AIChipMapExport Controls

NVIDIA SUPPLY CHAIN TRACE

CHIP DESIGNERS🇺🇸 United States · NVDA · NASDAQChokepoint

NVIDIA designs the AI accelerators that power the global AI buildout, holding roughly 80% of the AI chip market with its H100, H200, and Blackwell B200 GPUs. Its CUDA software ecosystem creates powerful switching costs that reinforce its hardware dominance, while its supply chain — TSMC CoWoS capacity, SK Hynix HBM, Ibiden substrates — is itself a map of the most constrained nodes in AI infrastructure. US export controls have blocked NVIDIA from selling its most advanced chips to China, creating a cat-and-mouse cycle of downgraded variants and subsequent bans.

→ View company page

UPSTREAM — SUPPLIERS

Ibiden

Dominant supplier of IC substrates for advanced packages

Chokepoint🇯🇵
MATERIALSsupplies FC-BGA substrates to

TSMC

World's largest contract chip manufacturer

Chokepoint🇹🇼
FOUNDRIESfabricates GPU dies for

Samsung Foundry

Second-largest advanced foundry, competing at 3nm

🇰🇷
FOUNDRIEShistorically fabricated Ampere GPU dies for (moved to TSMC)

SK Hynix

Leading HBM3/HBM3E supplier for AI accelerators

Chokepoint🇰🇷
MEMORY (HBM)supplies HBM3E memory to

Samsung Memory

World's largest DRAM and NAND supplier, ramping HBM3

🇰🇷
MEMORY (HBM)supplies HBM3E memory to (qualified Q3 2025, ramp ongoing)

Micron

Only US-based DRAM and HBM maker

🇺🇸
MEMORY (HBM)supplies HBM3E memory to

Arm Holdings

Dominant CPU/NPU IP licensor; virtually all AI chips in mobile, edge, and cloud use Arm instruction set

Chokepoint🇬🇧
CHIP IPlicenses CPU/NPU IP to

Cadence

Leading EDA software provider; IC design tools (Virtuoso, Genus, Innovus) used by every advanced chip designer; US export-controlled to Chinese entities since 2022

Chokepoint🇺🇸
EDA TOOLSprovides EDA tools to

Synopsys

Largest EDA software provider; IC design and verification tools (Design Compiler, PrimeTime, VCS) essential for all advanced chip tapeouts; US export-controlled to Chinese entities

Chokepoint🇺🇸
EDA TOOLSprovides EDA tools to

ASE Technology

World's largest OSAT; packages AI chip dies from TSMC and Samsung for NVIDIA, AMD, Apple, and Qualcomm

🇹🇼
OSAT / PACKAGINGprovides packaged GPU modules to

Amkor Technology

Second-largest OSAT; packages dies for NVIDIA, Qualcomm, Apple, and Broadcom; operates advanced 2.5D/3D packaging for AI chips

🇺🇸
OSAT / PACKAGINGprovides packaged GPU modules to

Shinko Electric

Major IC substrate and lead frame supplier

🇯🇵
MATERIALSsupplies IC substrates to

↓ all supply to NVIDIA ↓

NVIDIA

Dominant AI GPU designer (H100, H200, B200)

UPSTREAM — SUPPLIERS ↑ · DOWNSTREAM — CUSTOMERS ↓
Chokepoint~80% AI accelerator market36–52 weeks
→ View company page

↓ NVIDIA supplies to ↓

Super Micro

Leading GPU server ODM for AI data centers

🇺🇸
SERVER ODMssupplies GPUs to server ODM

Wiwynn

Major ODM server manufacturer for hyperscalers

🇹🇼
SERVER ODMssupplies GPUs to server ODM

Quanta Computer

World's largest ODM by revenue; primary AI server manufacturer for Google, Meta, Amazon, and Microsoft

🇹🇼
SERVER ODMssupplies GPUs to server ODM

Foxconn Industrial Internet (FII)

Foxconn's AI server ODM division; assembles GPU clusters for hyperscalers and enterprise AI deployments globally

🇹🇼
SERVER ODMssupplies GPUs to server ODM

Dell Technologies

Major AI server vendor (PowerEdge XE9680); routes NVIDIA, AMD, and Intel Gaudi GPUs to enterprise customers; #2 AI server brand globally

🇺🇸
ENTERPRISEsupplies GPUs to AI server vendor

CoreWeave

GPU-native cloud provider; operates one of the world's largest NVIDIA H100/H200/B200 clusters for AI training and inference

🇺🇸
CLOUD PROVIDERSsupplies H200/B200 GPUs to GPU cloud provider

xAI

Elon Musk's AI company; built Colossus — world's largest GPU cluster (200k H200 + B200); develops Grok LLM; direct NVIDIA buyer at hyperscale

🇺🇸
AI CONSUMERSsupplies H200/B200 GPUs for Colossus cluster to

ByteDance

TikTok/Douyin parent; one of the world's largest AI compute buyers; cut off from NVIDIA H100/H200 by Oct 2023 BIS advanced computing rules

🇨🇳
CLOUD PROVIDERSsupplies H20 (export-controlled, downgraded GPU) to

Alibaba Cloud

China's #1 cloud provider; deploys Alibaba Hanguang 800 AI chip (TSMC-fabbed) and Huawei Ascend; operates Qwen LLM

🇨🇳
CLOUD PROVIDERSsupplies H20 (export-controlled, downgraded GPU) to

Tencent Cloud

China's #2 cloud provider; major AI GPU buyer pivoting to H20/Huawei Ascend after BIS export controls; operates Hunyuan LLM

🇨🇳
CLOUD PROVIDERSsupplies H20 (export-controlled, downgraded GPU) to

DOWNSTREAM — CUSTOMERS

About this company

QWhat is NVIDIA's role in the AI chip supply chain?

Dominant AI GPU designer (H100, H200, B200)

▲Allocation queue 6–9 months

View full company profile →