Eli Lilly opens its $1B drug discovery models to biotechs

In AI drug discovery, code is cheap. High-quality wet lab data is what costs billions.

Most biotech startups hit a wall before they even begin: they have clever algorithms, but lack the proprietary experimental data required to train models that actually predict molecule behavior.

Eli Lilly just stepped in to change that equation. The pharma giant announced Lilly TuneLab, an AI/ML platform giving external biotech companies direct access to discovery models trained on decades of Lilly's in-house laboratory research.

A $1B dataset behind a federated wall

Lilly estimates that the models included in this debut release represent over $1 billion in internal research investments. The underlying system draws on preclinical, drug disposition, and safety datasets derived from hundreds of thousands of unique molecules.

Lilly's chief scientific officer Daniel Skovronsky pitched TuneLab as an industry equalizer, meant to hand small biotech teams the exact tools Lilly's internal scientists rely on daily.

There is, of course, a catch. Access is not unconditional. Selected partners must contribute their own training data back into the system to refine the models over time.

To prevent either side from exposing proprietary intellectual property, the platform is run by an independent third party and powered by federated learning. Models train across decentralized data without forcing anyone to hand over raw datasets. Lilly also plans to follow up with exclusive in vivo small molecule predictive models.

Why it matters

Lilly isn't doing this out of pure goodwill. It is quietly constructing an unbeatable gravity well for early-stage biotechnology.

For cash-strapped biotechs, the upside is massive. Having access to predictive intelligence built on hundreds of thousands of real-world molecular tests lets small teams kill failing drug candidates months earlier, cutting burn rates before heading into expensive animal studies.

However, startups should proceed with clear eyes. Participating requires feeding proprietary data into a shared flywheel. Even with federated learning keeping raw code hidden, founders are effectively helping sharpen models that Lilly and other partners can leverage.

Strategically, this cements Lilly's Catalyze360 platform, which already combines seed capital through Lilly Ventures and wet-lab space via Gateway Labs. By positioning itself as the data operating system for the next generation of biotech founders, Lilly gains early visibility and influence over promising drug targets long before competitors even see them coming.