Nvidia · one lens
Hamilton Helmer on Nvidia
Which of the seven Powers Nvidia actually holds, tested against its own rising competitor-edge share.
Helmer’s test for a real competitive advantage is narrow: a Benefit plus a Barrier that survives someone actually trying to arbitrage it away (7 Powers). Applied to Nvidia, three of his seven categories hold up against the record better than a vague “moat” story would.
Scale Economies: chip design and the CUDA software stack are fixed costs spread over the largest unit volume in the industry, the same arithmetic Helmer uses to explain Netflix’s content spend — a smaller rival amortizes the same R&D over far fewer units and can’t close the resulting cost gap without first matching volume it doesn’t have.
Switching Costs: CUDA is close to a textbook case — the cost of leaving isn’t a contract, it’s the retraining and re-tooling. SemiAnalysis called it a “monopoly… nowhere close to being broken” even as OpenAI, Meta, Microsoft, and Google jointly built Triton specifically to route around it, and Nvidia has since ported CUDA to RISC-V rather than let a new instruction set become an exit ramp (texxr.com/833971, texxr.com/865597, texxr.com/888127). A coalition of Nvidia’s own biggest customers building an alternative, rather than simply switching to one, is itself a measure of how expensive switching still is.
Cornered Resource, more provisionally: leading-edge fabrication capacity at TSMC functions like one, though it’s shared rather than exclusive, and Nvidia’s own record shows the limit of that claim — it tested Intel’s 18A process as a second source and walked away without shipping on it, a non-replicable test cutting the other way.
None of this is fixed. Nvidia’s competitor-predicate share of its own coverage graph rose from zero to 23% of tracked edges between 2024 and early 2026, as AMD, Qualcomm, and MediaTek all now contest the data-center CPU seat. Helmer’s framework never claims a Power is permanent — only that the question is whether the Barrier holds when someone tries to break it. On the record so far, CUDA’s has.
The seven Powers, one by one
| Power | Status | Why |
|---|---|---|
| Scale Economies | held | chip design and CUDA are fixed costs spread over the industry's largest unit volume; a smaller rival can't close the cost gap without first matching volume it doesn't have |
| Network Economies | absent | not argued in the record |
| Counter Positioning | absent | not argued in the record |
| Switching Costs | held | CUDA's retraining/re-tooling cost; customers built Triton to route around it rather than simply switch off it |
| Branding | absent | not argued in the record |
| Cornered Resource | contested | TSMC leading-edge capacity functions like one but is shared, not exclusive — and Nvidia's own Intel 18A test, walked away without shipping, cuts against it |
| Process Power | absent | not argued in the record |
An editorial application of Hamilton Helmer's published framework — not their actual view, affiliation, or advice.
Where this lens comes from
Hamilton Helmer's framework is set out at the lens page, drawn from 7 Powers. This page applies it to Nvidia and nothing else — the company's full record is in the dossier.