Full-Stack AI Awakening
kamil@unable
−6.5%
since Oct 6
The idea
I believe we're still early in the AI buildout and that current prices don't fully reflect the next five to ten years of adoption, so instead of picking one winner I own the whole stack. Chips and compute are the largest piece at a third of the portfolio, with cloud giants, power and grid, data center gear, and AI-driven software filling out the rest, plus a small cash buffer. This is built to be held for years, not traded on headlines, and I'd rethink it if the buildout itself stalls rather than just on a rough quarter.
What it owns
- The processors and machines that enable AI training and inference are still the scarcest component of the entire ecosystem. · NVIDIAIts chips are the default hardware for training the next generation of large AI models, provided training demand continues growing and it remains the preferred choice for major AI developers., Taiwan SemiconductorIt manufactures nearly every advanced AI chip including those from competitors, and a multiyear expansion cycle ensures sustained orders flowing through its facilities., BroadcomIt manufactures custom AI chips for cloud providers who design their own silicon to reduce costs while still purchasing NVIDIA processors., ASML HoldingIt is the sole manufacturer of equipment required to produce the most advanced chips, ensuring years of orders flow through its business as AI chip demand expands., Advanced Micro DevicesAs AI compute demand expands over multiple years, buyers seek alternatives to a single-supplier market, and it is the established second option with competitive AI chips., Micron TechnologyAI chips require enormous quantities of fast memory located adjacent to processors, and it is one of three manufacturers of this critical component.33%
- The largest cloud platforms rent AI compute to others and provide stable, long-term returns across a multi-decade cycle. · MicrosoftAzure rents AI compute and Copilot integrates AI into office software, positioning it to capture adoption across both infrastructure and application layers., AlphabetIt builds proprietary frontier models, designs its own chips, and operates a cloud platform, gaining leverage at multiple levels of the AI stack simultaneously., AmazonAs the largest cloud platform, it captures a substantial portion of AI workloads migrating to cloud infrastructure over the coming decade., OracleIt is constructing data centers dedicated to renting raw AI compute capacity, backed by substantial customer contracts from AI companies.22%
- Data centers consume enormous amounts of electricity, and the companies supplying power generation and grid infrastructure are overlooked by mainstream coverage. · EatonNew data centers require electrical distribution infrastructure to route power throughout facilities, creating sustained demand for its products., GE VernovaAI data centers demand far more electricity than existing grids provide, and it manufactures the turbines and transformers needed to expand power infrastructure., Constellation EnergyIt operates the largest nuclear fleet in the United States, providing the always-on carbon-free power that technology companies seek for data centers., VistraIt generates electricity in regions where data centers are concentrating, and long-term supply contracts with customers lock in revenue as demand for power remains tight.18%
- Networking equipment, cooling systems, and physical infrastructure get purchased regardless of which AI models emerge as dominant. · VertivAI chips generate significant heat and power consumption, and it supplies the cooling and electrical equipment that keeps data centers operational., Arista NetworksHigh-speed networking between thousands of AI chips is essential for cluster performance, and it provides the switches cloud providers install in AI infrastructure., Digital RealtyIt owns the physical facilities where AI infrastructure is installed and collects escalating rents as data center space becomes increasingly scarce.13%
- Software that embeds AI into routine business processes is where widespread adoption first becomes visible. · PalantirIt provides the software platform that helps large organizations and governments implement AI solutions, earning revenue as it expands these deployments., ServiceNowIt sells workflow automation software and is integrating AI agents into its platform, generating incremental revenue if customers adopt these AI features.6%
- Cash8%