Leading European Financial Services Organizations Embrace Open Models to Build Proprietary Intelligence

22 October 2026 3 min read هدهد تك | Hudhud Tech

At GTC Berlin, banks, payment companies, and financial data providers are demonstrating the capabilities of production-scale intelligence, built on NVIDIA accelerated computing and powered by open weight models running on their own infrastructure.

Revolut is in the process of overhauling its technology stack by adopting a single foundation model known as PRAGMA, which is also powered by NVIDIA accelerated computing. The company is utilizing the NVIDIA cuDF toolkit and Dynamo-Triton software on NVIDIA Blackwell Ultra GPUs to scale PRAGMA in production, initially serving 80 million customers and designed to scale threefold.

Pavel Nesterov, head of AI at Revolut, stated, “Revolut processes billions of transactions — the opportunity was always there, but the question was whether foundation models could operate at that scale and across domains. PRAGMA answers that. Built in collaboration with NVIDIA, it’s the clearest proof yet that the gap between a research breakthrough and a production banking system can be closed.”

Training on Accelerated Infrastructure

Starling Bank has become the first institution to train an agentic model on U.K. sovereign cloud infrastructure, developing “Starling Intelligence,” which serves as the intelligence layer for its AI capabilities, using its own transaction data on Civo’s U.K. cloud.

Lloyds Banking Group has created ATUM, a transaction foundation model that transforms billions of banking transactions into reusable intelligence, enhanced with NVIDIA RTX PRO 6000 Blackwell GPUs to improve efficiency across various bank-wide applications.

SBS has launched a banking sovereign intelligence platform that allows banks to run generative and agentic models while keeping their data secure within their own systems.

Open Models at a Fraction of the Cost

Open weight models enable European financial institutions to develop proprietary intelligence based on their own data, while controlling costs. NVIDIA’s accelerated computing platform provides the necessary infrastructure.

Neobank bunq has integrated its agentic assistant Finn with open weight models, achieving performance comparable to leading closed models at 60% of the cost per response, thanks to NVIDIA accelerated computing deployed on Nebius.

NatWest is developing technology to recognize a wide range of British accents, dialects, and financial terminology, utilizing NVIDIA’s speech models, including Nemotron 3.5 ASR.

Modulr is creating specialized models for various aspects of payment operations, based on the NVIDIA Nemotron-3-Ultra-550B open model.

Santander is among the early adopters of the new NVIDIA Kumo family of open weight foundation models for structured financial data, leveraging them to create specialized intelligence for core banking tasks.

Accelerated Computing Transforms Financial Research

LSEG is collaborating with NVIDIA to enhance the analysis of large volumes of historical tick data, achieving speeds up to ten times faster with the Polars GPU engine, accelerated by NVIDIA cuDF on NVIDIA RTX PRO 6000 Blackwell GPUs.

RavenPack has developed a dataset generation pipeline on Bigdata.com that converts plain-language research hypotheses into labeled, backtest-ready datasets, utilizing a fine-tuned NVIDIA Nemotron model. Investors can use these datasets to identify, analyze, and visualize market trends.

Speeding Data Pipelines to Improve Fraud Detection

NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs and cuDF facilitate accelerated data processing across multi-terabyte-scale, regulated sources, significantly reducing the time from data acquisition to decision-making for fraud detection and other applications.

HSBC is now able to provide near-real-time client insights that support customer processes and on-demand decision-making while minimizing risk and fraud, powered by five times faster data processing on Starburst Enterprise’s GPU Engine.

Stripe, a global provider of programmable financial services, has reported a 42x increase in data processing speed and a 92% reduction in costs for its Stripe Radar fraud-detection platform, utilizing NVIDIA GPU-accelerated Ray Data in offline testing. This advancement allows Stripe Radar’s machine learning pipelines to continually optimize model training and enhance fraud detection capabilities.