CoreWeave signs multibillion-dollar Hudson River Trading deal, underscoring Wall Street’s scramble for Nvidia’s latest AI chips

CoreWeave Inc., a provider of cloud infrastructure built around graphics processing units (GPUs), has landed a multiyear agreement with Hudson River Trading, a Wall Street trading firm, in what is described as a multibillion-dollar deal. The arrangement centers on access to Nvidia Corp. AI chips, a commodity that has become increasingly strategic for firms that need large-scale computing to run models, do research, and power automation.

The report frames the agreement as part of a broader competition among market participants for the newest Nvidia chip offerings. In that race, capacity is not simply a procurement issue but also a timing issue, because the pace of AI workloads and model training and inference can depend on what hardware is available and when. CoreWeave’s role, as described in the reporting, is to channel that hardware into an AI-ready service environment that customers can use.

For Hudson River Trading, the practical upshot is access to accelerated compute through a third-party infrastructure provider rather than relying only on in-house hardware purchases and deployments. That can shorten timelines, shift some operational burden to a specialized provider, and help the firm align spending with evolving demand for AI and machine learning capabilities.

CoreWeave’s deal also adds to the growing sense that Nvidia’s ecosystem is being pulled into a procurement-driven contest between customers with different procurement strategies. Some buyers take a direct route, acquiring their own servers and managing deployments. Others, including data-center and service providers like CoreWeave, position themselves as intermediaries, using buying power and supply relationships to deliver usable capacity to customers that need compute quickly.

Nvidia remains the core supplier in this setup. Nvidia’s GPUs and related platforms are widely used for AI training and inference, and the company’s “newest” chip generations typically bring improvements in performance per unit of power and speed of data handling, factors that can materially affect the cost and time required to run large workloads. Even when software and infrastructure can be portable across generations, the incremental benefits of each new hardware wave create pressure for customers to upgrade and gain access early.

The reporting does not provide, in the available text, additional specifics such as the exact chip model(s), the contracted quantities, pricing terms, or the precise start and end dates of the multiyear period. It also does not disclose whether the agreement is structured as a fixed commitment with guaranteed capacity, a demand-based arrangement, or a hybrid. Those distinctions can matter for how much certainty the customer is buying, and for how revenue is recognized by the service provider.

Still, the size and duration described in the report underline how AI infrastructure spending is increasingly tied to long-lead supply. When supply is constrained, access agreements can become a way to manage risk and to keep compute pipelines full, especially for firms that require consistent throughput for computationally intensive tasks.

What to watch next is whether this type of arrangement becomes more common across Wall Street and whether other customers follow similar routes to secure Nvidia-based capacity. For investors and industry observers, the key questions are whether CoreWeave’s customer wins translate into sustained growth, how quickly customers can scale usage from contracted capacity, and what this means for Nvidia’s downstream partners as the AI chip cycle continues.