---
title: AI Supply Chain Bottlenecks | Sumedh Habbu
description: Explore an interactive map of the global AI supply chain, its concentrated dependencies, critical bottlenecks and disruption scenarios.
image: https://sumedhhabbu.in/hubfs/ai_supply_chain_bottlenecks_linkedin.png
---

[Sumedh Habbu](https://sumedhhabbu.in?hsLang=en-in) [← Visit my website](https://sumedhhabbu.in?hsLang=en-in)

An interactive systems map · November 2026

# The AI ecosystem has more than one choke point.

AI depends on a tightly coupled industrial stack: materials, chip design, manufacturing, compute, power, cooling, networks and cloud operations. Select a node to inspect its dependencies—or choose a disruption scenario to see where the shock could travel.

Explore the ecosystem Stress test: Taiwan / TSMC disruption Stress test: NVIDIA unavailable Stress test: Qatar helium disruption Stress test: power / transformer bottleneck Stress test: major cloud-region outage

Reset map

Critical concentration High exposure Medium / capability bottleneck Scenario-affected node

## Dependency map

Upstream → downstream

On smaller screens, swipe horizontally to explore the full map.

Critical materials& industrial gasesHelium · gallium · etc. EDA softwareSynopsys · CadenceSiemens EDA EUV opticsCarl Zeiss SMT EUV lithographyASML Leading-edgefabricationTSMC · Samsung · Intel HBM memorySK hynix · SamsungMicron AdvancedpackagingCoWoS · ASE · Intel AI accelerators& softwareNVIDIA · AMD · ASICs Networking& opticsSwitches · fibre · optics Power & gridTransformers · switchgear Cooling & waterCDUs · pumps · heat Cloud & datacentresAWS · Azure · GoogleCloud · OCI · others Data, evaluation& human feedbackData quality · expertise

Selected dependency

### Start with a node

Click any box in the map to see why it matters, who is involved and what alternatives exist.

## How to use this map

Choose a stress test above to highlight exposed parts of the ecosystem. The map is a conceptual dependency model, not a live supply-chain monitor.

**How to read the risk tiers:** qualitative editorial assessment of concentration, substitutability and time-to-recover—not a probability score or a claim that a supplier is literally irreplaceable. Supplier lists are illustrative, not exhaustive. Dependencies vary by chip design, geography, workload and data-centre architecture. “Unavailable” does not mean already-installed equipment instantly stops working.