JAX / XLA
Google's ML framework — the TPU-native alternative to PyTorch
JAX with the XLA compiler is Google DeepMind's primary training framework and the software that makes TPUs first-class, powering Gemini. It is the main non-PyTorch training stack and a strategic lever letting Google escape the CUDA/NVIDIA dependency the rest of the industry lives under.
Backs
Gemini on TPU
Compiler
XLA
How it fits the stack
JAX / XLA 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.
JAX / XLA in the AI stack. JAX / XLA 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) — 4 entities, 3 relationships
- Google DeepMind —uses→ JAX / XLA
- Google TPU —used by→ JAX / XLA
- JAX / XLA —competes with→ PyTorch
Context — capital, rivals, policy · · 1