Uber deploys AWS custom chips for real-time rider matching and AI model training

Uber has rolled out Amazon's purpose-built Graviton4 and Trainium3 chips across its platform at scale. The Graviton4 processors power the matching engine that connects riders and drivers, processing millions of assignments daily with sub-second latency while reducing infrastructure costs. Trainium3 accelerates the continuous training of machine learning models that enhance every ride experience. The deployment underscores AWS's strategy of delivering custom silicon solutions that enable customers to scale efficiently and accelerate AI innovation at competitive price points.