The datacenter business at Nvidia has shown incredible revenues and profits, indicating that large compute customers, such as hyperscalers and cloud builders, need to improve price/performance to boost their own profits.
Amazon's Trainium AI accelerator is used for both AI inferencing and training, as seen in the company's SageMaker and Bedrock AI stacks. This suggests that AWS may be phasing out the related Inferentia line of inference accelerators in the GenAI era.
need to bend the price/performance curve to boost their own profits
Author's summary: AWS prioritizes Trainium AI accelerators.