![\"photograph](\"https:\/\/www.europesays.com\/dk\/wp-content\/uploads\/2026\/08\/photograph-taken-november-12-2025-shows-sign.jpg\")
<\/p>\n<\/p>\n
This photograph taken on November 12, 2025 shows a sign bearing the logo of Novo Nordisk next to the company’s factory in Hillerod.
\nSERGEI GAPON\/AFP via Getty Images<\/p>\n
Novo Nordisk and Amazon Web Services announced a strategic AI partnership<\/a> that puts AWS engineers and scientists physically inside Novo Nordisk’s London research facility \u2014 and deploys Amazon Bio Discovery, a platform that lets bench biologists design drug molecules using plain-language instructions rather than code, directly against the company’s live drug discovery pipeline. The deal also names AWS as Novo Nordisk’s preferred cloud provider, making it the Danish pharmaceutical giant’s third major AI infrastructure partner in roughly twelve months.<\/p>\n “Partnering with AWS is about making AI a driver of innovation across Novo Nordisk,” said Mike Doustdar, president and CEO of Novo Nordisk. “By combining our scientific expertise with AWS’ tools and purpose-built services for the life sciences industry, while also creating a dedicated co-innovation hub within our existing facility in London, we are accelerating discovery and developing responsible AI solutions that will create real progress for people living with serious chronic diseases.”<\/p>\n The partnership arrives at a pivotal moment for Novo Nordisk. The company reported two significant pipeline setbacks in late July: its cardiovascular drug ziltivekimab failed its primary endpoint<\/a> in the Phase 3 ZEUS trial involving more than 6,300 patients, and it ended development of obesity candidate monlunabant after neuropsychiatric safety concerns. The failures come as Novo Nordisk faces intensifying pressure from Eli Lilly in the diabetes and obesity drug markets, where it has ceded significant market share since 2024. The AWS deal signals that pipeline velocity \u2014 getting to first-in-human testing faster \u2014 is now a stated strategic priority, not just a long-term aspiration.<\/p>\n How Amazon Bio Discovery Works: Biologists in the Driver’s Seat<\/p>\n At the center of the partnership is Amazon Bio Discovery, an agentic AI application AWS launched in April 2026<\/a> at its annual Life Sciences Symposium. The platform is built around a three-component architecture designed to solve a specific and well-documented problem in pharmaceutical R&D: the expertise barrier that separates wet-lab scientists from the computational tools that could accelerate their work.<\/p>\n The first component is a catalog of more than 40 biological foundation models<\/a> \u2014 large AI systems trained on vast biological datasets, each specialized for a different aspect of drug molecule design. These models handle tasks including antibody binding affinity prediction, protein structure prediction, developability scoring (whether a candidate molecule can be manufactured reliably at scale without degrading), and de novo nanobody design using diffusion-based generative methods. One model in the catalog, developed in partnership with the Gray Lab at Johns Hopkins, was trained on what AWS describes as the most diverse and largest antibody dataset currently available in scientific literature.<\/p>\n The second component \u2014 and the one that fundamentally changes who can use the platform \u2014 is a conversational AI agent. A medicinal chemist or wet-lab biologist can describe what they want to achieve in plain language: design antibodies against a specific cancer target, identify which binding-critical amino acids to optimize, or benchmark which biological foundation model best predicts their compound’s stability profile. The agent walks through the key decisions interactively, constructs an “experiment recipe” \u2014 a step-by-step workflow combining multiple models in sequence \u2014 and explains its reasoning in plain language rather than returning raw model outputs that require data science training to interpret.<\/p>\n The third component closes the loop between computational design and physical chemistry: direct integrations with contract research organizations Twist Bioscience and Ginkgo Bioworks, which synthesize and test the highest-ranked molecule candidates. Results are automatically routed back into the platform, enabling researchers to fine-tune biological foundation models on their own experimental data with each iteration \u2014 while keeping that proprietary data inside an isolated environment that AWS says it does not use for hosted model training.<\/p>\n The performance case for this architecture was established before Novo Nordisk signed on. Memorial Sloan Kettering Cancer Center used Amazon Bio Discovery to design antibody candidates for potential pediatric cancer therapies, generating nearly 300,000 novel antibody molecules and routing the top 100,000 for synthesis testing \u2014 a process that previously took up to a year<\/a> using traditional methods \u2014 in a matter of weeks.<\/p>\n The London Hub: When a Cloud Partnership Becomes a Physical Presence<\/p>\n What distinguishes the Novo Nordisk deal structurally from a standard software licensing agreement is the co-innovation hub: a dedicated workspace within Novo Nordisk’s existing London facility where engineers, AI specialists, and applied scientists from AWS work alongside Novo Nordisk’s R&D teams, not remotely. The hub deploys not just Amazon Bio Discovery but also Amazon Bedrock \u2014 for connecting genomic, imaging, and clinical data to accelerate target identification \u2014 and Amazon Bedrock AgentCore, which runs autonomous AI agents across Novo Nordisk’s operational workflows.