NVIDIA announced at GTC Berlin on October 22 that Revolut is scaling its PRAGMA foundation model in production for an initial 80 million customers. Other projects cover bank-owned models, payment operations and accelerated data processing, ranging from live systems to development work and offline tests.
Key Takeaways
- Revolut’s PRAGMA is serving an initial 80 million customers and is engineered to scale to three times that level.
- Banks and payment providers are developing proprietary intelligence using their own data, open-weight models and accelerated infrastructure.
- NVIDIA reported substantial processing gains, including Stripe’s offline test results of 42 times faster pipelines and 92% cost savings.
Table of Contents
- European financial AI moves into production
- Open models support assistants and specialized workloads
- Financial research gets faster data tools
- Fraud-processing gains include offline test results
European financial AI moves into production
Revolut is rebuilding its technology stack around PRAGMA as a single foundation model. According to NVIDIA, the bank is using its cuDF toolkit and Dynamo-Triton software on Blackwell Ultra GPUs to scale the system in production. The stated threefold scaling capacity is an engineering target, not a reported expansion of the current customer base.
The announcement also highlighted institutions building intelligence from their own transaction records. Starling Bank developed Starling Intelligence, the layer underpinning its AI capabilities, on Civo’s U.K. cloud using the bank’s transaction data. NVIDIA described Starling as the first bank to train an agentic model on U.K. sovereign cloud infrastructure.
Lloyds Banking Group built ATUM, a transaction foundation model that turns billions of banking transactions into reusable intelligence. NVIDIA said RTX PRO 6000 Blackwell GPUs accelerate the model, which is improving efficiency across bank-wide use cases. Separately, SBS launched a sovereign banking intelligence platform designed to let banks run generative and agentic models without moving data outside their own walls.
Open models support assistants and specialized workloads
Neobank bunq’s Finn assistant now runs on open-weight models. NVIDIA said its deployment on Nebius matches the quality of leading closed models at 60% of the cost per answer. That figure describes the relative cost, rather than a 60% cost reduction.
Other projects remain under development. NatWest is building speech technology designed to recognize British accents, dialects and financial terminology using NVIDIA speech models, including Nemotron 3.5 ASR. Modulr is developing specialized models for different payment operations based on the Nemotron-3-Ultra-550B open model.
Santander is an early adopter of NVIDIA’s new Kumo family of open-weight foundation models for structured financial data. It is using the models to build specialized intelligence for core banking workloads.
Financial research gets faster data tools
LSEG is working with NVIDIA to help financial institutions analyze large volumes of historical tick data up to 10 times faster. The work uses the Polars GPU engine, accelerated by cuDF on RTX PRO 6000 Blackwell GPUs.
RavenPack has built a dataset-generation pipeline on Bigdata.com using a fine-tuned NVIDIA Nemotron model. It takes a research hypothesis expressed in plain language and returns a labeled dataset ready for backtesting, which investors can use to identify, analyze and visualize market trends.
Fraud-processing gains include offline test results
HSBC uses Starburst Enterprise’s GPU Engine to support near-real-time client insights and on-demand decision-making while minimizing risk and fraud. NVIDIA reported fivefold faster data processing.
Stripe’s results were explicitly limited to offline testing. NVIDIA said GPU-accelerated Ray Data delivered 42 times faster data pipelines and 92% cost savings for Stripe Radar’s fraud-detection platform in those tests. The reported gains concern machine-learning data pipelines, rather than a quantified improvement in fraud-detection accuracy.
Read More News
Disclaimer
This FinanceVelix article provides general information, not personalized financial, investment, legal, or tax advice. Market prices, interest rates and economic conditions can change; investments carry a risk of loss. Facts reflect the material available when published and may change. Check important information independently and consult a qualified professional when needed. Mentioning an asset, company, or service is not an endorsement.
