The CEO of Amazon Web Services highlighted their strong partnership with Anthropic and the rapid advancements of their Trainium chips, emphasizing AWS’s integrated technology stack and commitment to expanding compute capacity to meet growing AI demands. AWS supports diverse hardware needs, continues to scale infrastructure, and sees generative AI and agent technologies as key future drivers, with many customers already trusting AWS for large-scale AI workloads.
The CEO of Amazon Web Services (AWS) discussed the company’s strong partnership with Anthropic and highlighted the rapid progress made with their Trainium chips, emphasizing cost, performance, and efficiency improvements over previous generations and competitors like NVIDIA GPUs and Google’s TPUs. AWS controls the entire technology stack—from silicon development to data centers—allowing them to deploy large clusters quickly and deliver impressive performance. This integrated approach enables AWS to bring new capabilities to market rapidly and meet the growing demand for compute power.
AWS is committed to an annual cadence of new training chip generations, driven by the insatiable hunger for more compute power in AI applications. The company focuses on iterating technology quickly to increase compute density within existing power footprints, enabling customers to build innovative applications efficiently. AWS also supports NVIDIA GPUs, recognizing that different use cases may benefit from different hardware, and maintains a long-standing partnership with NVIDIA to provide customers with stable, high-performance clusters at scale.
Capacity expansion is a major focus for AWS, with plans to double their compute capacity to around eight gigawatts by the end of 2027. The company added 3.8 gigawatts in the past year alone and will continue to scale based on customer demand. While AWS listens closely to customer needs to guide their investments, they acknowledge that supply constraints exist across the industry due to the rapid growth of AI, affecting chips, power, data centers, and other components. Despite these challenges, AWS has strong partnerships and supply chain management to secure the necessary resources.
The CEO highlighted that AWS is already seeing financial benefits from their Trainium technology, with rapid growth in services like Bedrock, where over half of all inference tokens are processed on Trainium servers. The partnership with Anthropic is particularly strong, with Anthropic relying heavily on AWS for their compute needs through Project Rainier. Although Anthropic also uses other cloud providers to meet their massive compute demands, AWS remains their primary cloud partner, collaborating closely to support their current and future AI models.
Looking ahead, AWS sees generative AI and agent technologies as key drivers of enterprise value, though adoption will take time as organizations adjust workflows and processes. The CEO noted that while not all customers are ready for this shift, there is strong excitement and recognition of the efficiency gains agents can provide. AWS believes they are well-positioned in the market, with many customers choosing to run production workloads on their platform after evaluating multiple options, signaling confidence in AWS’s infrastructure and services for AI at scale.