In this case study, NVIDIA, the $130B+ market cap leader in AI computing and GPU infrastructure, partnered with Eli Lilly, the $700B+ market cap pharmaceutical company with the industry’s most powerful AI supercomputer, to solve drug discovery’s fundamental integration gap: AI models and wet lab experiments operate in separate silos, with results rarely fed back into model training in real time. In January 2026, the two announced a first-of-its-kind AI co-innovation lab with joint investment up to $1B over five years. Based in South San Francisco, the lab co-locates Lilly’s biology and medical experts with NVIDIA AI engineers to build a continuous learning system connecting automated wet labs with computational dry labs for 24/7 AI-assisted drug discovery. We evaluated this partnership to help your team structure similarly ambitious AI-industry collaborations.
1. Executive Summary
In January 2026, NVIDIA and Eli Lilly announced a first-of-its-kind AI co-innovation lab with joint investment of up to $1 billion over five years. The lab, based in South San Francisco, co-locates Lilly domain experts in biology and medicine with NVIDIA AI engineers to build a continuous learning system connecting agentic wet labs with computational dry labs — enabling 24/7 AI-assisted experimentation for drug discovery.
- Subject: NVIDIA ($130B+ market cap, AI computing leader) and Eli Lilly ($700B+ market cap, top pharmaceutical company)
- Problem: Drug discovery is slow and expensive — AI models are not yet integrated with experimental wet lab data in a continuous, real-time loop
- Solution: Co-located co-innovation lab with joint investment, shared infrastructure on NVIDIA BioNeMo platform, and a continuous learning system connecting AI models to automated experiments
- Result: First-of-its-kind continuous learning system for drug discovery; $1B commitment signals confidence; BioNeMo platform validated for pharmaceutical use
2. The Challenge
Drug discovery requires 10-15 years and billions of dollars per approved drug. Despite significant advances in AI, the technology has not yet transformed this timeline at scale because AI models and wet lab experiments operate in separate silos — experimental results are not fed back into model training in real time, and model predictions are not directly tested by automated experiments.
- Time and cost problem: Drug discovery requires 10-15 years and billions of dollars per approved drug. AI has not yet transformed this timeline at scale.
- Siloed workflows: AI models and wet lab experiments operate in separate silos — results aren’t fed back into model training in real time. The learning loop is broken.
- Infrastructure gap: No pharmaceutical company has the in-house AI infrastructure to build a continuous learning system alone. The required compute, AI talent, and integration expertise exceed any single company’s capabilities.
Both sides recognized that incremental improvements would not close the gap. Lilly needed a step-change in AI capability that its internal team could not deliver alone. NVIDIA needed a marquee pharmaceutical partner to validate BioNeMo as the industry-standard platform for AI drug discovery.
3. The Strategy
Rather than funding a traditional software licensing deal or building separate AI initiatives, NVIDIA and Lilly designed a co-innovation lab with co-located teams, joint investment, and a continuous learning system architecture that connects automated wet labs directly to AI model training — creating a closed loop that accelerates with each iteration.
- Continuous learning system: AI models trained on experimental results design the next experiment automatically. Lilly’s automated wet labs feed data directly into NVIDIA’s BioNeMo platform, which designs the next experiment — creating a 24/7 discovery cycle.
- Co-location of domain experts and AI engineers: Lilly biologists and NVIDIA AI engineers share a single South San Francisco facility. This physical proximity accelerates iteration cycles that remote collaboration would slow significantly.
- Joint investment structure: $1B split between both companies creates shared risk and aligned incentives. Both sides win only if the lab produces valuable drug candidates — a structural departure from fee-for-service AI consulting.
The lab extends beyond drug discovery into manufacturing, with digital twins of Lilly’s production lines built on NVIDIA Omniverse. This vertical integration — from AI model training through wet lab experimentation to manufacturing optimization — creates a comprehensive platform that no other pharmaceutical company has.
4. The Results
The NVIDIA-Lilly AI co-innovation lab, announced at the J.P. Morgan Healthcare Conference in January 2026, represents the most ambitious AI-drug discovery partnership ever formed — though measurable drug discovery output will take years to materialize.
- Landmark partnership: World’s most ambitious AI-drug discovery partnership, announced at J.P. Morgan Healthcare Conference — the most prominent stage in healthcare investing — signaling strategic importance to both companies.
- Integrated AI infrastructure: Lilly’s AI supercomputer (most powerful in pharma) integrated with NVIDIA BioNeMo platform, creating a vertically integrated system from compute to wet lab.
- Beyond discovery: Partnership extends into manufacturing digital twins using NVIDIA Omniverse, indicating the scope reaches from drug candidate identification through production optimization.
The partnership’s $1B commitment signals confidence in future output, but no drug candidates or publications are yet attributable to the lab. Long-term output — clinical candidates, approved drugs, BioNeMo platform adoption — will take 3-7 years to materialize.
5. The Melan Approach
Melan advises structuring partnerships like this one when the capability gap requires deep integration of AI platforms with domain-specific operations — the co-innovation lab model works best when both sides contribute capital and talent and the shared goal requires combining complementary technologies that neither partner can build alone.
- Governance model: Co-located teams with joint investment creating aligned incentives. Melan would recommend adding an academic advisory board of leading computational biology and chemistry faculty to provide external validation and frontier research access that pure corporate structures risk missing.
- Risk allocation: Joint investment shifts risk equally between both partners. Melan recommends allocating 10-15% of the budget for independent validation of AI model predictions by academic or contract research organizations — ensuring the continuous learning system’s outputs are rigorously verified.
- Shared goal: Transform drug discovery through continuous AI-wet lab integration while making BioNeMo the industry-standard platform for pharmaceutical AI. Melan recommends documenting platform-agnostic benchmarks to ensure the partnership produces transferable knowledge, not just platform lock-in.
This co-innovation lab model is replicable at lower investment levels ($50-100M) for mid-cap pharmaceutical companies and AI platform providers that cannot commit $1B but want the same structural advantages of co-located teams, joint investment, and continuous learning system architecture.
Building an AI-drug discovery co-innovation lab?
Melan helps pharmaceutical companies and AI platform providers structure co-innovation labs with continuous learning system architectures, joint investment frameworks, and academic advisory boards.
