Most AI stock content falls into one of two buckets. Either it's hype, a chart with an arrow and a ticker, or it's so hedged it tells you nothing. Margin of Signal is my attempt at the space in between: specific enough to be useful, honest enough to show where I could be wrong.
What I cover
The focus is the physical build-out behind AI, the companies that sell the picks and shovels. I think about it as four layers:
- Compute: the chips that train and run models.
- Memory: DRAM and high-bandwidth memory, often the real bottleneck.
- Networking: the switches, optics and interconnects that tie clusters together.
- Power: the grid, cooling and energy supply every data center depends on.
Demand moves through these layers at different speeds, and the gaps between them are where the interesting setups tend to show up.
How each post is built
Every analysis follows the same structure so you can compare them over time:
- The numbers. Revenue, margins, valuation and what's actually changed, with sources.
- The thesis. One clear claim about what the market might be missing.
- The risks. What would prove the thesis wrong, stated up front, not buried at the end.
What this isn't
This isn't investment advice and it isn't a stock-picking service. I share how I'm thinking about the sector, including positions I may hold, so you can check my reasoning against your own. Do your own research before you act on anything here.
If that sounds useful, subscribe below and you'll hear when new analysis goes up.