$20 billion in annual revenue now comes from Amazon’s custom silicon unit, which designs Graviton CPUs and Trainium AI accelerators for AWS rather than selling chips broadly.
More than 100% yearly growth reflects Amazon’s push to replace outside processors with in-house designs that cut vendor dependence, lower cloud costs and tune performance for AI workloads.
$150 billion in AWS annualized first-quarter 2026 revenue gives the chip effort a massive captive customer base, with efficiency gains potentially translating into billions in savings and added capacity.
$50 billion is the implied run rate CEO Andy Jassy sees if Amazon sold the chips directly, putting the internal business on par with Broadcom’s AI revenue and above AMD’s and Intel’s data-center AI segments.
Nvidia still dominates AI accelerators with over $193 billion in data-center revenue, but wider external sales of Trainium could turn Amazon’s chip arm into a new growth engine beyond cloud computing.
As AI giants commit billions to Amazon's chips, are they trading Nvidia's dominance for a more powerful gatekeeper?
Amazon's AI chip empire relies on TSMC. What is its plan B if the Taiwanese supply chain is ever disrupted?
Amazon’s $200 Billion Bet: How Custom Silicon Is Transforming AWS, AI, and the Semiconductor Industry in 2026
Overview
Amazon Web Services (AWS) has quickly become a leader in semiconductor manufacturing by developing custom chips like Trainium and Graviton, which are designed for AI and CPU workloads. This strong focus on internal chip innovation has driven Amazon’s financial success, helping the company achieve its best quarter on record in Q1 2026 and boosting profit margins to around 38%. To support this growth and capitalize on the AI boom, Amazon announced a massive $200 billion capital expenditure plan for 2026, highlighting its commitment to expanding AI infrastructure and maintaining its competitive edge in the cloud industry.