Uber Utilizes Amazon's Custom Chips for AI Enhancements

UBER has expanded its partnership with Amazon Web Services, adopting Amazon's custom Graviton CPUs and Trainium AI processors to speed up computing and train the machine learning models that power its apps . Graviton will handle core infrastructure tasks including ride matching, dispatch systems, and delivery optimization, while Trainium3 handles AI model training — analyzing data from billions of rides to determine optimal driver routing, calculate arrival times, and recommend delivery options.

The shift away from rival cloud providers toward AWS signals a meaningful strategic realignment. Uber joins AMZN cloud customers Anthropic, OpenAI, and Apple in adopting Trainium-based infrastructure, validating the custom chip program as a credible alternative to NVDA GPU clusters. As AI workloads grow heavier and more expensive, specialized hardware like Trainium offers cost-efficiency advantages that standard GPUs cannot match at scale.

The move also fits Uber's broader capital-light strategy: rather than building its own self-driving fleet, the company partners with autonomous vehicle developers like Waymo and Rivian, minimizing capital expenditure while staying competitive . Leveraging AWS's chip infrastructure extends the same logic to AI compute, letting Uber focus engineering resources on product differentiation rather than raw infrastructure buildout.