Independent analysis of data centre power architecture, autonomous operations, and the infrastructure behind AI. By Adi Kumar.

The AI Power Chain

The AI Power Chain · six-layer market map

The AI Power Chain

A six-layer market map of AI data-centre power infrastructure, from grid interconnect through on-package delivery. Each layer has a market essay covering vendor structure + economics and a technical companion covering the underlying engineering. Reference material for practitioners specifying next-generation AI power architecture, PE sponsors underwriting vendor exposure, and hyperscaler strategy teams tracking supply concentration.

6market-side stacks
6technical companions
3cross-cutting essays

01The stacks (market essays)

  1. 01
    The Interconnect Stack
    Grid interconnect, transformer + switchgear + cable ecosystem
  2. 02
    The Thermal Stack
    DLC + immersion cooling: heat as binding constraint above 100 kW/rack
  3. 03
    The Wide-Bandgap Stack
    SiC + GaN: boom, Wolfspeed restructuring, AI-power rebuild
  4. 04
    The Capacitor Stack
    MLCC + supercapacitor + silicon-cap categories that bind AI supply
  5. 05
    The On-Package Delivery Stack
    How 48V becomes 0.8V at 2,250 A: multi-phase PoL + MPS/Vicor/Empower
  6. 06
    The Modular Datacenter Stack
    Prefab enclosures, skid-mounted power + cooling, on-site assembly

02Technical companions (engineering deep-dives)

  1. 01
    Interconnect: Technical Companion
    Transformer physics + switchgear + grid interconnection deep-dive
  2. 02
    Thermal: Technical Companion
    DLC vs immersion + CDU architecture + chemistry supply chain
  3. 03
    Wide-Bandgap: Technical Companion
    SiC + GaN device physics, foundry economics, packaging materials
  4. 04
    Capacitor: Technical Companion
    HSC + EDLC + MLCC + silicon-cap electrical characteristics
  5. 05
    On-Package Delivery: Technical Companion
    Vertical vs lateral PoL, on-chip cap physics, packaging tolerances
  6. 06
    Modular Datacenter: Technical Companion
    Prefab economics + on-site assembly + module-integration standards

03Cross-cutting essays

  1. 01
    The Vendor Screen
    58 vendors across every layer with moat + execution + chokepoint scoring
  2. 02
    Pricing Under Scarcity
    Where premium capture is compounding across the six layers
  3. 03
    The Services Inversion
    Why installed-base annuities now command product-like multiples

Also see: The DC-DC Transition series

13 essays on the 800V DC data-centre transition, the parallel architectural shift feeding the AI Power Chain's power-delivery layers.

DC-DC Transition →
Related hubs: The Investment Layer (capital-allocation essays), Case Studies (deal retrospectives), Tech Spotlights (vendor deep-dives).
FrameworkThe AI Power Chain framework, canonical definition of the six layers, with vendor concentration, standards state and lead-time exposure per layer.
Definition The AI Power Chain is a six-layer market and engineering map of AI data centre power delivery, from grid interconnection through facility distribution, rack shelves, on-package delivery, thermal management and modular build. This series covers 15 essays: six stack maps, six technical companions, and three cross-cutting analyses.