BNY has tapped Nvidia to build an AI factory to increase its computational power and deploy gen AI more efficiently within its operations. 

Nvidia AI factories are a set of graphic processing units (GPUs) embedded with Nvidia software that can be customized for each organization to run gen AI models efficiently, Malcolm deMayo, global vice president of financial services at Nvidia, told Bank Automation News.

The chipmaker built BNY’s AI factory in 22 weeks, and it went live in February, he said, without disclosing the cost of the factory. 

DeMayo declined to say how many GPUs are running at BNY’s AI factory, but the bank is moving toward having a “superpod” in the near future. 

A superpod is a collection of GPUs arranged to increase computational speed, deMayo explained. 

One AI factory node, or a set of four GPUs, can process the workload of four research assistants and can read 1,000 books a second each, deMayo told BAN, adding that a superpod has 1,024 nodes to it, which can increase the computing power exponentially. Banks typically start with one or two nodes and add more as they progress on their AI journey. 

BNY is using Nvidia’s AI factory to run gen AI models for to improve client experience and increase operational efficiencies, Sarthak Pattanaik, global head of AI and engineering at the $30 billion bank, said during a panel at Money20/20 on Oct. 27. 

New York-based BNY has found gen AI use cases in three areas: 

• Treasury management: The bank is using gen AI to “predict [treasury] transaction settlement fails hours before the market closes,” Pattanaik said, adding that the tool helps clients to “proactively take action to reduce transaction fails and reduces the charges.” 

• Hyper-personalization: BNY is deploying gen AI to suggest specific products to clients by looking at their current and potential needs, Pattanaik said. 

• Customer service: The bank is using gen AI to help its customer representatives service clients better, Pattanaik said, adding that the bank wants its operations to be “cheaper, faster and without any friction.” 

BNY has provided gen AI tools to 25% of its workforce for operational uses as well as creating customized agents to help them in workflows, Pattanaik said. 

While BNY has implemented gen AI tools in its operations, it’s too early to quantify efficiency gains, Pattanaik said during the panel. 

The GPUs that are running gen AI models have open architecture, meaning they can run OpenAI, Google Gemini and other AI models according to the user’s wish. That makes problem-solving even more efficient, Pattanaik said. 

The bank has also encouraged its workforce to come up with gen AI use cases as it expands the implementation of the tech, Pattanaik said. 

Register here for early-bird pricing for Bank Automation Summit U.S. 2025, taking place March 3-4 in Nashville, Tenn. View the full event agenda here.