+0.3%
since Oct 2
Practiced to +1.7% over 3 days before going real.
Practice chapter: simulated results; no money was invested.
The idea
I believe AI is moving out of the cloud and onto the devices we already own, as open models shrink enough to run well on consumer chips. This basket spreads across the layers that shift would favor: device makers, edge silicon, memory and storage for hungry local models, and a small slice of the open-model ecosystem itself. I'm keeping about a tenth in cash to add on dips, and I treat this as a concentrated two-year view rather than something I'd hold forever. If local inference doesn't catch on the way I expect, that's my cue to rethink it.
What it owns
- Companies that sell the finished devices running AI locally, from phones and laptops to workstations and custom servers, benefit as businesses and consumers upgrade fleets to handle on-device inference. · AppleIt has the largest installed base of devices with proprietary neural processors and has committed to running inference locally on iPhones and Macs rather than in the cloud., NvidiaRTX consumer GPUs are the default hardware for people running open models at home, making local inference demand the critical driver for this company's consumer business., Dell TechnologiesIt sells the corporate laptops and workstations that face replacement when enterprises decide their fleets need local AI capabilities, creating an upgrade cycle tied directly to on-device inference adoption., HP Inc.It is one of the largest PC makers on earth, so an AI-driven refresh cycle flows directly into laptop volumes and average prices across its product line., Super Micro ComputerIt builds the compact servers that businesses deploy on-premises to run AI locally rather than renting cloud capacity, capturing growth in edge and on-premise AI infrastructure., Corsair GamingIt sells the memory, power supplies, and cases that enthusiasts buy when building custom PCs to run AI models at home, making it the purest consumer DIY angle in the basket.33%
- Processors and chips designed to run AI inference at the edge of the network, rather than sending data to the cloud, become essential as devices prioritize local computation and power efficiency. · QualcommSnapdragon processors with dedicated NPU hardware are winning sockets in AI-capable phones and Windows laptops, putting this company at the center of the mobile and Windows-on-Arm shift., Advanced Micro DevicesRyzen AI processors with built-in NPUs compete directly in the AI laptop market, while this company's consumer GPUs also power home-based AI inference., BroadcomIt supplies the networking and connectivity chips that edge devices depend on, plus custom silicon design services for companies building their own AI processors., Arm HoldingsMost phones and a growing share of laptops use this company's processor designs, so higher volumes of AI devices translate directly into higher royalty revenue., Lattice SemiconductorIt makes tiny, low-power programmable chips that handle AI tasks in cameras, laptops, and industrial equipment where a full processor is unnecessary, capturing a niche in power-constrained edge AI., AmbarellaIt provides the vision processors that run AI inference inside cameras and vehicles, meaning on-device AI processing in these segments is this company's core business.27%
- Local AI models require more memory and storage than previous device generations, creating sustained demand for DRAM, NAND flash, and hard drives as AI adoption spreads. · Micron TechnologyDRAM content in AI-capable devices is materially higher than the prior generation, so memory supply and pricing hold the key to this company's growth as AI PC volumes scale., Western DigitalModel files and their associated data live on local storage, and this company manufactures the hard drives that store growing volumes of AI workloads in homes and small edge deployments., Seagate TechnologySeagate's drives store the growing volume of data that local AI creates and processes at the edge, so rising storage demand from AI adoption flows directly to this company's results., SandiskEvery AI-capable phone and laptop ships with NAND flash storage that this company provides, so AI devices carrying larger and faster storage drives higher demand and pricing.19%
- Open-source AI models and the infrastructure supporting them enable capable inference on consumer hardware, while manufacturers of the chips powering that hardware win regardless of which designer wins sockets. · Meta PlatformsLlama open models are the primary reason capable AI can run on consumer hardware, and this company's continued release of competitive open models sustains the entire local-AI premise., Taiwan SemiconductorNearly every chip in this basket is manufactured by this company, so it captures upside from edge AI volume growth regardless of which chip designer wins market share.11%
- Cash10%