Edge AI Investing Guide: Where Capital Goes After Cloud AI
Damnang
· Damnang’s Substack
· May 29, 2026 at 07:15
· ⏱ 8 min read
| Read on Substack ↗
Summary
Edge AI is emerging as a structural investment theme distinct from cloud AI, with capital shifting from data-center-centric spending to device-level semiconductor content. The market has not yet priced Edge AI as a unified narrative because revenue recognition follows device shipments with a lag, but the simultaneous crossing of technology thresholds (NPU performance, quantization, memory bandwidth) makes now the inflection point. Investing requires a layer-by-layer approach rather than a single bellwether stock.
•Edge AI devices contain five components: SoC/NPU, memory, storage, sensors, and communication — each seeing rising BOM content as AI capabilities increase.
•Phone NPU performance has jumped from ~5 TOPS three years ago to above 50 TOPS today; memory capacity is moving from 8GB to 12-16GB, with 24GB high-end configurations emerging.
•A 7-billion-parameter LLM quantized to INT4 requires roughly 4GB of memory — now feasible in phones, enabling on-device inference.
•Automotive Edge AI requires 8-12 cameras, 4-6 radar units, and 8-12 ultrasonic sensors per vehicle, driving SoC and memory content.
•Cloud inference cost rises linearly with users, giving Apple, Google, Samsung, and Microsoft a strong economic incentive to shift inference to devices.
•Revenue recognition for Edge AI lags 2-4 quarters for consumer devices and 24-36 months for automotive SoCs, creating a layer-by-layer timing opportunity.
Article identifies Qualcomm Snapdragon as the brain in phones and mentions NPU TOPS rising from 5 to 50+; as phones become primary Edge AI devices, Qualcomm's SoC content per phone increases.
Article identifies Qualcomm Snapdragon as the brain in phones and mentions NPU TOPS rising from 5 to 50+; as phones become primary Edge AI devices, Qualcomm's SoC content per phone increases.
Risk: Competition from Apple Silicon and potential slowdown in smartphone replacement cycles.
Article names Mobileye EyeQ as the automotive AI computer alongside NVIDIA Drive, and notes that autonomous driving requires real-time on-device inference with safety certification — Mobileye is a dom
Article names Mobileye EyeQ as the automotive AI computer alongside NVIDIA Drive, and notes that autonomous driving requires real-time on-device inference with safety certification — Mobileye is a dominant player in this regulated layer.
Risk: Long design-win cycles (24-36 months) and potential market share loss to NVIDIA or Chinese ADAS suppliers.
Article highlights LPDDR5X as current mainstream and LPDDR6 as next-gen for bandwidth per watt, with memory capacity directly setting the size of on-device models (e.g., 4GB for 7B-parameter model). M
Article highlights LPDDR5X as current mainstream and LPDDR6 as next-gen for bandwidth per watt, with memory capacity directly setting the size of on-device models (e.g., 4GB for 7B-parameter model). Micron is a leading supplier of LPDDR memory for phones and automotive.
Risk: Memory pricing cyclicality and potential oversupply; transition to LPDDR6 may take time.
This newsletter, published May 29, 2026,
features Damnang
discussing QCOM, MBLY, MU.
3 trade ideas extracted by AI with direction and confidence scoring.