<\/p>\n The physical embedding is enabled by AWS’s recently launched Forward Deployed Engineering organization<\/a>, backed by a $1 billion investment announced in June 2026. The FDE model places AWS engineers inside client companies for accelerated deployment engagements, prioritizing customer self-sufficiency after deployment. “Every part of our work with Novo Nordisk comes back to one thing: getting the right medicine to patients, faster,” said Matt Garman, CEO of AWS.<\/p>\n The hub also taps AWS Forward Deployed Engineers \u2014 the technical deployment unit \u2014 to embed directly with Novo Nordisk’s R&D staff, building institutional knowledge transfer alongside the technology deployment itself.<\/p>\n Novo Nordisk will additionally collaborate with Amazon Pharmacy, Amazon One Medical, and Amazon Ads through the broader partnership to improve how its approved therapies are marketed and delivered to patients.<\/p>\n Third Platform in a Year: What the Multi-Vendor Strategy Signals<\/p>\n The AWS announcement is the third major AI infrastructure commitment Novo Nordisk has made in roughly twelve months, and the pattern is not coincidental. In June 2025, Novo Nordisk partnered with the Danish Centre for AI Innovation to access the Gefion sovereign AI supercomputer<\/a>, built on NVIDIA DGX SuperPOD infrastructure, for computational drug discovery, molecular simulation, and physical AI applications \u2014 focusing on the compute layer. In April 2026, it announced a wide-ranging partnership with OpenAI to integrate generative AI<\/a> across drug discovery, manufacturing, supply chain, and commercial operations, with full deployment targeted by the end of this year. The AWS partnership adds embedded cloud infrastructure, agentic workflow orchestration, and the specific wet-lab interface that Bio Discovery provides.<\/p>\n The three partnerships address structurally different layers of the AI stack: the Gefion supercomputer provides the compute substrate; OpenAI provides general-purpose scientific reasoning models accessible to researchers; AWS provides the cloud infrastructure, the biologically specialized platform (Bio Discovery), and the physical engineering presence to deploy everything in production. Taken together, they represent a deliberate multi-vendor architecture rather than a bet on any single platform \u2014 a hedge against single-source dependency in a space where no provider has yet demonstrated full-stack superiority from target identification to clinical proof.<\/p>\n Mishal Patel, Group Vice President of AI and Digital Innovation in R&D at Novo Nordisk, articulated the logic earlier this year<\/a> in the context of the OpenAI partnership: “Life sciences research is complex, data-rich, and interdisciplinary. To deliver meaningful value for researchers, advanced AI models must be grounded in trusted scientific data, connected to validated tools, and integrated into the real-world workflows researchers use every day.”<\/p>\n The competitive pressure driving this investment is concrete. Eli Lilly signed a $2.75 billion deal with AI drug discovery firm Insilico Medicine in March 2026<\/a>, acquiring an exclusive worldwide license to develop and commercialize a portfolio of AI-designed drug candidates, at least 28 of which have already been generated by the AI platform, with nearly half in clinical testing. More than $11 billion flowed into AI drug discovery and licensing in 2025 alone across 348 funding rounds, according to data from DealForma<\/a> reported at the BIO 2026 International Convention.<\/p>\n What This Deal Means for AWS<\/p>\n For AWS, the Novo Nordisk partnership is a marquee win in pharmaceutical life sciences \u2014 a sector it has invested heavily to penetrate with purpose-built tooling. 19 of the top 20 global pharmaceutical companies run on AWS cloud infrastructure<\/a>, giving the company an entrenched position from which to sell higher-value AI services like Bio Discovery and Bedrock AgentCore upward.<\/p>\n The deal also reflects a broader shift in what enterprise AI deployment looks like in practice. The agentic AI layer \u2014 autonomous AI systems capable of completing multi-step scientific and operational tasks without hand-holding at every step \u2014 is moving from experimental pilots into production workloads at leading pharmaceutical companies. Amazon Bedrock processed more tokens in the first quarter of 2026 than in all prior years combined, with 170% growth in customer spend quarter-over-quarter, according to Amazon’s Q1 2026 earnings disclosures.<\/p>\n Does Deploying AI on Cloud Infrastructure Introduce New Risks for Pharmaceutical R&D?<\/p>\n This is not a hypothetical concern. In June 2026, Novo Nordisk disclosed that a limited quantity of clinical trials data had been copied without authorization in an unauthorized access incident<\/a> \u2014 part of a broader wave of pharmaceutical sector breaches that year. Separately, threat intelligence reporting documented a 2026 pharmaceutical incident in which attackers specifically obtained proprietary AI drug-discovery models and datasets using a single exposed developer credential, reflecting a paradigm shift: as pharmaceutical companies embed AI into core discovery workflows, the models and datasets themselves become high-value intellectual property that attackers treat as primary targets.<\/p>\n