Physical AI infrastructure is not a single market. It is a stack of interdependent layers, each essential, each with distinct investment characteristics. Understanding the structure of the stack — and the dependencies between layers — is the foundation of a coherent investment and business thesis in this space.
THE STACK VISUALISED
LAYER 5 — PERCEPTION: LiDAR, cameras, HD mapping, spatial AI
LAYER 4 — SAFETY SOFTWARE: Certified OS, functional safety, cybersecurity
LAYER 3 — TEST & MEASUREMENT: Chip testing, network validation, system certification
LAYER 2 — CONNECTIVITY: 5G/6G, private networks, industrial protocols
LAYER 1 — COMPUTE: Edge chips, AI accelerators, memory
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
FOUNDATION — PHYSICAL WORLD: The environment Physical AI operates in
The numbering is intentional. Layer 6 sits at the top because it is the ultimate enabler — without power, every other layer is inert. Layer 1 sits above the physical world because compute is what bridges the gap between raw sensor data and intelligent action.
LAYER 1 — COMPUTE
What it is: The intelligence processing at the edge — AI inference chips, microcontrollers, neural processing units embedded in Physical AI devices.
Key players: ARM (architecture licensing — present in virtually every edge device), Qualcomm (integrated SoCs combining compute, modem and AI acceleration), MRAM specialists (non-volatile memory for edge persistence).
Investment characteristic: High volume, competitive, subject to rapid capability improvement. The Jevons Paradox applies — more efficient chips accelerate deployment rather than reducing demand.
LAYER 2 — CONNECTIVITY
What it is: The communication infrastructure that allows Physical AI devices to share data, receive updates and coordinate with cloud and edge intelligence.
Key players: Cisco (enterprise and industrial networking), Nokia (private 5G for industrial environments, 6G development), Belden (industrial cables and connectivity).
Investment characteristic: Infrastructure-like, long replacement cycles, 6G transition creates upgrade demand through 2030. Private 5G networks for smart factories are a high-growth segment.
LAYER 3 — TEST AND MEASUREMENT
What it is: The validation infrastructure that certifies every Physical AI component before deployment. Semiconductors must be tested. Networks must be validated. Safety-critical systems must be certified.
Key players: Keysight Technologies (electronic test equipment, network testing), Teradyne (semiconductor test, industrial automation testing).
Investment characteristic: Often overlooked but structurally essential. Every new chip generation, every new network standard, every new Physical AI application creates test demand. Revenue is less cyclical than semiconductor production because testing occurs throughout the product lifecycle.
New chip architectures require new test equipment
6G transition requires entirely new RF test infrastructure
Autonomous vehicle certification is test-intensive
Revenue grows with Physical AI deployment, not just production
LAYER 4 — SAFETY-CRITICAL SOFTWARE
What it is: The software foundation for Physical AI systems operating in environments where failure has physical consequences — vehicles, industrial machinery, medical devices, infrastructure.
Key players: BlackBerry QNX (certified RTOS in 235M+ vehicles, expanding to industrial and robotics), plus specialised cybersecurity for connected physical systems.
Investment characteristic: High switching costs, long certification cycles create durable competitive positions. QNX's presence in automotive creates a platform for expansion into adjacent Physical AI markets.
LAYER 5 — PERCEPTION AND SPATIAL MAPPING
What it is: The sensory layer that converts the physical world into machine-readable data — cameras, LiDAR, radar, and the mapping intelligence that contextualises sensor data.
Key players: TomTom (HD mapping for autonomous navigation), Aeva (FMCW LiDAR delivering 4D perception including velocity), plus the broader sensor ecosystem.
Investment characteristic: Technology differentiation matters significantly. FMCW LiDAR's velocity measurement capability addresses limitations of time-of-flight systems. Mapping data has network effects — more vehicles generating data improve maps for all users.
LAYER 6 — POWER INFRASTRUCTURE
What it is: The electrical infrastructure that enables every other layer — grid connections, high-voltage transformers, EV charging networks, industrial power upgrades, data center power supply.
Key players: Grid operators (RTE, National Grid), transformer manufacturers (ABB, Siemens Energy, Pauwels), electrical contractors, site developers with confirmed grid capacity.
Investment characteristic: Scarcest layer. Longest lead times. Cannot be substituted or accelerated by software. Grid-connected assets with confirmed capacity are permanent scarcity plays regardless of which Physical AI applications win.
THE DEPENDENCY STRUCTURE
Each layer depends on all layers below it. Layer 5 (perception) is useless without Layer 1 (compute) to process sensor data and Layer 6 (power) to operate. Layer 2 (connectivity) requires Layer 6 power for every base station, router and switch. Layer 3 (testing) requires Layer 6 power for every test facility.
This dependency structure has a critical investment implication: the scarcity of Layer 6 creates a binding constraint on the deployment of all other layers. You can have the most advanced LiDAR, the fastest 6G network and the most efficient AI chip — none of it deploys without power infrastructure.
THE GRIDREADINESS POSITION
GridReadiness focuses on Layer 6 — the power infrastructure layer that controls all others. We provide intelligence, site identification and equipment sourcing for the physical AI economy's most constrained resource. Our newsletter covers developments across all six layers, with depth analysis on the power infrastructure implications of Physical AI deployment at scale.