NVIDIA's BioNeMo platform has secured partnerships with Eli Lilly and Thermo Fisher, positioning itself as the global default compute layer for pharmaceutical drug discovery.
The consolidation mirrors a pattern already seen in enterprise cloud. When Amazon, Microsoft, and Google captured infrastructure, software firms built on top rather than beneath. BioNeMo is replicating that model in biology — owning the hardware and APIs that specialized models depend on, without needing to own the science itself.
Denmark's Novo Nordisk made the logic explicit this month. The company closed its internal cell therapy unit and licensed the program to Cellular Intelligence, an AI-native biotech targeting Parkinson's disease.1 Shares rose 25% over the following month — investors read the exit from internal development as discipline, not retreat.1
The foundation model layer is crowding globally. Natera, Basecamp Research with its EDEN model, Boltz, Owkin, and Edison Kosmos have each launched platforms targeting distinct pipeline stages. BioNeMo's advantage sits beneath all of them: specialized models run on NVIDIA hardware regardless of origin or geography.
For pharmaceutical companies across the US, Europe, and Asia, the math is consistent. Internal AI teams take years to build, carry high retention costs, and compete for talent against technology firms. Licensing platform access converts fixed R&D spend into variable cost and compresses timelines.
Novo Nordisk's simultaneous pivot back to its GLP-1 franchise — the category generating most of its global revenue — while outsourcing cell therapy to a specialist reflects the new model. Big Pharma is becoming a portfolio manager of AI-native programs, not a builder of every capability.
Infrastructure lock-in through dependency, not exclusivity, is NVIDIA's durable position. The 25% market signal from Novo Nordisk suggests investors globally are pricing that transition in — and placing NVIDIA at the center of it.
