JPMorgan Chase selects SambaNova for in-house AI inference

JPMorgan Chase has chosen artificial intelligence chip company SambaNova as an infrastructure partner for on-premises AI inference, adding specialist hardware to run production AI workloads within the bank’s own systems. The deployment covers SambaNova’s SN40 and SN50 systems and is intended to give the bank tighter control over data, performance and auditability as its use of AI expands.

The agreement places SambaNova inside one of the world’s largest financial technology environments, where banks face strict requirements over data handling, operational resilience and model governance. On-premises inference means AI models process requests within infrastructure controlled by the institution, rather than relying only on external cloud environments. For a bank, that can support internal oversight of sensitive data and create a clearer record of how AI systems are used.

Darrin Alves, chief information officer for infrastructure platforms at JPMorgan Chase, said the bank’s AI infrastructure must satisfy demanding standards for “performance, control and reliability.” Alves said JPMorgan Chase was deploying SambaNova’s reconfigurable dataflow unit, or RDU, architecture and would test its speed and security for in-house inference across enterprise AI workloads.

Inference is the stage at which a trained AI model produces outputs from new prompts or data. As companies move from experimentation into production, the computing requirements can shift from training large models to running them repeatedly, securely and at scale. SambaNova markets its full-stack platform for enterprise inference, cloud-style AI providers, AI laboratories, service providers and sovereign AI projects.

JPMorgan Chase has been a prominent buyer and builder of AI systems. The bank has used models from Anthropic and OpenAI and has doubled its AI use cases over the past year, with work focused on customer service and technology staff, according to Finextra. In October, chief executive Jamie Dimon said the bank spends about $2 billion a year on AI development, equal to roughly one-tenth of its technology budget, and that the technology is saving the bank about the same amount.

The partnership was disclosed as SambaNova raised $1 billion in strategic financing through a Series F round that valued the company at $11 billion. General Atlantic led the round, with other blue-chip investors also participating, according to Finextra.

SambaNova is among a group of AI infrastructure start-ups seeking to compete with Nvidia, whose graphics processors have dominated much of the AI compute market. The company’s proposition centres on providing both hardware and software for premium inference, rather than selling chips alone.

Martín Escobari, co-president and head of global growth equity at General Atlantic, said SambaNova’s platform was built for a market in which inference had become central to enterprise and industrial AI adoption. He said the company was showing commercial momentum as demand for inference capacity rises faster than supply.

This story draws on original reporting from Finextra Research.