SCOI Atlas · L0 Infrastructure
L0E Edge & On-Device Compute
NPUs in phones, laptops, cars, wearables and industrial devices.
TL;DR · Direct answer
L0E Edge & On-Device Compute is a sublayer of L0 Infrastructure in the Supply Chain of Intelligence™ (SCOI) by Anand Arivukkarasu. It is a required capability but not on its own a durable moat. If your product roadmap assumes cloud-only inference for consumer use cases, you are pricing yourself out of the next generation of surfaces.
What actually matters at L0E
- The default frontier user in 2027 will do most turns on-device.
- Battery and thermal are the real constraints, not FLOPs.
- OS vendors (Apple, Google, Microsoft) sit between you and the NPU.
The startup lens
If your product roadmap assumes cloud-only inference for consumer use cases, you are pricing yourself out of the next generation of surfaces.
Vertical lens — how this plays across categories
Consumer AI apps
Ship a hybrid model — small local + cloud fallback.
Automotive
Own the vehicle-side agent; cloud can only ever be a copilot to it.
Healthcare wearables
On-device inference resolves most HIPAA data-egress issues.
How to defend L0E
- Distilled proprietary models targeted at consumer NPUs.
- OS-relationship strategy alongside product strategy.
Other sublayers in L0 Infrastructure
L0A
Silicon & Memory
Accelerators (Nvidia, AMD, TPU, Trainium, custom), HBM, and the memory hierarchy that dominates training cost.
L0B
Data Centers
Powered shells, colo, hyperscale — the physical containers of intelligence.
L0C
Interconnect Fabric
NVLink, InfiniBand, Ethernet-based scale-out — the pipes between accelerators.
L0D
Compute & State Infrastructure
Vector DBs, object stores, feature stores, KV caches — persistent state around ephemeral compute.