bartek@aws: ~/news
$ whoami
$ AWS Architect · DevOps · Cloud
Thursday, November 19, 2026

Exa Cranks Up Embedding Speed by 50% on AWS AI Chips

Exa's neural search engine just got a serious performance boost by optimizing embedding inference on AWS Neuron hardware using custom NKI kernels. The result? A 47% improvement in both performance and cost efficiency when indexing billions of documents—perfect for anyone building large-scale vector search systems.

source: [aws/machine-learning-blog]