−11.8%
since Oct 3
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 device makers, edge silicon, memory and storage, and a small sleeve in the open-model ecosystem, since local inference needs more compute and more RAM everywhere at once. It is spread across many names rather than betting on one company, with a slice of cash kept aside to add on dips. I see this as a concentrated two-year view, not a permanent holding, and I'd rethink it if on-device models stall or the cloud keeps winning instead.
What it owns
- Companies selling devices where AI runs locally rather than in the cloud, capturing an upgrade cycle as inference moves from servers to consumer hardware. · AppleIt has the largest installed base of devices with dedicated neural processors and is pushing inference onto iPhones and Macs instead of relying on cloud compute., NvidiaIts RTX consumer graphics cards are the standard hardware for running open models on home computers, winning as local inference drives consumer GPU demand., Dell TechnologiesIt is one of the largest PC sellers to corporations and will capture upgrade volume and pricing when companies refresh fleets for on-device AI capabilities., HP Inc.It is among the largest PC sellers globally, so an AI-driven refresh cycle in laptops and desktops flows directly through unit volumes and pricing., Super Micro ComputerIt builds compact on-premises servers businesses deploy to run AI locally instead of using cloud capacity, winning as on-prem and edge AI adoption grows., Corsair GamingIt sells memory, power supplies, and cases for consumer-built AI computers, capturing the most direct exposure to home enthusiasts running models locally.33%
- Processors and chips designed to run AI tasks on the device itself rather than sending data to distant servers, winning as on-device inference becomes the default. · QualcommIts Snapdragon processors with dedicated neural processors are winning sockets in AI phones and Windows-on-Arm laptops, capturing the mobile and mobile-first compute shift., Advanced Micro DevicesIts Ryzen AI processors with built-in neural processing units are shipping in AI-capable laptops and desktops, plus consumer GPUs enable local model inference., Arm HoldingsMost phones and a growing share of laptops use its processor designs, so expansion in AI devices means more royalty revenue as Arm remains the standard for power-efficient inference., BroadcomIt supplies networking and connectivity silicon that edge devices depend on, plus custom chip design for companies building proprietary AI processors., Lattice SemiconductorIts low-power programmable chips enable AI inference in cameras, laptops, and industrial equipment where full processors are unnecessary, capturing edge AI spread., AmbarellaIts computer vision chips process AI inference inside cameras and vehicles directly, capturing the shift to on-device processing in security and driver-assist systems.27%
- Local AI models require more RAM per device and generate more data to store, so memory and storage suppliers benefit as capacity per device rises. · Micron TechnologyLocal AI models consume far more RAM than traditional workloads, so AI PCs with higher DRAM specifications support stronger memory demand and pricing., Western DigitalLocal AI model files and the data they operate on must reside on storage drives, so rising capacity per device drives demand for larger local storage., Seagate TechnologyIts drives store the growing volume of data that local AI systems create and consume at home and in small edge servers, benefiting as adoption rises., SandiskIts NAND flash storage goes into every AI PC and phone, so demand benefits as devices carry larger and faster solid-state drives for model files.19%
- Companies making capable AI models freely available and the foundries that manufacture chips for everyone else, winning regardless of which specific device wins. · Meta PlatformsIts Llama open models are the primary reason capable AI can run on consumer hardware, and releasing competitive models also supports Meta's own business interests., Taiwan SemiconductorIt manufactures nearly every chip in this take regardless of designer, so it wins on edge AI chip volume without exposure to which specific architecture dominates.11%
- Cash10%