−11.8%
since Oct 8
The idea
I believe AI compute demand keeps compounding faster than the market expects over the next couple of years, and that no single company captures it all, so this basket spreads across the whole supply chain while leaning hardest on choke points: power generation and electrical gear data centers need to run, and the wafer equipment and foundries every chip has to pass through. Memory and networking round it out, with hyperscalers included as proof the demand is real.
What it owns
- Advanced chip manufacturing is the narrowest choke point in the chain, so the take needs leading-edge fabs to stay capacity-constrained and the equipment makers supplying them to keep shipping tools at full speed. · Taiwan SemiconductorIt manufactures nearly every advanced AI chip from every maker, so the take needs its advanced-node capacity to stay sold out and AI-related revenue to keep climbing., ASMLIt is the only maker of EUV lithography machines needed to print leading-edge chips, so the take needs foundries to keep expanding advanced capacity and its order book to reflect that expansion., Lam ResearchIt sells the etch and deposition equipment fabs need to build logic chips and the high-bandwidth memory AI accelerators depend on, so the take needs continued fab spending to drive its orders., Applied MaterialsIt is the broadest supplier of chipmaking equipment across tool types, so the take needs overall equipment spending by foundries and memory makers to keep rising., KLAIt dominates inspection and measurement tools that catch defects and become more essential as chips grow more complex and expensive to ruin, so the take needs advanced-node production to keep growing.21%
- The take rests on accelerator demand growing faster than supply, so it needs the companies selling training and inference chips to keep landing orders and staying the default choices for building out AI infrastructure. · NVIDIAIt sells the accelerators at the center of nearly every AI buildout, so data center revenue and its chips staying the default choice for training and running large models drive its role in the take., BroadcomIt builds custom AI chips for hyperscalers and the networking silicon that ties clusters together, so the take needs custom accelerator programs at major cloud companies to keep expanding., Advanced Micro DevicesIt offers the credible alternative accelerator and the take assumes the pie grows fast enough that second place still wins, so it needs buyers wanting a second supplier and real hyperscaler orders for its MI-series chips., TeslaIt is betting its future on AI for self-driving and robotics and buying enormous amounts of compute to get there, so the take needs its AI programs to keep progressing toward products people pay for.20%
- Data centers are running hot and power grids are tightening in key regions, so the take needs electricity demand from AI buildouts to keep rising and long-term power deals to keep flowing to generators and utilities. · GE VernovaIt sells the gas turbines and grid equipment being ordered to power new data centers, so the take needs data center power demand to keep outrunning supply and its backlog to keep building., VistraIt owns power generation in regions where data center demand is tightening the grid, so the take needs electricity demand from AI to keep pushing power prices and long-term contracts in its favor., Constellation EnergyIt runs the largest nuclear fleet in the US and hyperscalers are signing long-term power deals for that clean, always-on capacity, so the take needs those data center contracts to keep coming., EatonIt makes the switchgear, transformers, and power distribution equipment every new data center needs regardless of owner, so the take needs its data center orders and backlog to keep growing.16%
- Networking, cooling, and connectivity inside AI clusters are essential and getting more specialized, so the take needs hyperscalers to keep choosing best-of-breed suppliers for each piece of the buildout. · Arista NetworksIt sells network switches that connect thousands of accelerators into usable clusters, essential gear regardless of chip choice, so the take needs hyperscalers to keep selecting it for AI network buildouts., VertivIt sells the liquid cooling and power management systems that AI-density data centers cannot run without, so the take needs the shift to liquid cooling to keep accelerating and drive its order growth., Astera LabsIt makes the connectivity chips that move data between accelerators, memory, and processors inside AI servers, so the take needs AI server volumes to keep climbing and its parts to stay designed into leading platforms., CoherentIt makes optical transceivers and lasers that carry data between racks as clusters scale beyond what copper can handle, so the take needs the move to optical networking inside AI data centers to keep accelerating., Credo TechnologyIt sells high-speed cables and connectivity chips linking servers inside AI clusters, so the take needs its hyperscaler customers to keep expanding orders as cluster counts grow.15%
- High-bandwidth memory and storage are tight enough to support pricing power, so the take needs demand for AI training and inference to keep outrunning the supply of both fast memory on accelerators and cold storage for datasets. · Micron TechnologyIt makes the high-bandwidth memory packaged directly onto AI accelerators, and tight supply means the take depends on HBM demand outstripping supply so pricing and margins hold up., Western DigitalIt sells the drives that store data AI models train on and generate, the cheaper end of the memory squeeze, so the take needs data center storage demand to keep tightening the market.8%
- The take assumes AI demand compounds, so it needs the largest cloud and AI spenders to keep investing capital at scale and showing revenue growth from AI products and services. · MicrosoftIt is spending tens of billions quarterly on AI infrastructure and selling results through Azure and Copilot, so the take needs its AI capital spending and cloud growth to stay robust as proof demand is real., AlphabetIt builds its own TPU chips, spends heavily on data centers, and monetizes AI through search and Google Cloud, so the take needs sustained capital spending and cloud growth to confirm compounding demand.8%
- Cash10%