Hudson River Trading Makes a Multibillion-Dollar Bet on CoreWeave Compute

Hudson River Trading is committing to CoreWeave’s AI cloud in a multiyear deal worth multiple billions of dollars, raising the computing capacity behind its trading research and model development. The pact also places HRT among CoreWeave’s earliest customers with wide-scale access to Nvidia’s Vera Rubin AI chips.

The agreement, announced August 20, expands a relationship that began when HRT became a CoreWeave customer in March. CoreWeave Chief Revenue Officer Jon Jones called the new commitment a material expansion but did not disclose a precise contract value or other financial terms.

From client to capacity commitment

HRT will use the platform for AI research, model training and AI-driven trading research. Its head of research and development, Kevin Lee, said added compute lets researchers test more ideas; it can also help firms handle more trades across longer time horizons.

That is the operational logic behind the purchase: more processing power expands the set of research experiments a quantitative trading firm can run. It does not establish which models or research approaches HRT will deploy, nor whether the additional capacity will produce better trading results.

The system HRT is buying into

The deployment is more than a chip allocation. CoreWeave says it will include clusters of Nvidia Vera Rubin NVL72 and HGX B200 GPU systems, plus a network connection intended to move data between HRT’s own environment and the cloud platform.

The proposed deployment has three layers

• Compute: clusters of Nvidia Vera Rubin NVL72 and HGX B200 GPU systems for large-scale AI workloads.
• Networking: Spectrum-X Ethernet and Direct Connect connectivity between HRT’s on-premises environment and CoreWeave’s platform.
• Workloads: AI research, model training and AI-driven trading research running on CoreWeave’s platform.

The networking component is designed to support high-throughput data movement between the two environments. CoreWeave says the arrangement is intended to minimize latency and make that movement reliable and predictable, a company objective rather than a disclosed performance result.

A financial-services test for CoreWeave

For CoreWeave, HRT is part of a deliberate move toward financial customers after the company relied on Microsoft for more than 70% of sales. CoreWeave says the agreement expands its presence among financial-services institutions and customers operating AI workloads at production scale.

There is already evidence that the sector can support unusually large capacity commitments. Jane Street, HRT’s largest competitor, agreed to spend $6 billion on CoreWeave data-center capacity and took a $1 billion equity stake in the provider.

The next test is delivery. HRT selected CoreWeave in part because Lee said the provider had delivered infrastructure on time or ahead of schedule and run Nvidia chips efficiently. The new deal makes that execution record central to a much larger commitment, but its buildout schedule and the volume of capacity HRT will receive remain undisclosed.