Industry·
European Banks Turn Open Models into Proprietary Intelligence on GPU Clouds
Major European financial institutions are deploying open‑weight foundation models on NVIDIA‑accelerated infrastructure to build custom AI that stays inside their own data centers.
Open Models Drive Proprietary Intelligence
At GTC Berlin, banks such as Revolut, Starling, Lloyds and NatWest demonstrated production‑scale AI built on open‑weight models running on NVIDIA Blackwell Ultra and RTX PRO 6000 GPUs. Revolut’s PRAGMA foundation model now serves 80 million customers and is engineered to scale threefold, thanks to cuDF and Dynamo‑Triton on Blackwell Ultra.
Starling Bank trained its “Starling Intelligence” agentic model on U.K. sovereign cloud infrastructure from Civo, keeping all transaction data on‑premise. Lloyds’ ATUM model converts billions of transactions into reusable insights, accelerated by RTX PRO 6000 GPUs.
Neobank bunq runs its Finn assistant on open models hosted on Nebius, achieving closed‑model quality at roughly 60 % of the cost per answer. NatWest is fine‑tuning NVIDIA Nemotron 3.5 ASR to understand the full range of British accents and financial terminology.
Santander adopts the new NVIDIA Kumo family for structured banking data, while Modulr builds payment‑specific models from Nemotron‑3‑Ultra‑550B. LSEG leverages the Polars GPU engine with cuDF on RTX PRO 6000 GPUs to analyze tick data up to ten times faster.
- aigpu
- ai gpu
- ai gpu cloud
- aigpu dubai
- financial services
- open models
- nvidia
- accelerated computing
By AiGpu Editorial · Editorial rewrite based on public reporting (NVIDIA Blog)
← All articles