Overview

• Amazon's 'Build on Trainium' program makes 40,000 first-generation Trainium chips available to researchers with $110 million in cloud credits.
• The program seeks to attract researchers from top universities like Carnegie Mellon and UC Berkeley to develop AI models using Amazon's chips.
• AWS's strategy includes open-sourcing research outcomes and providing detailed chip programming documentation, contrasting Nvidia's closed CUDA system.
• Amazon's investment in custom AI chips is part of a broader trend among tech giants to reduce reliance on Nvidia and enhance control over AI infrastructure.
• Despite its push for Trainium, AWS is also negotiating a $475 million deal with IBM to supply Nvidia GPUs, highlighting ongoing demand for Nvidia's technology.