Nvidia's real moat is not what you think

CUDA. It's always been CUDA. The chips are the vessel. The developer lock-in is the moat.

Why the AMD "threat" is overblown

Switching costs in AI infrastructure aren't about the silicon. They're about seven years of tooling, docs, and internal Slack threads at every hyperscaler.

The trade

Long the dip. Stop-loss on a break of the 200-day. Take partials on any new all-time high.