Investing in AI today is mostly investing in two things: companies that train large models, and companies that build the infrastructure those models run on. Neither is cheap, and the winners list is anything but settled.
Mistakes I try to avoid
- Concentrating in one name. NVIDIA's run is impressive, but the tape teaches you not to fall in love with a single ticker. A 30% drawdown in a 30% portfolio position is a 9% portfolio drawdown.
- Buying the loudest names. The most-talked-about stocks are usually already priced for perfection.
- Ignoring economics. AI infrastructure has insane capex; cash conversion lags revenue. Watch free cash flow, not just top-line growth.
My personal stance
I'm staying with broad Nasdaq exposure (TQQQ + QLD). It captures the AI winners (NVDA, MSFT, GOOG, META, AVGO) without me having to be right about which one wins. If a clear thematic ETF with reasonable structure shows up I might add a small dedicated bucket — until then, the index is enough.