Novo Nordisk has chosen Amazon Web Services as its preferred cloud provider and strategic AI partner, deepening a technology relationship that now reaches from workplace automation into the scientific process of identifying and developing new medicines.

The Danish pharmaceutical company and AWS have established a co-innovation hub at Novo Nordisk’s King’s Cross facility in London, where research teams will work directly with engineers, AI specialists, and applied scientists from Amazon. Their aim is to shorten the path between identifying a potential drug target and reaching the first human dose by reducing hand-offs between scientific research and the teams building AI systems around it.

Novo Nordisk will use AWS technologies including Amazon Bio Discovery, Amazon Bedrock, and Bedrock AgentCore across research and development, commercial operations, and enterprise IT. The agreement extends beyond infrastructure hosting because AWS engineers will work alongside Novo Nordisk teams on applications while the pharmaceutical group expands its use of AI agents and biological models.

The companies are entering the partnership with a sizeable internal deployment already in place. More than 25,000 Novo Nordisk employees use a generative AI system built with Amazon Bedrock to retrieve information, draft documents, build chatbots, and support other non-regulated activities. Novo Nordisk also says its existing work with AWS has reduced time spent producing clinical documentation, although it has not disclosed a comparable figure for that improvement.

Moving AI further into pharmaceutical research changes the consequence of an error. A workplace assistant that produces an unhelpful document can be corrected with relatively little cost, whereas systems used to evaluate drug candidates, connect biological datasets, or inform experimental choices operate much closer to expensive and heavily governed scientific decisions.

The London hub is therefore more consequential than a conventional preferred-cloud agreement. Drug discovery brings together biologists, chemists, clinicians, data specialists, and engineers, with each group working through different evidence and tooling. Novo Nordisk and AWS are trying to reduce the delay between computational findings and physical experiments by placing the people responsible for both within the same operating environment.

Amazon Bio Discovery is intended to give scientists access to biological AI models that can help generate and evaluate potential drug candidates before linking computational design with laboratory testing. Bedrock AgentCore, meanwhile, provides infrastructure for agents that can work across enterprise data and multi-step processes, creating a route for automation to move into operational tasks as well as research.

The arrangement also illustrates how enterprise AI adoption separates into different risk classes within a single organisation. Novo Nordisk can give tens of thousands of employees generative tools for non-regulated work while applying stronger controls to systems touching scientific development, clinical activity, manufacturing, or other regulated processes. Treating both as one corporate AI programme would obscure the different validation and oversight requirements.

Pharmaceutical research is unusually suited to the infrastructure advantages large cloud providers are trying to sell around AI. Genomic, imaging, clinical, and molecular datasets can be computationally intensive, while model development and inference require specialised computing that few companies want to operate entirely as isolated internal infrastructure.

Concentrating more of that work on one strategic provider creates a corresponding dependency. Making AWS the preferred cloud provider and AI partner should simplify integration and joint development, but it also gives one technology supplier a broader role across computing, models, agent infrastructure, and specialist life-sciences applications.

The commercial value will consequently depend less on how many AI products Novo Nordisk deploys than on whether they shorten expensive scientific processes without weakening the evidence needed to make decisions. Drug discovery contains many tasks that can be accelerated computationally, but a candidate still has to survive laboratory work, clinical development, regulatory review, and manufacturing requirements.

The ambition to compress the journey from drug target to first human dose therefore sits within hard biological and regulatory limits. AI can reduce search and coordination costs, help scientists explore more possibilities, and automate parts of information processing, but it does not remove the experimental stages that distinguish a plausible computational result from a medicine.

The London location gives the partnership a tangible UK component even though Novo Nordisk is Danish and AWS is American. Rather than establishing another broad AI centre with an undefined brief, the hub is attached directly to existing research and development activity, with AWS’s technical teams working alongside scientists on drug-discovery and operational problems.

Novo Nordisk already has tens of thousands of employees using generative AI. The harder test begins as the technology moves nearer the scientific decisions that determine which potential medicines consume years of research and substantial capital.