Research Note: The Minimum Off-Grid Stack Has Five Budgets
Published July 2026
Question
What is the smallest stack that remains useful when the network is gone? Pockot should not answer with one chip. A durable pocket AI needs compute, memory, storage, energy, and trust budgets at the same time.
Source-Backed Data Points
- FAA PackSafe guidance identifies 100 Wh as a common boundary for lithium-ion batteries carried by airline passengers, with larger batteries requiring airline approval. Source: FAA Batteries.
- NVIDIA lists the Jetson Orin Nano Super Developer Kit at 67 INT8 TOPS, 8 GB LPDDR5 memory, storage through SD card and external NVMe, and a 7 W to 25 W power range. Source: NVIDIA Jetson Orin Nano Super.
- Raspberry Pi documentation lists AI HAT+ 2 at 40 TOPS with 8 GB onboard memory and local LLM/VLM support. Source: Raspberry Pi AI HAT documentation.
- Meta describes Llama 3.2 1B and 3B as edge and mobile models. Source: Meta Llama 3.2.
Reading
The minimum stack is not the fastest stack. It is the smallest configuration that can answer useful questions, retrieve local sources, survive a power budget, and explain what it can and cannot do. That means a 100 Wh battery, a small local model, a curated corpus, a writable user notebook, and a restrained tool layer may be more meaningful than a bigger accelerator with no storage discipline.
Five budgets interact. Compute determines possible model speed. Memory determines model and context fit. Storage determines source packs and indexes. Energy determines runtime and duty cycle. Trust determines whether the outputs are usable without network verification.
This is why Pockot's calculator should not produce one "good/bad" number. It should show which budget is currently the limiter. A stack can be compute-rich but energy-poor, storage-rich but source-stale, or memory-limited even when TOPS looks high.
Tool Rule
Version 1 should label bottlenecks by budget. The minimum viable stack is the smallest stack with no red budget for the intended scenario, not the stack with the largest TOPS value.