−10.4%
since Sep 30
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 local chips. This basket spreads across device makers, the edge silicon that powers on-device inference, the memory and storage that hungry local models demand, and a small stake in the open-model ecosystem itself. No single name dominates, and I'm holding 10% in cash to add on dips. I see this as a concentrated two-year view rather than a permanent holding, and I'd rethink it if local inference fails to catch on.
What it owns
- Companies selling end devices, phones, laptops, desktops, servers, that shift from cloud-dependent to running AI locally. · AppleIt has the largest installed base of devices with dedicated neural silicon and is pushing inference onto iPhones and Macs instead of relying on cloud servers., NvidiaIts RTX consumer GPUs are the default hardware for running open models at home, and demand depends on local inference staying central to the category., Dell TechnologiesIt sells the corporate laptops and workstations that companies replace when deciding their fleets need to run AI locally, benefiting from the refresh cycle., HP Inc.As one of the world's largest PC sellers, it captures flows from an AI-driven refresh cycle through laptop volumes and pricing power., Super Micro ComputerIt builds compact servers that businesses deploy on-premises to run AI locally instead of renting cloud capacity., Corsair GamingIt sells memory, power supplies, and cases for people building PCs to run models at home, capturing the DIY enthusiast angle of local AI.33%
- Chip designers and makers of processors, NPUs, and custom silicon optimized for inference on devices rather than in data centers. · QualcommIts Snapdragon chips embed dedicated AI processors in phones and Windows laptops, and NPU-equipped designs need to keep winning new sockets., Advanced Micro DevicesIt ships Ryzen AI processors with built-in NPUs into AI PCs, and consumer GPUs for local models, relying on share gains in AI-capable laptops and desktops., BroadcomIt supplies networking and connectivity chips edge devices depend on, plus custom silicon design for companies building proprietary AI processors., Arm HoldingsMost phones and a growing share of laptops run on its chip designs, so more AI devices on its architecture means more royalties., Lattice SemiconductorIt makes tiny, low-power programmable chips that run AI tasks in cameras, laptops, and industrial hardware where large processors are unnecessary., AmbarellaIts chips run computer vision AI inside cameras and cars, with inference happening on the device by design rather than in the cloud.27%
- Suppliers of DRAM, NAND flash, and disk storage that see higher capacity per device as AI workloads move on-premises. · Micron TechnologyLocal AI models demand more RAM per device, and AI PCs are shipping with materially higher DRAM specifications than prior generations., Western DigitalModel files and the data they process live on local drives, and it supplies the hard drives storing growing model and data capacity per device., SandiskIt sells the NAND flash going into SSDs in AI PCs and phones, benefiting as AI devices ship with bigger and faster storage., Seagate TechnologyIts drives store the growing volume of data that local AI creates and consumes at home and in small edge servers.19%
- Companies releasing and supporting open-source AI models and the manufacturing infrastructure that makes local inference possible. · Meta PlatformsIts Llama open models are the primary reason capable AI can run on consumer hardware, and spreading open models also serves its own business model., Taiwan SemiconductorNearly every chip in this take is manufactured by it, so it wins on edge AI volume regardless of which chip designer gains share.11%
- Cash10%