Nvidia has entered into a major multi-year agreement with Amazon Web Services (AWS) to supply approximately 1 million GPUs, along with advanced networking and inference chips, through the end of 2027. The deal, confirmed by an Nvidia executive to Reuters, represents a significant expansion of AWS's AI infrastructure buildout across its global cloud regions.
The collaboration extends beyond simple hardware supply. AWS and Nvidia will work together on integrating networking hardware like Connect X and Spectrum X, as well as Nvidia's NVLink Fusion platform, into AWS data centers. The goal is to build systems "capable of reasoning, planning, and acting autonomously across complex workflows," with a focus on agentic AI systems. The rollout is set to commence this year and continue through 2027.
Industry observers note the deal signals a shift in AI demand, with inference—the process of running trained AI models in real-time—now accounting for roughly two-thirds of AI compute, up from about a third in 2023. The market for inference-focused chips is projected to exceed $50 billion by 2026, according to Deloitte estimates cited by analysts.
The partnership highlights a nuanced dynamic of "cooperation and competition." While AWS is a major buyer of Nvidia chips, it continues to develop its own AI chips, such as the Trainium series, which it claims offer 30–40% better price-performance than GPU-based instances. This hybrid, multi-chip model allows AWS customers more choice and is seen as a key differentiator.
Financially, the terms of the deal were not disclosed, but it follows Nvidia CEO Jensen Huang's previous comments about a potential $1 trillion sales opportunity for its Rubin and Blackwell chip families. The news prompted a slight dip in Nvidia's (NVDA) stock, which analysts attribute to investor caution over supply chain impacts and margins, despite the deal underscoring Nvidia's central role in cloud AI infrastructure.
The agreement comes amid broader geopolitical tensions, as U.S. prosecutors pursue a case alleging Nvidia chips were smuggled to China. Since 2022, U.S. controls have limited the export of Nvidia's most advanced chips to China as part of a strategy to curb China's progress in advanced computing and AI.
