AI Factory Supply Chain Visualization

The AI Factory

From silicon to skyscrapers — explore the massive global supply chain powering the AI revolution

50+

Chip Companies

Design & fabrication

200+

Datacenter Operators

Colocation & cloud

1GW+

Power & Cooling

Per mega-campus

50+

Construction Firms

Specialized builders

The AI Factory Supply Chain

Every AI model runs on a physical stack spanning 7 layers — each with specialized companies that must work in perfect synchronization to deliver exascale compute

Chip Design

Architecture, RTL, verification

NVIDIA AMD Google TPU Amazon Trainium Microsoft Maia Cerebras Groq Graphcore

Fabrication

Wafer fab, EUV lithography

TSMC (3nm/5nm) Samsung Foundry Intel Foundry GlobalFoundries

Advanced Packaging

HBM integration, 2.5D/3D stacking

TSMC CoWoS Amkor ASE Intel EMIB/Foveros Samsung I-Cube

High-Bandwidth Memory

HBM3/3E/4, 1.2TB/s+ bandwidth

SK Hynix (HBM3E) Samsung (HBM3E) Micron (HBM3E)

Server & Rack

HGX/H100/B200 platforms, liquid cooling

Foxconn Quanta Wistron Inventec Supermicro Dell HPE Lenovo

Datacenter Campus

Colocation, GPU clouds, 100MW+ campuses

Equinix Digital Realty CoreWeave Lambda Labs Crusoe Applied Digital QTS CyrusOne

Power & Utilities

GW-scale, 24/7 carbon-free, 500kV substations

Nuclear (SMRs) Solar + Storage Wind Natural Gas Grid operators

Construction & EPC

EPC, specialized MEP, 18-month builds

Turner Hensel Phelps Mortenson DPR Jacobs Bechtel Fluor Black & Veatch

Semiconductor Layer

The Silicon Foundation

  • NVIDIA: Blackwell B200, H100, GB200 NVL72
  • TSMC: 3nm/5nm EUV, CoWoS-L/S/R
  • SK Hynix: HBM3E 12-hi, HBM4 roadmap
  • ASML: High-NA EUV, Twinscan EXE:5000
  • Broadcom: Networking, custom ASICs
  • Cadence/Synopsys: EDA tools, IP
Explore Semiconductor Layer

Infrastructure Layer

Datacenter & Cloud

  • Equinix/Digital Realty: 300+ locations globally
  • CoreWeave: 28K+ GPUs, $7B+ valuation
  • Lambda Labs: GPU cloud, on-demand clusters
  • Crusoe: Flared gas powered, climate-aligned
  • Microsoft/Azure: Maia 100, NDv5-series
  • AWS: Trainium2, UltraClusters, P5en
Explore Infrastructure Layer

Construction Layer

Building the Factory

  • Turner Construction: $10B+ data center portfolio
  • Hensel Phelps: Hyperscale specialists
  • Jacobs/Bechtel: EPC, nuclear integration
  • Vertiv/Schneider: Power & thermal management
  • Delta/AAVID: Liquid cooling, CDU/manifolds
  • Substations: 500kV, 1-2 year lead times
Explore Construction Layer

Critical Dependencies & Bottlenecks

The AI factory is only as fast as its slowest link — these are the choke points determining global AI capacity

EUV Lithography

~180 tools globally

ASML High-NA EUV: $380M each, 1-2 year waitlist. TSMC consumes ~70% of global supply.

ASML TSMC Intel Samsung

CoWoS Packaging

Capacity sold out 2025

TSMC CoWoS-L/S/R: 2.5D/3D stacking for HBM. NVIDIA B200 needs CoWoS-L. 2-year backlog.

TSMC Amkor ASE NVIDIA AMD

HBM Memory

3 suppliers, 100% allocated

SK Hynix, Samsung, Micron. HBM3E 12-hi: 1.2TB/s. HBM4 (2025) needs new fab investment.

SK Hynix Samsung Micron NVIDIA

Power Delivery

1-2 GW per campus

500kV substations: 18-24 month lead. SMRs: 2030+ deployment. Grid interconnection queues: 5+ years.

Utilities GE Vernova Siemens Energy NuScale Oklo

Ready to Explore the Full Supply Chain?

Dive deep into each layer with interactive diagrams, company profiles, investment data, and real-time capacity tracking.