Situational Awareness Conservative
Michael Homerick@homerick
−10.4%
since Jul 28
The idea
I believe AGI's real bottleneck is physical: power, chips, data centers, and storage, not the model companies themselves. So this basket leans into the buildout, with large-cap semiconductor names for stability, utilities and data center operators for the infrastructure layer, and a memory/storage position to round it out. I keep a meaningful cash reserve to buy into weakness, avoid concentrating too heavily in any single name, and treat major shifts in the underlying 13F holdings this idea is based on as a signal to revisit the thesis.
What it owns
- Physical facilities and GPU cloud platforms that provide the computational capacity for AI development and deployment at scale. · Core ScientificIt operates data center facilities with existing power infrastructure, repurposing industrial capacity for artificial intelligence and high-performance computing workloads., CoreWeaveIt leases graphics processing unit capacity to artificial intelligence laboratories and researchers who prefer renting compute rather than building their own facilities., Applied DigitalIt builds and operates data centers equipped with graphics processors to serve customers developing artificial intelligence applications and services., IREN LtdIt develops data center facilities powered by renewable energy sources, serving artificial intelligence and high-performance computing customers., CleanSparkIt operates data centers with substantial power infrastructure capacity, repurposing industrial-scale facilities for artificial intelligence and computing workloads., Riot PlatformsIt owns and operates extensive power infrastructure in Texas that can be redirected toward artificial intelligence and high-performance computing applications.26%
- The foundational layer of artificial intelligence infrastructure, where chip designers and manufacturers enable all downstream AI applications. · NvidiaIt designs the graphics processors that dominate artificial intelligence applications, with pricing power reflected in very high gross margins., BroadcomIt manufactures custom artificial intelligence accelerators for major cloud companies, diversifying away from reliance on a single chip supplier., IntelIt operates one of the few advanced semiconductor factories in the United States, providing a domestic alternative for critical chip manufacturing., TSMC (ADR)It manufactures chips designed by others, including processors for artificial intelligence applications, winning by serving every major design ecosystem.24%
- Electricity generation and delivery systems that supply the enormous and growing power demands of data centers running AI workloads. · Constellation EnergyIt generates carbon-free electricity and has secured long-term contracts to power data centers, directly benefiting from artificial intelligence infrastructure growth., Vistra CorpIt produces electricity from nuclear and gas generation and holds long-term power agreements with major artificial intelligence companies., Bloom EnergyIt manufactures fuel cells that generate power on-site at data centers, providing electricity independent of the grid when it faces congestion.17%
- Data storage systems that handle the vast volumes of information required to train and operate large language models and AI workloads. · SanDisk CorpIt produces NAND flash memory essential for storing the massive datasets used to train and operate artificial intelligence models.8%
- Broad market exposure and infrastructure companies that provide protection if AI adoption progresses more slowly than expected. · GE VernovaIt builds gas turbines and electrical grid infrastructure that form the physical backbone supporting electricity generation for data centers and other facilities., Vanguard S&P 500 ETFIt provides broad exposure to major companies developing artificial intelligence applications and services, while protecting against slower-than-expected adoption.4%
- Cash20%