NVIDIA lines up European banks, carmakers and retailers at GTC Berlin

NVIDIA used its GTC Berlin event to announce a wide set of European deployments, from banks to robotaxis. The figures below come from NVIDIA's own press release and blog posts, and have not been independently checked ^[1][2][3][4].

Robotaxis and driver assistance

NVIDIA said Bolt with Lucid, Mercedes-Benz, Volkswagen Commercial Vehicles and Einride are showcasing autonomous-mobility use cases on its Hyperion computing platform. Volvo Cars is adopting Hyperion and NVIDIA's Alpamayo Stack software ^[1]. Bolt plans to begin gathering real-world driving data in Madrid in 2027, subject to regulatory approval, as part of a planned European rollout of at least 25,000 level 4-capable future Lucid vehicles ^[1]. NVIDIA also introduced Halos SAFE, a framework for generating traceable safety evidence for autonomous vehicles ^[1]. Its release carries the usual warning that these plans are forward-looking and may change ^[1].

Banks and open models

NVIDIA says Revolut is rebuilding its technology on a single foundation model, PRAGMA, which serves an initial 80 million customers ^[2]. Starling Bank trained an agentic model on UK sovereign cloud infrastructure, and Lloyds Banking Group built a transaction foundation model called ATUM ^[2]. Neobank bunq says its assistant on open-weight models matches leading closed models at 60% of the cost per answer, and Santander is an early adopter of NVIDIA's new Kumo family of models for structured financial data ^[2]. Stripe reported 42x faster data pipelines and 92% cost savings for its fraud-detection platform in offline testing ^[2].

Retail

In retail, NVIDIA says Shopify post-trained an open model that, on Shopify's internal benchmark, beat a leading frontier model on accuracy while cutting serving costs by 90% and responding 7x faster ^[3]. Zooplus reached 86% search accuracy in a production-ready prototype, and Mirakl says its catalogue tool has onboarded more than 70 million products ^[3]. Otto Group is using digital twins to test warehouse operations in simulation before changing anything physically ^[3].

Secure computing

NVIDIA also expanded its Confidential Computing with Confidential Bare Metal, aimed at model builders who run sensitive workloads on third-party infrastructure while keeping control of their models and data. Cohere plans to use it, and Google Cloud is extending support to its A4X Max instances ^[4].

What it means for businesses

The common thread is a push towards open models run on infrastructure companies control, often framed as sovereignty and lower cost. For European firms that worry about where data goes, that is a clear message. But these are vendor-reported results from selected customers, and several are prototypes, internal benchmarks or offline tests. Buyers should ask for results on their own data before assuming similar savings.

It is also a reminder of NVIDIA's reach. Many of the named companies depend on its chips, software and reference designs, which suits NVIDIA commercially. Smaller businesses are unlikely to build robotaxi stacks, but the cost and accuracy claims around open models and fraud detection show where practical AI spending is heading in finance and retail.

Source: NVIDIA — European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion