From AI Insider · James Dargan · · 1m 30s
AWS Challenges Nvidia's PyTorch Edge with Trainium
AWS launches TorchNeuron to compete with Nvidia's PyTorch dominance while Nvidia releases open-weight Nemotron models for agents.
In brief
AWS launches TorchNeuron to compete with Nvidia's PyTorch dominance while Nvidia releases open-weight Nemotron models for agents. AWS introduces TorchNeuron to run PyTorch apps on Trainium chips, while Nvidia expands open-weight Nemotron models for AI agents. Originally reported by AI Insider.
AWS releases TorchNeuron
Amazon Web Services launched TorchNeuron, a tool enabling AI startups to move applications to Trainium, Amazon's AI chip.
TorchNeuron works with PyTorch
TorchNeuron integrates with PyTorch, allowing apps to run on Trainium by changing one line of configuration without code rewrite.
85% of AI apps use PyTorch
Amazon estimates 85% of AI applications are written in PyTorch, where Nvidia's CUDA software has long been embedded.
Trainium offers cloud access
AWS positions Trainium as a stronger choice than Nvidia-aligned neoclouds like CoreWeave, Nebius, and Firmus for startups.
Colin Brace on Trainium value
Colin Brace, AWS VP of Technology Software, says Trainium handles both training and inference on one platform, improving audio processing efficiency.
Nvidia releases Nemotron 3.5 Super VL
Nvidia launched Nemotron 3.5 Super VL, a 120-billion-parameter model for enterprise agents matching Nvidia's larger Nemotron 3 Ultra in instruction-following.
Super VL outperforms competitors
Nvidia claims Super VL completes tasks faster than Anthropic and OpenAI models when post-trained on agentic tasks like document intelligence and financial accounting.
Nvidia introduces Bolt models
Nvidia released two lightweight Bolt models for local AI tasks on RTX Spark chips and DGX Spark workstations.
The takeaway
AWS introduces TorchNeuron to run PyTorch apps on Trainium chips, while Nvidia expands open-weight Nemotron models for AI agents.
