Novo Nordisk closed its internal cell therapy unit and licensed the program out to AI-driven firm Cellular Intelligence rather than build it in-house.1 Its stock rallied nearly 25% over the following 30 days as the company refocused on its GLP-1 franchise.1
The deal signals a shift in pharma strategy: outsourcing R&D to AI-native biology platforms instead of running programs internally. Novo Nordisk's Parkinson's-focused AI partner is advancing separately, while the core business narrows around GLP-1.1
The restructuring coincides with a race among AI biology firms to own the data layer underneath drug discovery. MindWalk Holdings, NVIDIA BioNeMo, Boltz Lab, Owkin, Basecamp Research and Edison Scientific are each building platforms meant to organize scattered biological data into a usable form for drug programs.2
MindWalk's HYFT Technology is built on roughly 660 million biological patterns — what the company calls "a proprietary, function-aware representation of biology built on roughly 660 million biological patterns, universal fingerprints that encode conserved relationships between sequence, structure, and function."3 MindWalk recently filed a new patent, EP26187897.9, which it says is "the computational layer built on the original foundation, not a re-filing of it," building on its earlier HYFT patent EP3881326A1.4
The company's pitch to pharma partners centers on data itself as the scarce asset: "the representation of biology itself is the hard part, and the place to build a durable advantage."3 MindWalk also gave its first public demonstration of its ReefIQ platform this month.2
NVIDIA has anchored lab-in-the-loop partnerships with Thermo Fisher and Eli Lilly, pairing its BioNeMo platform with pharma manufacturing and research infrastructure. Together, the moves point to large pharma companies licensing AI capability rather than building it, while a smaller set of AI-native firms compete to supply the underlying biological data layer.
