Research and information only — nothing here is investment advice or a recommendation. Public markets have organized themselves around this database's layers; this article maps the segments so the Stocks page reads like the rest of the site.
The segments, layer by layer
Silicon: designers (NVDA, AMD, custom-ASIC houses AVGO/MRVL), the foundry (TSM), memory (hynix/MU/Samsung), equipment (ASML and the tool quartet). Systems: server builders (SMCI, Dell, Taiwanese ODMs), networking (ANET, optics names, connectors). Facilities: data-center REITs (EQIX, DLR) and the private giants behind them. Power: merchant nuclear (CEG, VST, TLN), turbines and grid gear (GEV, Siemens Energy, ETN, VRT), SMR speculatives (OKLO, SMR). Platforms: hyperscalers and the listed neoclouds (CRWV, NBIS, IREN). Most frontier labs remain private — exposure to them routes through their suppliers and investors.
What 2023–25 actually rewarded
The market paid, in rough order: the chip monopoly, then the picks-and-shovels around it (Vertiv-class thermal/electrical, optics), then the power trade (the 2024 surprise — utilities as AI stocks), then the contract-backed neoclouds. Each wave started when a bottleneck became legible in earnings calls — a pattern worth noticing more than any single name.
The honest uncertainties
Concentration: a handful of buyers (four hyperscalers plus OpenAI) fund most of the chain — their capex decisions are systemic risk. Circularity: vendors investing in customers (the bubble debate) makes demand harder to read. Depreciation: GPU fleets age against an annual roadmap. Efficiency shocks: R1-style results reprice everything in a day. And policy: export rules rewrite addressable markets overnight. None of these has a settled answer; our Trends page tracks the labeled numbers on all of them.
Disclosure once more: this site is a research database. It does not recommend, rate or advise on securities. Verify every figure against primary sources.
