Ground Floor and the Stack
lindo@mac_matz
−2.7%
since Oct 6
The idea
I believe the next decade favors two things: businesses with real advantage in the physical world, and Africa's mispriced ground floor, starting with South Africa. So this take splits roughly evenly between African miners, infrastructure and transition-metal names priced for decline, and an AI stack running from energy, water and grid through machinery and chips to applied systems and physical-world data moats. It also tracks a sleeve mirroring Coatue's largest public positions as a signal worth watching.
What it owns
- South African and African businesses with real cash flows and transition metal reserves trade as if structural improvement never arrives. · iShares MSCI South Africa ETFIt holds South African banks, insurers, telecoms and retailers with real cash generation valued as if the country never improves., Sibanye-StillwaterIt owns platinum group metals essential to the energy transition while the market prices the company as if that demand never materializes., SasolIt operates South African energy and chemicals assets generating cash that the global market undervalues relative to the cash flows they produce., KaroooooIt gathers fleet data from vehicles across Africa, a physical information advantage that competitors cannot replicate from a desk or scrape online., IHS HoldingIt owns telecom towers that African connectivity depends on, infrastructure that compounds returns as mobile data demand grows across the continent., Gold FieldsIt is a South African mining house with real cash flows trading as if in decline, currently positioned as an acquirer in a major deal., Freeport-McMoRanIt supplies copper for both the energy transition and AI buildout, with demand from grids, data centers and electrification outrunning new supply., Jumia TechnologiesIt builds commerce and logistics infrastructure across Africa, a bet on the path to profitable payments and delivery networks becoming viable.39%
- The machines and silicon that the AI stack requires, from lithography equipment through networking infrastructure. · NVIDIAIt is the compute standard for AI training and inference, the center of the technology stack from chips through applied systems., ASML HoldingIt is the only maker of advanced lithography machines that next-generation chip production requires, a monopoly in irreplaceable equipment., Taiwan SemiconductorIt is the foundry bottleneck through which the entire advanced chip supply chain flows, concentrated in one location., Arista NetworksIt builds the routing that connects AI clusters together, the networking layer between chips, orchestration and cloud infrastructure.18%
- The power generation, grid infrastructure and water systems that every data center needs before running a single model. · GE VernovaIt builds the turbines and grid equipment that data centers and energy infrastructure require as the physical ground floor of the AI stack., XylemIt provides water cooling, treatment and infrastructure that data centers and the grid assume, with rising demand from infrastructure spending., VistraIt sells the electricity that AI systems require to operate, the foundational layer of power that the entire stack depends on.13%
- The largest public positions held by Coatue this quarter, adopted as a sleeve to align with their reading of the AI supply chain. · Lam ResearchIt makes the wafer equipment that chipmakers need to expand capacity, with rising capital spending as the AI buildout continues., Micron TechnologyIt produces high-bandwidth memory that AI compute consumes faster than the industry can expand supply, keeping pricing firm., SpaceXIts edge is entirely physical: rockets, launch sites and satellites, with economics that compound as launch and satellite services scale., Applied MaterialsIt sells the equipment that every fab expansion requires, with rising demand as semiconductor capital spending continues to grow.10%
- Companies whose competitive advantage comes from information gathered in the physical world, information that AI tools cannot scrape from the internet. · SamsaraIt collects operational data from trucks, sites and equipment, physical-world information that AI tools cannot find on the internet., CoStar GroupIt gathers property data through people in the field, an asymmetric information advantage that competitors cannot buy or scrape online.8%
- Cloud, data and software businesses that turn AI capabilities into work actually completed, standing in for applied AI leaders. · MicrosoftIt rebuilds how work gets done at scale, with cloud infrastructure underneath and AI agents on top, converting AI products into enterprise revenue., PalantirIt is software that acts inside real operations rather than waiting to be used, the closest listed proxy for applied AI in commercial work., ServiceNowIt transforms workflows into AI agents that do the work directly, a systems-level change where software executes rather than merely assists.7%
- Cash5%