Lilly’s model-sharing deal with 1STBIO could widen the gap in AI drug discovery

This collaboration matters because it moves AI drug discovery deeper into the partnership economy. Large pharma firms increasingly possess not only capital and compound libraries, but also proprietary models built from years of internal learning. Sharing access to those models can accelerate a partner’s work dramatically, but it also makes the model itself a strategic asset.

In practical terms, the deal suggests a new kind of asymmetry in biotech. Smaller companies may gain leverage by accessing elite models, while larger firms can extend the reach of their platforms without directly scaling headcount. That could compress timelines for some programs and make partnerships more competitive, especially in areas where prediction quality meaningfully affects hit selection or lead optimization.

But model-sharing also raises a deeper question: if the most useful AI systems remain proprietary, does the industry end up concentrating advantage even further? Unlike public datasets or academic benchmarks, internal models are difficult to inspect, compare, or independently validate. That makes it harder to know whether a collaboration is transferring genuine capability or just providing a polished interface around a black box.

Even so, the deal is strategically important. It shows that AI in drug discovery is no longer just about building better tools inside the company that made them. It is becoming a tradable capability, and that may be one of the clearest indicators that the field is entering a more mature and commercially consequential phase.