AI Comes Back Home
jason reynolds@joeylicorice
−2.2%
since Oct 16
The idea
I think the cloud-data-center AI boom cracks within a couple of years, with work shifting to companies running open-source models on their own hardware. So this skips the cloud giants and chipmakers tied to their spending, and instead owns the servers, storage, chips, networking and software businesses need to run AI in-house. It holds about a fifth in cash, expecting these names to dip too in a crash, so there's room to add on weakness. No single stock dominates, and the view only changes if AI workloads stay centralized in the cloud rather than moving on premise.
What it owns
- Open-source models running inside company buildings need software to manage, search, protect and deploy data on machines the business owns. · NutanixIt provides the software that turns a company's own servers into a private cloud capable of running AI models without relying on external cloud giants., IBMIt owns Red Hat, whose open-source software lets companies run AI models on their own hardware, and IBM builds its own open models for on-premises use., MongoDBIt offers a database platform that companies can run on their own servers, widely used to build and support AI applications on private infrastructure., CommvaultIt backs up and protects company data wherever it sits, including in private data centers where an increase in in-house AI would create more data to safeguard., ElasticIt makes search software that companies often run themselves, widely used to feed private documents and data to in-house AI models., TeradataIt operates data warehouses that many large companies keep on their own premises and is positioning itself to run AI directly where that data lives.26%
- Businesses choosing to run AI themselves need to buy their own servers and storage hardware instead of renting compute and capacity from cloud giants. · Dell TechnologiesIt sells the AI servers and storage systems that regular companies deploy in their own data centers instead of renting from cloud providers., Hewlett Packard EnterpriseIt sells servers, storage and private cloud platforms that let companies run AI on hardware they own instead of renting compute from a cloud giant., NetAppIt provides storage systems that hold a company's data in its own data center, keeping information close to local AI models that need fast access.18%
- These chip makers sell into company servers and laptops, enabling AI to run on machines businesses own rather than in distant cloud data centers. · Advanced Micro DevicesIt makes server and AI chips that companies install in their own machines, and powers AI chips in many new laptops for local model inference., QualcommIt makes chips with built-in AI engines for laptops and phones, allowing AI models to run on the device itself rather than in a remote data center., IntelIt makes the chips that power most servers inside regular companies and builds AI-capable laptop processors for running models locally.17%
- AI systems running on a company's own hardware need fast, reliable networks to move data and models securely inside the building. · Cisco SystemsIt builds the networking equipment inside company offices and data centers that in-house AI systems need to move data between machines reliably., F5It makes software and hardware that secure and manage company applications across in-house data centers, protecting AI systems running on owned infrastructure.11%
- Powerful personal computers and laptops can run open AI models locally, letting employees work without sending data to cloud services. · AppleIt designs Macs and software that run open-source AI models directly on the machine, and builds its own AI to stay on the device rather than in the cloud., HP Inc.It sells business laptops and workstations powerful enough to run AI models locally, letting employees work with AI tools on machines they own.9%
- Cash20%