CUDA

CUDA is NVIDIA’s computing architecture and developer platform for GPU-accelerated computing, framed by Jensen Huang as the strategic bridge from graphics chips to general accelerated computing and AI infrastructure.

Key facts

• Type: GPU computing architecture / developer platform
• Maker: NVIDIA
• Jensen traces CUDA through earlier steps: programmable pixel shaders, FP32 in shaders, Cg, and then CUDA ^[src-065].
• The existential bet was putting CUDA on GeForce, even though it raised costs for a consumer product and hurt gross margins, because NVIDIA wanted to become a computing architecture company ^[src-065].
• Jensen argues that developers choose platforms because of install base; GeForce put CUDA in front of students, researchers, scientists, and university labs before cloud AI infrastructure existed ^[src-065].
• CUDA’s durability comes from balancing specialization and generality: it accelerates workloads while evolving fast enough to match changing AI algorithms ^[src-065].

Related entities

• NVIDIA
• Jensen Huang
• GeForce
• Nvidia Blackwell NVL72
• NVIDIA Vera Rubin

Related concepts

• Install-Base Moats
• GPU Supply as AI Strategy
• Extreme Co-Design

Source references

• ^[src-065] Lex Fridman – “Jensen Huang: NVIDIA – The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494” (2026-03-23)