Uber expands AWS contract, betting on Amazon chips over Google and Oracle

Uber is expanding its contract with Amazon Web Services and moving more of its service's functions onto Amazon chips. It is a major win for AWS against Google Cloud and Oracle, which also competed for the business. The decision by the world's largest ride-hailing aggregator tells the market that Amazon's custom AI processors are starting to seriously challenge NVIDIA's standard GPU solutions in the enterprise segment.

Amazon has been developing its own AI chips for several years. Trainium is optimized for training neural networks, Inferentia for running them in production. These are the processors Uber is now shifting part of its AI workloads to. Previously, those tasks required standard GPU clusters or the computing capacity of other cloud providers.

AWS positions its silicon as a cheaper alternative for inference tasks: at comparable throughput, the cost of inference on Inferentia is lower than on NVIDIA H100, which is critical for high-frequency production services.

The expanded contract targets real-time AI workloads: driver-matching algorithms, demand forecasting by district, dynamic pricing, fraud detection systems and anti-abuse filters. All of these components require constant ML model inference with minimal latency. Uber runs such workloads around the clock in dozens of cities at once, and inference cost is one of the key lines in the company's cloud budget. Cutting that line by 20-30% means tens of millions of dollars in savings every year.

For Google and Oracle, the news reads as a public jab. Both companies are aggressively building out AI infrastructure and courting enterprise customers. Google Cloud is betting on fourth-generation TPUs and A3 clusters based on NVIDIA H100, while Oracle is betting on A100/H100 superclusters, positioning them as the highest-performing AI infrastructure in the cloud. Uber's choice of AWS and its custom chips signals that the economics of AI infrastructure are starting to work in Amazon's favor.

Amazon Web Services remains the largest cloud provider, with about 30% of the market by revenue. But AI has become a new field of competition: enterprise customers are rebuilding their infrastructure and looking for ways to cut inference costs. Every major move to AWS custom silicon is a market signal and a marketing case for negotiations with other businesses. In that sense, Uber is the ideal showcase story for Amazon's sales team.

Uber is among the most technically demanding platforms in the world. The company handles millions of trips every day in 70+ countries, runs Uber Eats and Uber Freight, and operates large-scale R&D programs in autonomous driving. At that scale, any platform choice follows detailed technical and financial analysis. When Uber picks a particular type of AI chip, it is not a marketing partnership but an engineering decision with consequences lasting years.

The expanded AWS partnership is further evidence that the race for AI infrastructure has long moved beyond the familiar NVIDIA-versus-everyone-else standoff. Amazon, Google and Microsoft are building their own processors, and large technology companies must make a concrete choice: whose silicon will underpin their next generation of AI. Uber chose Amazon.