Pharmaceutical companies are discovering that their decades of accumulated research data is worth billions when paired with artificial intelligence. Rather than simply buying AI startups or developing drugs the traditional way, major pharma firms are now licensing their proprietary datasets to train powerful AI models. These data-licensing deals represent a fundamental shift in how drug discovery works, turning internal research archives into strategic assets that can accelerate the entire industry.

Why Are Pharma Companies Suddenly Treating Data Like Currency?

For years, pharmaceutical companies kept their research data locked away. Screening results from millions of compound tests, genetic information from clinical trials, protein structures, and chemical optimization records remained internal tools used only for individual drug programs. But the rise of foundation models, large AI systems trained on massive diverse datasets, has changed everything.

Foundation models work differently than traditional software. They're pre-trained on enormous amounts of data, then fine-tuned for specific tasks, much like how GPT-4 learns language patterns before being adapted for specialized uses. Building these models requires vast, high-quality datasets spanning multiple types of information, from molecular structures to patient outcomes. Pharma companies, it turns out, sit on exactly the kind of data these AI systems need.

"Foundation models are only as good as the underlying training data they are built upon," noted Kim Branson, GSK's AI head.

Kim Branson, AI Head at GSK

This realization has sparked a wave of licensing agreements. Instead of keeping data proprietary, companies now recognize that licensing it to AI partners can accelerate drug discovery across the entire industry, while generating substantial revenue through upfront payments, milestone bonuses, and royalties.

What Do These Multibillion-Dollar Data Deals Actually Look Like?

Recent high-profile agreements reveal the structure and scale of pharma's data-licensing boom. Incyte, a biopharmaceutical company, expanded its collaboration with Genesis Molecular AI in 2026 with a deal worth $120 million upfront, plus over $1 billion in potential milestone payments and royalties. This agreement explicitly licenses Incyte's experimental data for training large-scale AI foundation models.

The Incyte-Genesis deal isn't an outlier. AstraZeneca and Tempus announced a $200 million data-licensing agreement to build a specialized AI model for oncology research using Tempus's dataset of 7.3 million patient records. GSK licensed Noetik's cancer "virtual cell" foundation models for $50 million upfront, plus additional license fees. Recursion Pharmaceuticals signed a $160 million data-licensing deal with Tempus for AI-ready clinical and molecular data.

These deals follow a consistent pattern. The pharma company grants an AI partner rights to use its proprietary datasets for training, while retaining rights to any resulting drug discoveries. Payment structures typically include three components:

• Upfront Fees: Large cash payments or equity stakes paid immediately upon signing, ranging from $50 million to $200 million depending on data quality and scope.
• Milestone Payments: Additional payments triggered when the AI model achieves specific development goals, such as identifying promising drug candidates or advancing compounds into clinical trials.
• Royalties: Ongoing payments based on the commercial success of drugs discovered using the licensed data, ensuring pharma companies benefit from long-term outcomes.

The inclusion of equity stakes in some deals, like Incyte's arrangement with Genesis, signals how valuable pharma companies view their data. By taking equity positions, they're betting that AI-driven drug discovery will generate substantial returns.

What Types of Data Are Worth Billions?

Pharmaceutical companies generate diverse datasets across their research operations, and each type has distinct value for AI training. Understanding what data is being licensed reveals why these agreements command such high prices.

• Chemical and Screening Data: Results from screening millions of compounds against disease targets, including molecular structures and activity measurements like IC50 values, which indicate how effectively a compound binds to its target.
• Medicinal Chemistry Data: Detailed structure-activity relationship information showing how specific chemical modifications affect biological activity, along with proprietary chemical series and optimization histories.
• Protein and Biology Data: Experimental results on protein targets, including binding affinities, crystallography images, protein interaction networks, and phenotypic screening results that reveal how compounds affect cell behavior.
• Genomics and Omics Data: Patient-derived datasets including genomic sequences from clinical trials, transcriptomics data showing which genes are active, proteomics data revealing protein levels, and links between genetic variations and drug responses.
• Clinical and Real-World Evidence: Electronic health records and outcomes data from actual patients, providing context for how drugs perform outside controlled laboratory settings.

The combination of these datasets is what makes pharma data so valuable. A single company might have decades of screening results, but pairing that with genomic data and clinical outcomes creates a comprehensive training resource that AI models can use to learn patterns humans might miss.

How to Structure a Pharma Data-Licensing Deal: Key Considerations

For companies considering data-licensing agreements, several structural elements determine whether a deal creates value or creates problems. Based on recent high-profile agreements, here are the critical components:

• Data Exclusivity Terms: Determine whether the AI partner can use the data exclusively for one pharma company or can license similar data from competitors. Exclusive arrangements command higher prices but limit the AI partner's ability to build diverse models.
• Intellectual Property Rights: Clarify who owns the resulting AI models, the discoveries made using those models, and any patents generated. Most deals allow pharma companies to retain rights to drug candidates while the AI partner retains rights to the underlying model technology.
• Data Security and Compliance: Establish protocols for protecting sensitive patient information and proprietary research, including data anonymization requirements and compliance with regulations like HIPAA for health data and GDPR for European patient information.
• Performance Metrics: Define how the AI model's success will be measured and how milestone payments will be triggered, such as identifying compounds that advance to specific development stages or demonstrating predictive accuracy above certain thresholds.

"High-quality data is among the most valuable inputs for advancing molecular AI," stated Incyte's CEO.

Incyte CEO

These structural decisions reflect a broader shift in pharma strategy. Rather than viewing data as a byproduct of research, companies now see it as a strategic asset that can generate revenue, accelerate discovery timelines, and create competitive advantages when paired with the right AI partners.

What Does This Mean for the Future of Drug Discovery?

The explosion of pharma data-licensing deals signals a fundamental transformation in how new medicines will be discovered. Instead of individual companies working in isolation, the industry is moving toward collaborative models where data flows to AI specialists who build foundation models, which then flow back to pharma companies as tools for accelerating their own research.

This shift mirrors earlier transitions in pharma history. Just as companies moved from discovering drugs through trial-and-error to rational drug design based on understanding molecular targets, the industry is now moving toward AI-accelerated discovery powered by data. The companies that accumulate the highest-quality datasets and structure the most favorable licensing agreements will likely gain the greatest advantage in bringing new drugs to market faster and more cost-effectively.

The scale of investment in these deals suggests the industry believes AI-driven drug discovery will deliver substantial returns. With agreements ranging from $50 million to $200 million upfront, plus billions in potential milestone payments, pharma companies are betting that licensing their data will accelerate discovery timelines and reduce the time and cost required to bring new medicines to patients.