Amazon’s investment in custom silicon is quickly turning into one of the company’s most important growth engines. In the wake of its latest earnings report, the company revealed that its custom chip portfolio, including Trainium AI accelerators, Graviton CPUs, and Nitro networking processors, has now exceeded a $20 billion annual revenue run rate.
At this point, one thing is obvious: Amazon Web Services (AWS) is evolving into a fully integrated AI infrastructure provider from a cloud platform it was years ago.
The milestone underscores Amazon’s long-term strategy. The tech giant, launched three decades ago as an online bookstore, has already outgrown its online retailer status and is striving to control more of the AI technology stack. Instead of relying solely on third-party hardware suppliers, AWS has spent years building proprietary chips optimized for AI training, inference, cloud computing, and networking. CEO Andy Jassy said the business is growing at triple-digit rates and argued that if the chip division were treated as an independent semiconductor company selling externally, its annualized revenue would approach $50 billion.
Demand appears far from slowing. Amazon disclosed that Trainium has accumulated more than $225 billion in customer commitments, while Trainium2 is largely sold out, Trainium3 is nearly fully subscribed, and much of Trainium4 capacity has already been reserved ahead of launch. Major AI developers including Anthropic and OpenAI have committed to large-scale deployments on AWS infrastructure, reflecting growing confidence in Amazon’s alternative to NVIDIA-powered systems.
The custom silicon momentum also aligns with AWS’s accelerating financial performance. Amazon reported AWS revenue growth of 37% year over year in the latest quarter, its strongest expansion in more than a four-year period. With these numbers in mind, the company may be bound to increase planned capital expenditures for 2026 to approximately $220 billion, primarily to expand AI infrastructure and data center capacity.
AI-powered payment orchestration, fraud detection, digital identity verification, agentic commerce, and real-time transaction processing increasingly rely on specialized AI infrastructure capable of handling massive inference workloads at lower cost. By designing its own chips, Amazon can reduce operating costs while offering enterprise customers greater price-performance for AI applications.
