Large companies are rethinking how they run artificial intelligence workloads in the cloud. Uber is one of the latest examples, expanding its use of AWS chips to support its AI systems.
At the centre of this change are AWS-designed chips like Graviton and Trainium. Reuters reports Uber is increasing its use of the hardware to power AI models and backend systems for its ride-hailing and delivery platforms. Uber’s AI models work on core functions like matching riders with drivers, estimating trip times, setting prices, and managing food delivery routes. Such tasks rely on large volumes of data and constant updates, which can push up cloud costs.
Custom chips offer a way to manage price pressure. AWS says Graviton can improve price-performance compared to traditional x86-based instances, while Trainium is designed to lower training costs. The hardware may help companies like Uber run more AI tasks without a similar rise in spending.
How custom chips change cloud use
The decision to explore alternative hardware ties closely to scale for Uber. The company operates in dozens of countries and processes millions of transactions each day. Even small gains in efficiency can matter in a network of that size.
According to Reuters, Uber is using AWS chips to improve both training and inference workloads. Training refers to how AI models learn from data, whereas inference is how those models make decisions in live systems. Both stages can be costly, but inference often runs continuously in production, making efficiency particularly important.
Chips like Trainium are designed for high-throughput machine learning tasks, which can help minimise the time and cost needed to train models. Graviton, which is built on ARM architecture, is often used for general workloads that benefit from lower power use and better cost control. Together, they give enterprises more options in how they run AI systems in the cloud.
Balancing cost and flexibility
Cloud strategies are also changing. Companies are taking a more active role in how workloads are structured, from choosing instance types to tuning models for certain chips and balancing cost against performance.
This approach can add complexity, however. Developers need to adjust software for ARM-based processors or specialised AI chips, and it may require closer coordination with cloud providers.
Uber’s move comes at a time when AI workloads are expanding in many industries. From finance to retail, companies are using machine learning for tasks like fraud detection, demand forecasting, and customer support. As these systems grow, so does the need to manage the cost of running them.
Custom silicon is one response. Cloud providers like AWS are building their own processors, which gives them more control over pricing and performance. It also raises questions about flexibility. Companies that build around specific cloud chips may find it harder to move workloads between providers.
Uber’s use of AWS chips shows how these trade-offs are playing out in practice. Rather than moving away from the cloud, the company is using more specialised cloud hardware. Reuters does not detail the exact scale of Uber’s deployment, but it says the chips support important AI-driven functions in the platform.
Rising cloud costs are forcing more companies to rethink how they run workloads. Custom chips may not replace general-purpose compute, but they are becoming part of the mix.
Uber’s move reflects a broader change in how enterprises use the cloud. The focus is increasingly on running workloads more efficiently. Companies will need to balance cost and flexibility, and custom silicon is likely to play a larger role.
(Photo by Erik Mclean)
See also: Cloud costs rise as AI moves into core business systems
Want to learn more about Cloud Computing from industry leaders? Check out Cyber Security & Cloud Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
CloudTech News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
