Uber Taps AWS Chips for Ride Matching, AI Training
The company is moving more Trip Serving workloads onto Graviton4 processors and piloting Trainium3 for AI training as it looks to cut latency, improve efficiency and handle demand spikes more smoothly.
Topics
News
- Anthropic Poaches Microsoft AI Executive for Infrastructure Role
- Uber Taps AWS Chips for Ride Matching, AI Training
- OpenAI Urges Grid Upgrades, Stronger Safety Nets for AI Age
- Wipro Buys Mindsprint in $1 Billion-Plus Olam Deal
- Magicpin Fast-Tracks AI Rollout as LPG Crisis Disrupts Restaurants
- Coforge Names Sunil Fernandes COO Amid AI-Native Push
Uber is expanding its use of Amazon Web Services’ custom chips to run more of the real-time systems that match riders, drivers and deliveries, while also beginning early tests of AWS silicon for AI model training.
The company said it is moving more of its Trip Serving Zones, the backend infrastructure that processes location data and rapid-fire predictions for ride and delivery requests, onto AWS Graviton4 processors.
It has also started piloting AWS Trainium3 to train some of the AI models behind demand forecasting, driver and courier assignment, arrival estimates and delivery recommendations.
Uber said the Graviton4 shift is aimed at helping the platform scale more efficiently during demand spikes while cutting latency, energy use and costs.
“Uber operates at a scale where milliseconds matter,” Kamran Zargahi, the company’s Vice-President of Engineering, said in the AWS statement. “Moving more Trip Serving workloads to AWS gives us the flexibility to match riders and drivers faster and handle delivery demand spikes without disruption.”
Alongside infrastructure changes, Uber has also started piloting the use of AWS Trainium chips to train some of its AI models. These models are used to predict demand, assign drivers or couriers, estimate arrival times, and suggest delivery options based on historical trip data.
Training such models requires significant compute resources, and the company said Trainium could offer a more cost-efficient way to handle these workloads as it scales.
“By starting to pilot some of our AI models on Trainium, we’re building a technology foundation that will make every Uber experience smarter—so we can keep our focus where it belongs: on the people who use Uber every day,” Zargahi said.
AWS, which provides the underlying cloud infrastructure, positioned the collaboration as part of its broader push into large-scale, real-time applications.
“Uber is one of the most demanding real-time applications in the world, and we’re proud to be an important part of the infrastructure powering their global operations,” said Rich Geraffo, vice president and managing director of North America at AWS. “We’re helping Uber deliver the reliability hundreds of millions of people count on today—and the AI-powered experiences that will define ride-sharing and on-demand delivery tomorrow.”
