NVIDIA Expands UK Startup and Industry Work Across AI, Healthcare and Quantum

NVIDIA’s latest UK activity spans early-stage companies and large-scale deployments, tying startup investment to work in science, healthcare, financial services and quantum computing. The company says it plans to invest in five UK startups: Boltz, Foresight, Hiverge, Prima Mente and Riverlane. The range of projects shows where local teams are applying AI and accelerated computing, while also making clear that the work varies widely in maturity and purpose.

The investments sit alongside NVIDIA’s five-year, $250 million compute commitment to support UK research on agentic safety and security. NVIDIA has previously backed companies including Basecamp Research, Ineffable Intelligence, Synthesia and Wayve. It also joined the Earthshot Prize as a Global Alliance Founding Partner, contributing computing resources and technical expertise to teams working on climate and sustainability challenges.

Five startups, five different technical bets

Boltz, based in London and founded by MIT researchers, develops biomolecular models and AI agents. The tools are intended to help scientists predict molecular structures and interactions, then use those predictions to design candidate medicines. That puts the company’s work at the intersection of computational biology and drug research, where predictions can help narrow experiments but still need scientific validation.

Foresight, another London startup, applies AI to project planning for major infrastructure, including hyperscale data centers. Its software helps owners forecast completion dates, identify risks and move facilities toward revenue sooner. The practical value depends on how well its forecasts reflect changing project conditions and whether teams can act on the risks it flags.

Cambridge-based Hiverge is building an AI discovery engine that writes and optimizes algorithms. Founded by researchers from Google DeepMind and the University of Cambridge, it targets tasks such as speeding up AI training and planning complex logistics. The goal is to automate parts of algorithm design, though the source doesn’t report benchmark results or specify how its system compares with existing optimization methods.

Prima Mente, a London-based AI biology company, is developing biological foundation models to study complex diseases and identify therapeutic targets. It is starting with Alzheimer’s, Parkinson’s and ALS. As with other disease-modeling efforts, the key question will be whether model outputs can guide useful experiments and eventually improve the search for treatments.

Riverlane, based in Cambridge, makes hardware and software for real-time quantum error correction across major qubit types. Error correction is a central engineering challenge: quantum systems are prone to errors, and practical computation depends on detecting and correcting them as calculations run. NVIDIA’s planned investment adds to the attention on this work, while the company’s products aim to help quantum computers reach greater reliability.

AI moves into operational workloads

Several UK organizations are applying AI to customer operations and data infrastructure. Fuse Energy is moving its customer AI onto infrastructure it operates itself, built on NVIDIA B300 Blackwell Ultra. The company says the system supports more responsive customer service for more than 450,000 UK households. The source doesn’t provide response-time or cost comparisons, so the reported scale describes the service reach rather than measured performance gains.

Locai Labs has released Nemotron-Personas-UK, an open source dataset of one million synthetic personas designed to represent UK demographic, regional and cultural diversity. Developers can use synthetic data to build and evaluate systems where representative examples matter. The dataset’s size and intended coverage are clear, but those figures alone don’t establish how closely the personas reflect real users or how well systems trained with them perform.

Healthcare and scientific workflows

Healthcare projects cover triage, clinical documentation, speech recognition and hospital logistics. OneAdvanced is deploying what it describes as the UK’s first and most advanced fine-tuned large language model for primary care AI triage. Built with NVIDIA Agent Toolkit software and open models, the system is intended to support millions of NHS patients. The source gives no evaluation results or details on how clinicians review its outputs.

Heidi has integrated NVIDIA Nemotron open models to help clinicians find trusted medical evidence more quickly. Speechmatics is deploying Oak, a multilingual medical speech recognition model, with NVIDIA Triton Inference. Its setup lets healthcare organizations control where sensitive patient audio is processed, a practical deployment consideration for systems handling clinical conversations.

Apian, Synnovis and NVIDIA are using NVIDIA Cosmos for computing and digital simulations aimed at helping hospital robots move laboratory samples. The proposed benefit is to reduce routine transport work for medical staff. Separately, Danaher and NVIDIA are working on autonomous science, combining accelerated computing, scientific AI and physical AI in laboratory workflows and instruments. Danaher’s recently announced autonomous lab is designed to deliver up to eight times faster discovery of affinity reagents. It uses NVIDIA BioNeMo open models in protein design workflows that connect computational design with experimental feedback.

Finance, cities and quantum computing

In finance, the London Stock Exchange Group and NVIDIA are bringing GPU-accelerated market analytics to global institutions. Revolut’s PRAGMA model, introduced earlier in the year, was pretrained on 26 million anonymized user histories representing banking event sequences and has moved from research toward real-time scaling. Lloyds Banking Group has built ATUM, a transaction foundation model that turns billions of transactions into reusable AI. Starling is developing a banking foundation model on Civo’s UK sovereign cloud infrastructure.

NatWest is working with NVIDIA on Nemotron Voice to better understand how banking customers across the UK speak. Modulr is deploying a network of agents using Nemotron for separate payment operations. These projects apply models to distinct tasks, from speech understanding to transaction processing; the source doesn’t provide comparative accuracy or operating-cost figures.

Liverpool City Region is developing an NVIDIA-powered “City Brain” as a sovereign AI strategy that could serve as a blueprint for dozens of unitary authorities across the UK. Digital Jersey, Innovation City Belfast and Tees Valley Combined Authority are also exploring trusted AI computing for communities and public services.

Quantum computing offers a more experimental example of the same computing ecosystem. Riverlane has demonstrated GPU acceleration for recalibrating quantum error-correction decoders as hardware behavior changes. ORCA Computing unveiled its PT-3 photonic quantum accelerator at the Digital Realty Innovation Lab and used it to run NVIDIA’s Proteina model for protein design. ORCA reports that its hybrid quantum-classical extension improved the identification of useful proteins by 20%; the source doesn’t provide further details about the evaluation behind that result.

NVIDIA says more than 3,100 UK companies are part of its Inception program for startups. Across these examples, the practical test is whether access to computing and new models translates into measurable gains in scientific discovery, service quality or operational performance. The clearest next signals will be published evaluations, real-world deployment results and evidence of how these systems perform as workloads grow.