Particle.news
Download on the App Store

Uber Taps AWS Graviton4 for Real-Time Trips, Pilots Trainium3 for AI Training

The shift signals a broader move to custom cloud chips for cheaper, greener AI.

Overview

  • Uber, which announced the expansion Tuesday, is moving more real-time Trip Serving Zone compute to AWS Graviton4 to reduce latency and power use.
  • Trip Serving Zones are the backend systems that match riders with drivers and route deliveries in milliseconds during demand spikes.
  • Uber has begun piloting AI training on AWS Trainium3 to improve driver or courier selection, arrival time estimates, and in-app recommendations.
  • The Trainium work remains a test as Uber keeps a multi-cloud setup that includes Google Cloud and Oracle.
  • For AWS, Uber’s deployment serves as a proof point as big customers weigh custom chips over general GPUs for price and performance.