Home AI Edge Compute
Nikita
@nikita
· 17
+11.1%
since Aug 29
The idea
I believe AI inference is moving out of the cloud and into hardware people own, as open-source models keep getting smaller and better. I'm expressing this through device makers shipping AI PCs and consumer GPUs, backed by the memory, storage, and edge silicon that gate local models, plus a small sleeve in the open-model ecosystem. No single position dominates, I keep a 10% cash buffer to add on dips, and I'm running this as a concentrated, two-year view rather than a forever holding.
What it owns
- These companies manufacture the hardware platforms where local AI models run, from consumer laptops and phones to home servers and specialized workstations. · AppleIt runs AI models directly on its devices rather than in the cloud, making it the largest proof point that on-device inference drives hardware upgrades., NvidiaMost people running local models at home use Nvidia cards, so consumer GPU sales and gaming revenue growth indicate whether the shift to local inference is real., Dell TechnologiesIt ranks among the top two PC makers selling AI-capable devices, so its consumer and commercial PC metrics show whether the AI PC refresh is actually materializing., HP Inc.It is the other major AI PC seller, offering a higher-volume, lower-cost entry point to capture the hardware upgrade cycle., Super Micro ComputerIt manufactures compact AI servers and workstations for enthusiasts and small businesses running models on owned hardware rather than cloud services., Corsair GamingIt supplies cases, power supplies, and memory for DIY AI machine builders, capturing a small speculative bet on growth in the home builder community.33%
- Processors and chip designs that enable on-device inference, from smartphone neural processing units to low-power edge accelerators, power the shift away from cloud-dependent AI. · QualcommIts chips power on-device AI features in Android phones and AI-focused laptops, making it the most direct edge silicon exposure outside of GPUs., Advanced Micro DevicesIt designs chips combining CPU, GPU, and AI processing on a single die, the exact architecture that home AI systems need for efficient local inference., BroadcomIt supplies the WiFi and networking chips connecting home AI devices together, though most of its growth comes from cloud AI rather than edge AI infrastructure., Arm HoldingsNearly every smartphone and a growing share of laptops license its chip designs, so it collects royalties on most AI devices shipped regardless of brand winner., Lattice SemiconductorIt makes tiny, power-efficient programmable processors that run AI tasks in size and power-constrained devices like cameras and appliances., AmbarellaIts chips let cameras and vision systems run AI inference locally without sending video to the cloud, betting that vision is the first workload to leave cloud infrastructure.27%
- Local models require substantial memory and storage capacity, so suppliers of these components benefit regardless of which device platform dominates. · Micron TechnologyIt manufactures the memory that local models depend on, so rising per-device RAM requirements increase both pricing and volume., Western DigitalIt makes the storage drives that hold tens of gigabytes of model files for local inference, capturing demand from consumer and edge storage growth., SandiskIt manufactures the flash memory in laptops, phones, and compact AI devices, the storage layer closest to where local inference actually executes., Seagate TechnologyIt sells affordable high-capacity drives for storing model weights and inference data at home, capturing the value end of local storage demand.19%
- Open-source model releases drive adoption of local AI, while the foundry fabricating chips across this portfolio wins from any device category that gains share. · Meta PlatformsIts open-source Llama models make local AI practical and accessible, serving as the engine enabling the entire shift toward on-device inference., Taiwan SemiconductorIt fabricates nearly every processor in this portfolio across all device makers, so it benefits from whichever category wins while also concentrating geopolitical risk.11%
- Cash10%