established | Why are these connected?The page is current and undated; use launch material for an exact announced date. NVIDIA H100 GPUNVIDIA says H100 is based on the Hopper architecture.
|
Find an evidence path
Find a documented path
From NVIDIA · up to two relationship steps.
A path retains its original claims. Shared technology does not establish a supply contract.
25 nodes · 26 claims
Scope
Relationships may describe different dates. A path does not establish simultaneous activity. Bounded source-backed collection. Missing or filtered connections do not establish absence. Supported does not mean independently verified.