−10.2%
since Sep 30
The idea
I believe AI inference is moving out of the cloud and into boxes that schools, labs, and homes actually own, so I built around the portable but powerful machines making that possible, plus the chips, memory, and foundries that gate how much model fits on the edge. I added a smart home sleeve on the idea that AI accelerates that hardware category too, played through component makers rather than big platforms, backed by distributors that get parts into classrooms worldwide. No single name dominates, I keep some cash to add on dips, and I'm thinking in years, not months.
What it owns
- Portable AI machines powerful enough to run substantial models locally are the hardware that schools and home labs actually purchase for real workloads. · NvidiaIt builds the DGX Spark, a portable machine explicitly designed for schools and labs to run substantial AI models without cloud dependence., Advanced Micro DevicesIts Strix Halo chips deliver enough performance for real local models while staying portable enough for classroom deployment., Dell TechnologiesIt is the institution's path to purchasing and supporting AI workstations at scale, handling procurement and deployment across school districts and enterprises., HP Inc.It ships AI-capable PCs at the volume that schools and institutions actually buy, making this the distribution channel for local AI refreshes.29%
- Inference on devices rather than in the cloud requires silicon designed for that task, and a foundry capable of manufacturing it at scale. · QualcommIt designs the processors that run inference directly on laptops, smart home devices and edge hardware instead of sending data to the cloud., Taiwan SemiconductorIt manufactures the silicon for the chipmakers across this take, so it wins if on-device AI adoption accelerates across all segments., BroadcomIt provides the connectivity chips that link local AI machines, smart home devices and edge systems to each other and to networks., Arm HoldingsIts processor designs sit inside most edge and smart home silicon, collecting royalties on the shift of computing from cloud to device., Lattice SemiconductorIt supplies programmable silicon that small hardware makers use to prototype and build custom edge AI systems faster than designing chips from scratch.19%
- On-device intelligence makes smart home devices genuinely more useful, driving upgrades across cameras, speakers, controls and security gear in residential settings. · Silicon LabsIt makes the wireless radio chips that connect smart home devices to each other and the network, enabling more complex local interactions., AmbarellaIts processors enable cameras to run AI detection and recognition on the device itself without streaming video to external servers., SynapticsIt designs processors that sit inside smart home devices and edge hardware, the silicon layer where on-device AI actually gets implemented., Resideo TechnologiesIt already operates thermostats and controls in millions of homes, positioning itself to upgrade that installed base with on-device AI features., Arlo TechnologiesIt sells AI-enabled security cameras directly to households, embedding inference capabilities into the device rather than relying on cloud processing., SonosIts smart speakers become more useful if on-device AI capabilities improve voice recognition and task handling without constant cloud calls.18%
- Local models run only as large as the device's memory and storage allow, so machines built for on-device AI pull these components into every unit sold. · Micron TechnologyIt supplies the RAM that limits how large a model can run locally, making memory capacity the gating factor for local AI machines., Western DigitalIt provides the storage that local AI machines use to keep model files and data on the device instead of in cloud data centers., Seagate TechnologyIt covers bulk storage for school labs and home setups that keep models, datasets and video on local drives rather than in the cloud., SandiskIts flash memory goes into the portable machines and smart home devices that need fast local storage for models and on-device processing.16%
- Component distributors and specialty retailers supply the parts and finished systems that get local AI hardware into classrooms and home builds worldwide. · Arrow ElectronicsIt distributes semiconductors and components to builders and institutions across the globe, moving parts from manufacturers to the buyers who assemble or deploy them., AvnetIt supplies components to small builders, boutique hardware companies and school labs in every geography, moving parts to the decentralized buyers in this ecosystem., Corsair GamingIt sells the enclosures, power supplies and components that hobbyists and home lab builders use to assemble their own local AI hardware.8%
- Cash10%