+1.3%
since Sep 30
The idea
I think AI is moving out of the cloud and onto the devices we already own, laptops, phones, home GPUs, as open models shrink and get good enough to run locally for privacy, speed, and cost. So this basket spreads across the whole stack: the devices themselves at 33%, the edge chips that make local inference possible at 27%, the memory and storage that hungry local models need at 19%, and a smaller slice in whoever publishes the open models and fabricates the chips at 11%.
What it owns
- Companies selling the end devices and systems that run AI inference locally rather than in the cloud. · AppleIt has the largest installed base of devices with its own neural silicon and committed to running AI on the device, so inference adoption on iPhones and Macs drives its exposure., NvidiaRTX consumer graphics cards are the default hardware for running open models at home, so local inference demand supports consumer GPU sales., Dell TechnologiesCorporate laptop and workstation refresh cycles accelerate when businesses deploy AI locally, lifting both shipment volumes and pricing., HP Inc.As one of the largest PC sellers globally, an AI-driven refresh cycle flows directly through laptop volumes and margins as customers upgrade for on-device capabilities., Super Micro ComputerIt builds compact servers for businesses running AI inference on-premises rather than through cloud services, capturing the on-prem and edge server market., Corsair GamingMemory, power supplies, and cases sold for DIY PC builds targeting local model inference represent the home-enthusiast edge of on-device AI adoption.33%
- Chipmakers designing processors, neural engines, and connectivity silicon optimized for on-device AI. · QualcommIts Snapdragon chips with dedicated AI processors are winning sockets in both Windows on Arm laptops and premium phones as AI capabilities become standard., Advanced Micro DevicesRyzen AI processors with built-in NPUs compete in the AI PC market, and consumer GPUs from AMD support people running local models at home., Arm HoldingsNearly every phone and a growing share of laptops run Arm designs, so more AI devices mean higher royalty revenue as Arm-based chips become the standard for power-efficient on-device AI., BroadcomIt supplies networking and connectivity chips edge devices depend on, plus custom silicon for companies building their own AI chips., Lattice SemiconductorIts low-power programmable chips handle AI inference in cameras, laptops, and industrial equipment where dedicated processors would be inefficient., AmbarellaComputer vision AI in cameras and cars increasingly runs on-device using its chips, shifting inference from the cloud to the edge.27%
- Suppliers of DRAM, flash storage, and hard drives that store models and data on edge devices. · Micron TechnologyLocal AI models require substantially more DRAM than prior generations, and AI PCs are shipping with higher memory specifications as adoption scales., Western DigitalModel files and training data live on local drives, so increasing storage capacity per AI device drives demand for hard disk storage., Seagate TechnologyDrives store the growing pile of data that local AI creates and processes in home setups and small edge servers, supporting total storage demand., SandiskNAND flash in SSDs is standard across AI PCs and phones, so larger and faster storage requirements as AI adoption grows lift flash demand and pricing.19%
- Companies enabling the software and manufacturing foundation for open-source AI models to run anywhere. · Meta PlatformsLlama open models make capable AI feasible on consumer hardware, and the spread of open models accelerates adoption of on-device inference., Taiwan SemiconductorNearly every chip in the edge AI stack, from Apple to Qualcomm to AMD, is manufactured by TSMC, so edge AI volume scales directly to its production.11%
- Cash10%