Hudson River Trading's CoreWeave deal puts a quant fund at the front of the Vera Rubin queue

The undefined number is the money: the parties called the agreement multi-year and multi-billion-dollar without publishing a figure ^[1]. The defined number is the calendar. HRT first became a CoreWeave customer in March 2026 ^[2], and crowdfundinsider.com reported the expansion in August 2026 ^[3], which puts roughly five months between opening the account and committing for years at that scale ^[4].

What the speed buys is position. HRT is now among the first CoreWeave customers with broad access to NVIDIA's Vera Rubin chips ^[5], and the clusters it will use pair Vera Rubin NVL72 with HGX B200 systems ^[6]. Early access to a new accelerator generation is a rationing outcome rather than a product tier: at launch there is a finite pool, and the firms holding multi-year contracts are the ones standing in front of it. A trading firm holding that position is a trading firm that outbid, or out-committed, somebody who wanted the same racks.

The networking detail says what kind of buyer HRT is. The deal includes Direct Connect built on NVIDIA Spectrum-X Ethernet, providing dedicated high-bandwidth links between HRT's on-premises systems and CoreWeave's cloud, with lower latency and more predictable transfers as the stated aim ^[7]. That is not a migration. HRT keeps its own machines and bolts rented training capacity onto them. It is buying throughput, not a platform, and the research loop stays anchored to hardware it controls.

The dependency runs the other way too. Kevin Lee, HRT's head of research and development, said the foundation underneath AI and machine learning research matters as much as the models, and that CoreWeave was chosen partly for its ability to scale alongside HRT's goals ^[8]. Translated into procurement terms, a firm whose stated need is faster iteration cycles and larger data volumes ^[13] has tied the pace of its own research to a vendor's delivery schedule.

Then there is the company HRT is keeping. Jane Street has deepened its own CoreWeave relationship through capacity commitments and investments ^[9], and CoreWeave has been adding quantitative firms while pushing past its original technology and AI lab customer base ^[10]. Jon Jones, CoreWeave's chief revenue officer, framed financial services as a demanding proving ground for production AI ^[11]. From a model lab's side of the table, that reframes the negotiation. The marginal GPU-hour is no longer priced only against other labs' training runs but against a firm that treats compute as a cost of extracting trading revenue, and that measures the return on it weekly rather than at the next model release.