- Published on
Amazon just tripled its order of Nvidia chips over ‘surging demand’
- Authors
- Name
- DeepHorus
Amazon has significantly expanded its partnership with Nvidia, tripling its order of advanced AI chips to meet what the companies describe as "surging demand." The tech giants announced on Wednesday that Amazon will add another 2 million Nvidia GPU chips to its data centers, following a deal for over 1 million chips made just five months prior. The new order includes cutting-edge Blackwell Ultra, Rubin, and Rubin Ultra GPUs, scheduled for deployment in Amazon Web Services' data centers during 2027 and 2028.
The announcement, made during Nvidia's quarterly earnings call, underscores the intense global competition for AI computing power. Nvidia stated that demand for its initial 1 million GPU commitment had "exceeded those expectations," prompting the accelerated partnership. The deal, while financial terms were not disclosed, is estimated to be worth tens of billions of dollars based on typical GPU unit costs.
This expanded agreement is notable for its scope beyond a simple chip purchase. It integrates Nvidia's broader technology stack, including networking hardware, open models, CPUs, and robotics platforms, across AWS infrastructure. The collaboration is driven by robust demand from startups, enterprises, AI labs, and governments seeking to build and run advanced AI models.
The move comes despite Amazon's own efforts to develop competing AI chips, such as its Trainium deep-learning processors and Arm-based Graviton CPUs. Amazon's custom silicon business has reportedly reached a $25 billion annualized revenue rate, fueled by commitments from major clients like Anthropic and OpenAI. However, Nvidia remains the dominant force in the AI chip market.
As part of the deal, Nvidia will also deploy an unspecified number of its new Vera CPUs alongside the GPUs. Nvidia CEO Jensen Huang has projected the Vera CPU represents a "brand-new $200 billion" total addressable market. The partnership further extends to integrating Nvidia's full physical AI software stack to power Amazon's warehouse and logistics robot fleets.
The collaboration also boosts Amazon's enterprise AI offerings. AWS will host Nvidia's Nemotron family of open models on its Bedrock and SageMaker platforms. The announcement coincided with Nvidia reporting stellar quarterly results, including $96.2 billion in total sales and $89 billion in data center revenue, highlighting the unstoppable momentum of the AI infrastructure boom.
