−10.3%
since Sep 30
The idea
I believe AI is moving off the cloud and onto the devices we already own, as open models get good enough to run locally for privacy, speed, and cost. To play that, this basket spans four layers: the device makers building the hardware, the edge silicon enabling on-device inference, the memory and storage that hungry local models demand, and a smaller sleeve covering model publishers and chip fabrication. It is a concentrated two-year view rather than a forever hold, with 10% kept in cash to add on dips if the thesis keeps playing out.
What it owns
- Companies that sell consumer and corporate devices win when those devices are refreshed to run AI locally, provided the upgrade cycle actually materializes in unit volumes and pricing power. · AppleIt has the largest installed base of devices with proprietary neural silicon and has committed to running AI inference directly on iPhones and Macs rather than relying on cloud processing., NvidiaIts RTX consumer GPUs are the standard hardware for running open-source models at home, and sustained local inference demand is critical to consumer GPU sales., Dell TechnologiesIt sells corporate laptops and workstations that face replacement when enterprises decide their fleets need on-device AI capability, driving the refresh cycle in commercial PC segments., HP Inc.It is one of the largest PC makers globally, so an AI-driven upgrade cycle flows directly through its laptop unit volumes and margins in both consumer and commercial channels., Super Micro ComputerIt builds compact servers businesses deploy on-premises to run AI locally instead of renting cloud capacity, and wins as enterprises move inference from public cloud to edge infrastructure., Corsair GamingIt sells the memory, power supplies, and chassis for consumer-built PCs running models locally, representing the enthusiast segment of home AI infrastructure.33%
- Processors and connectivity chips designed for power-efficient on-device AI inference gain share as devices move compute away from cloud data centers toward the edge. · QualcommIts Snapdragon processors include dedicated AI accelerators going into phones and Windows laptops, winning sockets in the wave of AI-capable mobile and portable devices., Advanced Micro DevicesIt ships Ryzen AI processors with integrated NPUs into the same AI PC refresh cycle, plus consumer GPUs for local model inference, and gains as it takes design share in AI-capable systems., BroadcomIt supplies the networking and connectivity silicon edge devices depend on, plus custom chip design for companies building their own AI accelerators, and benefits from edge AI growth., Arm HoldingsIts instruction set is the basis for nearly every phone and a growing share of laptops, so more AI devices means higher royalty revenue from Arm-based chip designs., Lattice SemiconductorIt makes low-power programmable chips that handle AI tasks in cameras, laptops, and industrial equipment where a full processor is unnecessary, benefiting as edge AI spreads into constrained devices., AmbarellaIts chips run computer vision AI inside cameras and vehicles where inference must happen on the device, and it gains as security and automotive systems move to on-device processing.27%
- Local AI models demand more memory and storage per device than prior generations, and this basket benefits as those component specifications grow with AI adoption. · Micron TechnologyIt manufactures the DRAM that local AI models require, and AI PCs are shipping with materially higher memory specifications as inference moves to the device., Western DigitalIt manufactures hard drives that store model files and data for local AI workloads, and benefits as AI devices and home setups push storage capacity per device higher., SandiskIt produces the NAND flash inside SSDs that ship in every AI PC and phone, and gains as AI devices carry larger and faster storage to support model inference., Seagate TechnologyIt manufactures drives that store the growing volume of data local AI systems create and consume at home and in edge servers, and benefits as total storage demand rises with adoption.19%
- Open-source AI models and the manufacturers that enable them thrive when local inference becomes viable, creating a foundation for device-side AI that does not depend on cloud services. · Meta PlatformsIts Llama models are the primary reason capable AI can run on hardware people own, and open-model adoption accelerates the local-inference premise underlying this take., Taiwan SemiconductorIt manufactures nearly every chip in this basket, from Apple's neural engines to Qualcomm's and AMD's AI processors, benefiting from edge AI volume regardless of which designer wins.11%
- Cash10%