This page is the map of everything on this site: what each layer of AI infrastructure actually is, which company sits where, and which tool answers which question. If you landed here from a search result and want to know what else is here, start below.
What are you looking for?
- I am new to AI infrastructure — start with the layer map
- I want two companies compared — head-to-head posts
- I need to run a number — free calculators
- I want to use AI in my own research — prompts and tool comparisons
- I want something that keeps updating — the quarterly tracker
The AI infrastructure value chain
Money flows through these layers in order. A single headline usually only touches one of them, which is why the same news moves some stocks and not others.
1. Power generation
Before a data center computes anything, someone has to generate the electricity. This is where the demand shows up first and where the lead times are longest.
- Utility Stocks and Data Center Electricity Demand — who actually captures the demand
- SMR Stocks vs Uranium Stocks — two different bets on the same theme
- Oklo vs NuScale — the two most traded SMR pure plays
2. Transmission and distribution
Generated power still has to reach the rack. Transformers and switchgear have become the physical bottleneck, with lead times measured in years rather than months.
- Transformer Stocks — 160-week lead times and who benefits
3. Cooling
Dense GPU racks produce more heat than air can carry away. The shift to liquid is the single clearest structural change in data center design right now.
- Immersion vs Direct Liquid Cooling — how the two approaches differ
- Liquid Cooling Suppliers — who supplies the parts
- Data Center Water Stocks — the water side of cooling
4. Servers and racks
Where GPUs get assembled into something a hyperscaler can deploy. Margins here are thin and depend heavily on how much of the stack a company integrates itself.
- Best AI Server Stocks — where the margin stories diverge
- Super Micro vs Dell — concentrated exposure vs diversified
5. Memory
High-bandwidth memory is the component AI accelerators cannot run without, and supply is decided at chip design time rather than at purchase.
- HBM Memory Stocks Guide — the full supply picture
- Micron vs SK Hynix — the two listed ways to own HBM
- HBM4 Stocks — the next generation
Across the whole chain
- AI Power Infrastructure Stocks — five picks across layers
- AI Infrastructure ETFs — if you would rather not pick names
- AI Infrastructure Dividend Stocks — the income side
- Small Cap AI Infrastructure — higher risk end of the chain
Head-to-head comparisons
When two companies get mentioned in the same sentence, the useful question is usually not which is better but which risk each one carries.
- Micron vs SK Hynix — memory
- Super Micro vs Dell — servers
- Oklo vs NuScale — small modular reactors
- Immersion vs Direct Liquid — cooling methods
- SMR vs Uranium — two routes to nuclear
Things that keep updating
Most posts are written once. These are not.
- Hyperscaler Capex Tracker — revised after each earnings season
- Free calculators — tax, FX, retirement, position sizing
Using AI in your own research
I use AI to draft and check this site, and it fails in specific, repeatable ways. These posts are about those failures as much as the wins, because knowing where a tool breaks is what makes it usable.
- AI Stock Analysis Prompts — five prompts and the mistake that created each one
- Claude vs ChatGPT — same 10-K, same day, different answers
Desk gear
Kept deliberately separate from the investing posts. These are affiliate reviews of things on my own desk, compared on specs rather than hype.
How this site checks its numbers
Every figure here is meant to trace back to a filing, a regulator, or a company release rather than a summary of a summary. When a number turns out wrong, the correction stays in the post instead of quietly disappearing. Company business descriptions get re-checked against current filings, because a company that sold a division four years ago should not still be listed as a play on it.
That is also why comparisons here stay qualitative more often than you might expect. Prices and market caps go stale within days, so the parts that hold up longer are structure, contract terms and risk profile.
Last updated: 2026-08-07. This site is information only and is not investment advice.