Amazon is exploring a major expansion of its AI chip business. The company is considering selling its in-house AI chips to other businesses, rather than using them primarily within its AWS cloud platform, reports Bloomberg. Amazon CEO Andy Jassy recently indicated that demand for Amazonâs chips has become so strong that the company could eventually sell complete AI hardware systems and server racks to third-party customers. If it moves ahead, Amazon would become a direct competitor to Nvidia in the fast-growing AI hardware market.
According to Jassy, Amazonâs custom silicon division â which includes Trainium AI chips, Graviton server processors, and Nitro networking hardware â has grown to more than a $20 billion annual revenue run rate and is expanding at triple-digit growth rates. He also said that if Amazon sold these chips externally in the same way Nvidia sells its products, the business could eventually reach a $50 billion annual revenue run rate. At the same time, AWSâs overall AI business has already crossed a $15 billion annual revenue pace, clearly showing how rapidly demand for AI infrastructure is growing.
At the center of Amazonâs strategy is its Trainium family of AI accelerators. Trainium chips are designed to train large AI models, while Inferentia chips are built for inference. The firm believes custom silicon can significantly lower computing costs. Jassy has argued that widespread use of Trainium could save AWS tens of billions of dollars in annual capital spending over time. Amazon has already integrated these chips into many of its AI services, including Bedrock, its generative AI platform that allows businesses to access large language models through AWS.
Demand for the chips appears to be extremely strong. Amazon has revealed that available capacity for Trainium2 has effectively been sold out, while much of the upcoming Trainium3 production has already been reserved by customers. Interest in future generations, including Trainium4, is also growing well before those products are fully available. The company reported around 40% quarter-over-quarter growth in its chip business during the first quarter. Jassy also disclosed that some customers wanted to secure nearly all future Graviton processor capacity, but Amazon rejected those requests to ensure supply remained available for a wider customer base.
The push comes at a time when Nvidia dominates the AI hardware market as no company has before. Nvidiaâs GPUs are used by nearly every major AI developer, including cloud providers, startups, and large enterprises. This dominance helped Nvidia become one of the most valuable companies in the world. However, the largest technology companies are increasingly trying to reduce their reliance on Nvidia. For example, Google has developed its own TPU chips, Microsoft is investing in Maia AI processors, and Meta is building custom AI hardware as well.
Rather than trying to beat Nvidia purely on raw performance, Amazon is focusing heavily on cost efficiency. Amazon claims its Trainium chips can offer a better price-to-performance ratio for many AI tasks, helping businesses lower training and inference expenses. At the same time, the company expects to spend around $200 billion on capital expenditures in 2026, with a significant portion going toward AI infrastructure and data centers.
The Tech Portal is published by Blue Box Media Private Limited. Our investors have no influence over our reporting. Read our full Ownership and Funding Disclosure â
Ashutosh is a Senior Writer at The Tech Portal, largely reporting on new tech, and intersection of technology and business. Ashutoshâs career spans across nearly a decade of technology writing across multiple platforms and languages.
