Overview
Lambda announced that it has entered into a cloud services arrangement with high-frequency trading firm Hudson River Trading to provide access to more than 1,000 of Nvidia’s most recent Blackwell systems. The company said the systems to be rented were equipment Lambda had previously bought and deployed in a data center, rather than being new chip purchases specifically for this contract.
Background on the companies involved
Lambda, which counts Nvidia as a backer, completed a $1.5 billion fundraising round last year following a separate agreement to supply Microsoft with access to Nvidia chips. Hudson River Trading is a significant customer of Google Cloud and was reported last month to have recorded $12.3 billion in trading revenue in the prior year.
Deal specifics and statements
The arrangement will see Hudson River Trading rent more than 1,000 Blackwell systems from Lambda. Neither Lambda nor Hudson River Trading disclosed the financial specifics of the contract. Stephen Balaban, co-founder and chief technology officer at Lambda, said the systems involved in this deal are units Lambda already owned and had installed in a data center.
"It’s the only product that’s available in every one of the major cloud providers," Balaban said, referring to the ubiquity of Nvidia’s AI chips as a selling point for large customers such as Hudson River Trading.
Cloud and chip deployment notes
Hudson River Trading has an established relationship with Google Cloud but has only publicly announced use of Nvidia’s chips within Google’s cloud rather than Google’s own custom AI processors. Lambda emphasized the cross-cloud availability of Nvidia’s AI hardware as an important factor in attracting large-scale customers.
Implications for infrastructure
From an infrastructure and data center perspective, the deal highlights a commercial route in which pre-provisioned, on-premises cloud-like hardware is leased to compute-intensive customers. The use of systems already purchased and installed by Lambda underscores a model that leverages existing data center capacity and inventory to serve enterprise customers seeking large blocks of GPU compute.