−10.1%
since Oct 2
The idea
I believe AI is moving out of the cloud and onto the devices we already own, so local processing power becomes the thing worth owning. This basket spans device makers building the hardware, the edge chipmakers enabling on-device inference, the memory and storage suppliers that local models lean on, and a small slice of the open-model ecosystem including the chip fabricator behind most of it. I'm keeping it diversified across names rather than betting on one, holding roughly a tenth in cash to add on weakness, and treating this as a multi-year view rather than a quick trade.
What it owns
- Companies that sell computers and servers win when their customers replace fleets with AI-capable hardware and when consumers build or upgrade machines to run models at home. · AppleIt has the largest installed base of devices with dedicated neural processors and has committed to running inference on phones and computers rather than in the cloud., NvidiaIts consumer RTX graphics cards are the standard hardware for anyone running open AI models on home computers, and demand depends on local inference staying central., Dell TechnologiesIt sells the corporate laptops and workstations that companies replace when their fleets need to run AI locally, so the take depends on an upgrade cycle actually materializing., HP Inc.It is one of the world's largest PC sellers, so an AI-driven upgrade cycle directly lifts its laptop volumes and pricing if the refreshes actually happen., Super Micro ComputerIt builds compact servers that businesses deploy on-premises to run AI locally instead of renting cloud capacity, and the take needs on-prem and edge AI server demand to keep growing., Corsair GamingIt sells the memory, power supplies, and cases that people buy when building a computer to run models at home, making it the purest home-enthusiast angle and dependent on DIY AI builds becoming a real category.33%
- Chips designed to run inference on devices rather than in data centers win as AI shifts toward local processing across phones, laptops, cameras, and specialized hardware. · QualcommIts Snapdragon chips put dedicated AI processors into phones and Windows laptops, and the take needs these NPU-equipped designs to keep winning in AI-capable devices., Advanced Micro DevicesIt ships Ryzen AI processors with built-in neural processing units into laptops and sells consumer graphics cards for local model inference, so it needs to keep gaining share in both., BroadcomIt supplies the networking chips that edge devices depend on and builds custom silicon for companies making their own AI processors, so growth depends on device volumes and custom chip demand., Arm HoldingsNearly every phone and a growing share of laptops use its designs, so more AI devices mean more royalty revenue as long as Arm-based chips stay the default for power-efficient inference., Lattice SemiconductorIt makes small, low-power programmable chips that handle AI tasks in cameras and laptops where a large processor wastes power, so the take needs edge AI to spread into power-constrained devices., AmbarellaIts chips run computer vision inference inside cameras and vehicles where processing happens on the device by necessity, so growth depends on security and driver-assist systems moving to on-device AI.27%
- Running AI models locally requires more memory and storage per device than traditional computing, and the supply side benefits as this hardware content grows. · Micron TechnologyLocal AI models require more DRAM than previous generations, and the company wins if that memory content stays high as AI devices scale production., Western DigitalModel files and their working data live on local drives, and the company wins if AI devices keep pushing storage capacity per machine higher., SandiskIt supplies the NAND flash inside the solid-state drives that every AI computer ships with, and the take needs flash pricing and demand to benefit as devices carry bigger, faster storage., Seagate TechnologyIts drives store the growing volume of data that local AI creates and uses, and the take holds if total storage demand keeps rising with edge AI adoption.19%
- Open-source AI models make local inference practical and valuable, and companies enabling that ecosystem benefit from the broader shift to edge AI regardless of which specific chip wins. · Meta PlatformsIts Llama open-source models are the main reason capable AI can run on hardware people own, and the take depends on Meta continuing to release competitive open models., Taiwan SemiconductorNearly every chip in this take is manufactured by the company regardless of which designer wins, so it benefits from edge AI volume growth while concentrating risk in one geography.11%
- Cash10%