Amazon expands Nvidia GPU purchases, reported to add 2 million chips over two years

Amazon is adding a large new wave of Nvidia graphics processing units (GPUs), according to a market report published by Yahoo Finance. The outlet said Amazon has “tripled” its order of Nvidia chips, with an additional 2 million GPUs expected to be deployed in its data centers over the next two years.

The report ties the expansion to “surging demand,” a reference to the fast-growing customer appetite for cloud-based artificial intelligence workloads. GPUs are the specialized chips most commonly used to train and run AI models because they can process large volumes of parallel computations more efficiently than general-purpose CPUs.

While the piece frames the move as an incremental purchasing step, it also suggests the relationship between Amazon and Nvidia is deepening. Beyond simply buying more chips, the report states that the extended partnership stretches further than procurement, implying additional collaboration between the two companies. The specific forms of that collaboration were not detailed in the available account.

Nvidia, whose investor appeal is closely linked to the strength of AI-capable data-center demand, has been supplying GPUs at a pace that reflects industry-wide competition for the hardware needed to build AI factories. For Amazon, expanding GPU capacity is a practical response to higher utilization rates and to the need to keep pace with new AI services and customer deployments.

Even so, the report does not provide operational specifics such as which particular Nvidia GPU models Amazon will receive, how the chips will be allocated across Amazon’s cloud regions, or whether the arrangement includes any performance guarantees, exclusivity, or volume-based pricing. Those details matter because different GPU generations and configurations can materially affect total compute capability and cost.

The transaction also underscores how cloud providers are responding to demand not only for AI applications, but also for the supporting infrastructure required to deliver them. GPUs must be paired with data-center networking, memory, storage, and orchestration software to deliver usable AI performance. In that sense, additional chip orders often announcement broader capacity build-outs across the stack.

From a sector perspective, the report fits a broader pattern in AI hardware procurement where hyperscalers and other large buyers lock in supply to reduce bottlenecks. When companies “tripling” orders are reported, it typically reflects a shift from incremental scaling to more aggressive capacity planning, especially when customers are competing to secure compute for model training and inference.

What remains unclear is the level of visibility Amazon and Nvidia have provided to investors and customers about the timing, exact delivery schedule, and the end-to-end impact on Amazon’s cost structure and Nvidia’s near-term shipment trajectory. The report also does not quantify how much of the incremental 2 million GPUs are intended for immediate customer workloads versus longer-term capacity expansion. Observers will likely watch for additional disclosures in company filings, earnings materials, and any public updates from Nvidia’s product and supply communications.