+7.9%
since Sep 1
The idea
I think the AI buildout is fundamentally a construction problem, and the real pricing power sits with whoever controls the parts that take years to expand, not the chips everyone talks about. So this basket leans hardest into power and grid capacity, then spreads across fab tooling, networking, memory, and cooling infrastructure, with no single holding dominating any sleeve. I keep a slice in cash on purpose, since this space sells off hard when sentiment flips, and those selloffs are when I want room to add.
What it owns
- AI datacenters consume enormous amounts of electricity, and the generation, transmission, and distribution infrastructure cannot expand as quickly as demand. · GE VernovaIt manufactures the gas turbines and grid equipment that new datacenters need for power, with order backlogs extending years into the future., VistraIt owns natural gas and nuclear generation capacity in regions where datacenter power demand is most constrained, and new generation cannot be built quickly., EatonIt supplies the switchgear, transformers, and electrical distribution equipment inside every new datacenter, with lead times that stretch to years., Constellation EnergyIt operates the largest fleet of US nuclear reactors, providing the clean, always-on power that hyperscalers are signing long-term contracts to secure., Quanta ServicesIt builds the transmission lines and grid connections that deliver power from generation plants to datacenters, with permitting and crew availability constraining speed.27%
- Advanced chip manufacturing requires specialized equipment and facilities that take years to build and represent the deepest bottleneck in semiconductor production. · Taiwan SemiconductorNearly all advanced AI chips are manufactured by a single foundry, and building competing capacity at the leading edge requires half a decade and tens of billions of dollars., ASML HoldingIt is the only manufacturer of the extreme ultraviolet lithography machines required to print advanced semiconductors, making it the purest single-point constraint in the entire supply chain., Applied MaterialsIt supplies the deposition and etch tools that every new semiconductor fab requires regardless of which company operates it or what chips it makes., Lam ResearchIt dominates the equipment that etches semiconductor features as memory and processors stack in three dimensions, which is the direction AI chips are heading.24%
- Training large AI models requires thousands of processors connected by high-speed networks, creating demand for specialized networking hardware and optical components. · BroadcomIt supplies both the networking silicon inside AI datacenters and designs custom accelerator chips for major hyperscalers, occupying two critical roles simultaneously., Arista NetworksTraining large AI clusters requires thousands of GPUs wired together with specialized networking switches, and it holds the leading position at major cloud companies., CoherentData traveling between thousands of AI processors requires optical transceivers and lasers, and it is a leading supplier of those components.17%
- AI systems depend on the fastest possible data flow between processors and memory, creating a supply constraint at the chip level. · Micron TechnologyHigh-bandwidth memory is the critical link between AI processors and data, and only a handful of companies can manufacture it at the required speed and scale., NVIDIAIts processors and software stack define how most AI systems are built, making it the primary demand driver for all the other components in this take.13%
- Datacenters running AI workloads generate extreme heat that requires specialized cooling and construction expertise to handle at scale. · Vertiv HoldingsAI racks generate heat beyond what air cooling can handle, and it supplies the liquid cooling and power distribution systems that datacenters require., EMCOR GroupIt is one of the largest electrical and mechanical contractors building and outfitting datacenters, with skilled labor availability becoming a bottleneck.11%
- Cash8%