Amazon Web Services is preparing to directly challenge Nvidia’s dominance in the artificial intelligence chip market by exploring sales of its custom Trainium processors to external data center operators. According to AWS AI chief Peter DeSantis, the division is currently in preliminary discussions with potential third-party buyers, a strategic pivot initially outlined in CEO Andy Jassy’s April shareholder letter. Jassy noted that strong internal demand suggests a standalone chip business could generate approximately $50 billion in annual revenue, positioning Amazon as a formidable competitor in the global semiconductor space. This proposed shift marks a significant departure from AWS’s historical practice of reserving its semiconductor designs exclusively for internal use and cloud customers. Amazon has traditionally avoided direct chip sales because Trainium processors drive a broader ecosystem revenue model, encompassing cloud compute, storage, networking, and security services. Furthermore, AWS has struggled to satisfy surging internal demand, with current Trainium capacity fully allocated and next-generation Trainium4 production delayed beyond next year. Expanding external sales would require scaling manufacturing through third-party foundries such as TSMC, a process complicated by the facility’s current capacity constraints and competing priorities from Nvidia and Apple. Despite these supply chain and capacity hurdles, AWS leadership maintains that distributing its silicon to third parties is a viable long-term strategy. If executed, the initiative would establish Amazon as a direct competitor in the broader AI hardware sector, challenging Nvidia’s recent expansion into AI-optimized processors. While a $50 billion chip division would not immediately displace Nvidia’s $326 billion revenue run rate, it signals Amazon’s intent to monetize its semiconductor development beyond cloud infrastructure, fundamentally altering competitive dynamics in the enterprise AI market.