Triton (OpenAI)
Python GPU kernel language cracking the CUDA moat
Triton is OpenAI's open-source Python-embedded language for writing fused GPU kernels that rival hand-tuned CUDA, now the kernel authoring layer inside PyTorch 2.x's torch.compile/Inductor. It is central to portability efforts because AMD, Intel, and others target Triton to bypass CUDA lock-in. FlashAttention-style kernels and much of modern training efficiency flow through it.
Backs
torch.compile / Inductor
Portability
AMD, Intel targets
Origin
OpenAI
Role
PyTorch Inductor backend
How it fits the stack
Triton (OpenAI) with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
Triton (OpenAI) in the AI stack. Triton (OpenAI) with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 5 entities, 6 relationships
- AMD —used by→ Triton (OpenAI)
- PyTorch —used by→ Triton (OpenAI)
- PyTorch —uses→ Triton (OpenAI)
- PyTorch —depends on→ Triton (OpenAI)
- Triton (OpenAI) —competes with→ CUDA / software moat
- Triton (OpenAI) —designs→ OpenAI
Context — capital, rivals, policy · · 2