Leading European Financial Services Organizations Embrace Open Models to Build Proprietary Intelligence
Original reporting by NVIDIA Blog
Financial institutions are rapidly deploying production-scale intelligence built on NVIDIA accelerated computing, powered by open-weight models, and running on their own infrastructure. This significant shift, highlighted at GTC Berlin, demonstrates how cutting-edge AI is moving beyond research labs into the core operations of banks, payment companies, and financial data providers globally. These deployments are not just conceptual; they are driving tangible results, from enhancing customer service to detecting fraud with unprecedented speed and accuracy.
Leading the charge, Revolut is rebuilding its entire technology stack on a single foundation model, PRAGMA. Engineered with NVIDIA accelerated computing, PRAGMA is designed to serve an initial 80 million customers and scale significantly, proving that sophisticated AI breakthroughs can be integrated directly into production banking systems.
The Open Model Advantage
A key enabler of this transformation is the strategic adoption of open-weight models. These models empower European financial institutions with proprietary intelligence built on their own data, offering significant cost controls and flexibility. Neobank bunq, for instance, powers its agentic assistant, Finn, with open-weight models on NVIDIA accelerated computing, achieving quality comparable to leading closed models at 60% of the cost per answer. Across the financial sector, from speeding up multi-terabyte data pipelines for fraud detection at HSBC and Stripe to accelerating complex financial research at LSEG and enabling specialized models at Santander and Modulr, NVIDIA's accelerated computing platform is proving to be the essential engine behind this widespread AI revolution.
The demonstrations at GTC Berlin underscore a significant turning point where financial institutions are not just exploring AI, but actively deploying production-scale intelligence. The rapid shift to NVIDIA-accelerated computing, leveraging open-weight models on proprietary infrastructure, is enabling unprecedented operational efficiency and innovation. From Revolut's foundational PRAGMA model scaling to hundreds of millions of customers, to Starling Bank's sovereign cloud agentic models, and Lloyds' ATUM transforming transaction data, the industry is rapidly moving from theoretical potential to tangible, transformative applications. This concerted effort ensures not only scalability and cost-efficiency but also critical data security and sovereignty.
Redefining Financial Services
This wave of industrial-scale AI deployment marks a profound redefinition of financial services. The ability to process billions of transactions, detect fraud in near real-time, generate personalized insights, and accelerate complex financial research at a fraction of previous costs fundamentally alters the competitive landscape. For customers, this translates into more secure, efficient, and tailored experiences, while institutions gain powerful tools for risk management, operational optimization, and new product development. Looking ahead, this foundational shift will likely foster an era of continuous innovation, where AI-driven capabilities become standard, not just differentiators. It paves the way for increasingly autonomous financial systems, highly predictive analytics, and hyper-personalized client engagement, pushing the boundaries of what is possible in a secure and compliant manner. The strategic embrace of open models and powerful accelerated computing signifies a future where financial intelligence is not just faster and cheaper, but also more accessible and deeply integrated into the very fabric of global finance.
Frequently asked questions
- How are financial institutions leveraging NVIDIA accelerated computing and AI foundation models for operations?
- Financial institutions are utilizing NVIDIA accelerated computing to build and scale powerful AI foundation models, such as Revolut’s PRAGMA and Lloyds’ ATUM. These models process billions of transactions, generate reusable intelligence, and underpin advanced AI capabilities across various domains. The technology accelerates data pipelines for tasks like fraud detection, enables real-time customer insights, and supports the development of agentic assistants, significantly enhancing efficiency and decision-making capabilities in production environments.
- Why are open weight AI models becoming increasingly important for European financial institutions?
- Open weight AI models offer European financial institutions the ability to develop proprietary intelligence using their own data, ensuring data sovereignty and control over sensitive information. These models provide significant cost savings compared to proprietary alternatives, with examples showing similar quality at a fraction of the cost. This approach fosters innovation by enabling specialized model development for distinct operations like payment processing or nuanced speech recognition, while maintaining flexibility and security.
- How is accelerated computing enhancing fraud detection and risk management across financial services?
- Accelerated computing, particularly with GPUs and toolkits like NVIDIA cuDF, drastically speeds up the processing of multi-terabyte-scale data from diverse sources. This reduces the critical time from data analysis to decision-making, enabling near-real-time client insights and prompt risk assessments. Financial providers like Stripe have achieved significantly faster data pipelines and cost reductions, allowing continuous optimization of machine learning models for improved accuracy and efficiency in identifying and preventing fraudulent activities.