Ex-NVIDIA engineer recalls Jensen Huang backing his 2012 deep learning cluster
GregoryDiamos · x · 2026-11-15
Gregory Diamos, who worked on CUDA at NVIDIA, recalls that by 2012 it was clear to him AI was about to work, so he started building the deep learning cluster used to train Deep Speech.
Key moments:
- Jensen Huang visited, asked why he was building it, supported the work, and featured it in the GTC 2015 keynote (Diamos links the intro and Deep Speech segments).
- Huang then pivoted NVIDIA to AI in a way that made GPUs worse at traditional graphics, seemingly hurting the cash cow.
- Diamos thought at the time that no other CEO would have made that pivot — and it inspired him to learn about founding companies.
A first-hand industry anecdote about NVIDIA's pivotal AI bet.
More from Companies & People
- 40 Tech Newsletters Worth Following, Curated Across AI, Systems, and Product — bibryam · 2026-11-15
- Amazon planning World Product Graph to index the web's products for AI shopping agents — jomic01 · 2026-11-15
- Jensen Huang: China now likely produces the vast majority of the world's open source contributions — _AustinCalvert_ · 2026-11-15
- Critic clarifies: only Anthropic leadership, praises Opus 5.5 and Fable 5.2 for coding — DeryaTR_ · 2026-11-15
- NYT: AI startups and VCs hire 'ecosystem' leads to double down on in-person networking — nordicinst · 2026-11-15
- 39-minute interview: how Kimi's founder built a $5B+ AI company from China — MaxLi_Columbia · 2026-11-15