The diagnosis:
Jensen Huang's read of the GPU market was that speed of iteration, not individual chip performance, would determine the winner. Any single chip could be matched or beaten by a competitor given enough time. But a competitor that was always one generation behind could never catch up — by the time they studied an Nvidia chip and designed a response, the next Nvidia chip was already shipping.
The guiding policy:
Release a new GPU architecture every six months. Organize the company around this cadence. Accept the enormous cost and risk of running three parallel design teams.
The coherent actions:
- Three parallel design teams — Each team works on the next-generation architecture in a staggered pipeline. This required hiring and organizing talent in a way that most chip companies considered wasteful.
- Relentless focus on GPU — All resources went to graphics processing. No diversification into CPUs, memory, or other semiconductor markets (at the time). Concentration of effort.
- Developer ecosystem investment — CUDA and developer tools ensured that software was optimized for Nvidia's architecture, creating switching costs and network effects.
Why it worked as good strategy:
- It had a clear diagnosis (iteration speed wins)
- A guiding policy that created focus (six-month cadence)
- Coherent actions (three teams, GPU-only, developer tools)
- It was unexpected — competitors thought it was impossible, which gave Nvidia time to build the moat before anyone tried to copy it
The outcome:
Within a few years, most competitors were eliminated. 3dfx went bankrupt. ATI was acquired by AMD. Nvidia became the dominant GPU company — a position it has held and expanded into AI and data center computing decades later.