Thesis holds
NVIDIA Corporation
ConvictionConviction 4 out of 54/5
The compute layer of the AI build-out, protected by CUDA rather than by silicon alone.
Why I invest
I own NVIDIA because the moat is the software, not the chip. CUDA is fifteen years of accumulated tooling that every serious ML team already builds against, and switching means rewriting work that currently runs. Competitors can match raw FLOPS long before they match that. I expect data-centre revenue to keep compounding while gross margin stays above the mid-sixties; if margin slips materially, it tells me pricing power has gone and the moat argument was wrong.
Bull case
CUDA lock-in keeps customers on-platform even when rivals close the hardware gap
Data-centre demand broadening from training into inference
Pricing power visible in sustained gross margin
Bear case
Hyperscalers shipping credible in-house silicon at scale
Export controls removing a whole region of demand
Any quarter where margin falls without a one-off explanation
What would prove me wrong
- 1Gross margin drops below 65%Holding
- 2A major hyperscaler moves production workloads to in-house siliconNo evidence yet
- 3Export restrictions remove a region worth more than 15% of data-centre revenueHolding
Position
- $96.50
- $245.00
Decision journal
- BoughtConviction 4 out of 5
Opened the position after the annual report. The risk factors name customer silicon before I did, which I take as a sign the concern is real rather than mine alone.
- HeldConviction 3 out of 5
Read the Reuters piece. Holding, but this is the assumption I said I would watch, and I want to see it in a filing before I act on a press report either way.
Do this with your own holdings
Write down why you own something and what would change your mind. Okidor reads every filing, earnings call and material news item against it, and tells you when it matters. Three theses free.
Start for free