1 hour ago
2
JPMorgan Chase, the largest bank in the US by assets, has selected SambaNova Systems as its inference partner for on-premises AI workloads. The deployment includes SambaNova’s SN40 and SN50 Reconfigurable Dataflow Units, purpose-built AI accelerators designed to run models without shipping data to someone else’s servers.
The deal and what’s behind it
The partnership arrives at a moment when SambaNova is riding serious momentum. The company completed the first close of a $1 billion Series F funding round on July 8, 2026, pushing its post-money valuation to $11 billion. General Atlantic led the round, with Intel Capital also participating.
For a company founded in 2017 with the explicit goal of challenging Nvidia’s dominance in AI hardware, landing JPMorgan as a flagship client is the kind of validation that money alone can’t buy. SambaNova’s Reconfigurable Dataflow Units take a fundamentally different architectural approach, one that the company argues is better suited for enterprise inference workloads where consistency and security matter more than raw training throughput.
Rodrigo Liang, SambaNova’s CEO, highlighted JPMorgan’s commitment as a signal of the technology’s readiness for the most demanding enterprise environments. On the bank’s side, CIO Darrin Alves emphasized the need for high performance and reliability in AI infrastructure.
Why banks are moving away from the cloud
Financial institutions have spent the last decade migrating workloads to public cloud providers like AWS, Azure, and Google Cloud. For AI inference on sensitive financial data, the calculus is changing. Regulations around data privacy and security continue to tighten globally, pushing financial institutions to maintain tighter control over where their data lives and who can access it. Running AI models on third-party infrastructure introduces layers of risk that compliance teams increasingly don’t want to manage.
On-premises AI hardware solves this by keeping everything in-house. As AI becomes central to competitive advantage in finance, that tradeoff is starting to look more favorable, especially when vendors like SambaNova offer systems specifically architected for enterprise deployment rather than repurposed data center GPUs.
What this means for investors and the AI hardware landscape
SambaNova’s $1 billion fundraise signals that investors see room in the AI hardware market for more than one winner. Nvidia’s dominance remains overwhelming, particularly in training large models. But inference — the process of running trained models on new data in real time — is a workload profile where specialized chips can compete on efficiency, latency, and total cost of ownership.
The risk for SambaNova is execution. An $11 billion valuation prices in significant future growth, and delivering enterprise-grade reliability in a demanding banking environment is considerably harder than winning a benchmark competition. Regulated industries from healthcare to defense face identical data sovereignty pressures, and the JPMorgan deployment will serve as a closely watched reference case.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) · 