−17.1%
since Sep 10
The idea
I think AGI progress will be limited less by model breakthroughs than by physical constraints: chips, power, data centers, and storage. So rather than betting on the AI labs themselves, this basket owns the picks and shovels across four layers of the buildout, semiconductors, converted data center infrastructure, power generation, and memory storage, with a small slice of broad market exposure to soften single-company risk.
What it owns
- The take contains companies that build or operate the physical infrastructure, networking, and systems that enable AI data centers to function. · Core ScientificIt converted bitcoin mining facilities into AI data centers with long-term hosting contracts, positioning it to capture value from the shift in how those power-connected buildings are used., CoreWeaveIt rents Nvidia processors to AI companies on an hourly basis and represents the most direct ownership of underlying data center capacity demand and utilization., Applied DigitalIt constructs data center capacity in locations with low-cost electricity and leases the infrastructure to AI companies, with customer contract execution driving its growth., IREN LtdIt owns renewable energy powered sites originally built for cryptocurrency mining and is converting them to serve AI compute customers with signed contracts., CleanSparkIt holds large amounts of contracted electricity capacity and grid connections in a region where data centers are being built, positioning it as a conversion candidate for AI infrastructure., Riot PlatformsIt controls substantial electricity contracts in Texas, making it a potential acquisition or conversion target for companies seeking to deploy AI data center capacity in low-cost power regions.31%
- The take focuses on companies that design or manufacture the processors and chips powering artificial intelligence systems and data centers. · NvidiaIt designs the chips that form the core of nearly every AI data center buildout, so sustained acceleration in physical infrastructure translates directly into orders., BroadcomIt builds custom AI processors for the largest cloud companies and provides the networking equipment that connects thousands of chips together inside data centers., IntelIt operates the largest foundry capacity in the US, making it essential if domestic artificial intelligence chip manufacturing becomes a strategic priority, but success depends on winning external customers beyond internal demand., TSMC (ADR)It manufactures nearly all advanced AI chips globally, including for rivals, so it captures value across multiple chip designers and benefits from growth in data center silicon demand.29%
- The take includes energy providers and power management companies that supply the electricity needed to run data centers and AI infrastructure. · Constellation EnergyIt operates the largest nuclear fleet in the United States and is signing multiyear power supply agreements with technology companies building AI data centers., Vistra CorpIt generates electricity in the regions where data centers are expanding fastest, and rising demand for computing power increases the rates it can charge for generation., Bloom EnergyIt manufactures fuel cells that allow data centers to generate their own power on site rather than waiting for grid connections, eliminating a major constraint to deployment speed.21%
- The take includes businesses that provide the storage and memory infrastructure required to support large-scale AI and data center operations. · SanDisk CorpIt supplies the flash memory that stores the enormous datasets required for AI training, so sustained data center growth keeps demand for storage capacity firm.9%
- The take includes positions that provide portfolio diversification and downside protection. · GE VernovaIt produces gas turbines and grid infrastructure required to expand electricity supply, winning on the infrastructure buildout regardless of which specific generator or data center operator advances., Vanguard S&P 500 ETFIt tracks the broad market and provides diversification to offset volatility from concentrated exposure to the data center infrastructure theme.5%
- The take holds additional positions that support the broader artificial intelligence infrastructure ecosystem. · MRVLIt produces the networking and custom processors that enable data movement within AI data centers, with growth tied to cloud customer adoption of proprietary silicon designs.5%