MindWalk Holdings filed patent EP26187897.9 on July 15, adding a computational layer atop its foundational HYFT patent, EP3881326A1.1 The filing organizes biological meaning into a form reusable across AI workflows.1
HYFT encodes roughly 660 million biological patterns — fingerprints linking sequence, structure and function.2 MindWalk's wager: representing biology itself, not raw compute, is the harder problem worth owning.2
NVIDIA is chasing the same ground from Silicon Valley. Its BioNeMo platform now anchors lab-in-the-loop deals with Thermo Fisher and Eli Lilly, both US firms, positioning NVIDIA as the link between wet labs and AI models. Rivals span three continents: UK-based Basecamp Research's EDEN, Boltz Lab, Owkin's OwkinZero and Edison Scientific's Kosmos are each building competing foundation-model platforms for biology.
The clearest sign of the shift sits in Denmark. Novo Nordisk, Europe's most valuable pharmaceutical company, closed its internal cell therapy unit and licensed a Parkinson's disease asset to the AI-driven Cellular Intelligence platform.3 The move accompanied strong core earnings and a 24.9% 30-day stock gain.3 Investors read it as additive outsourcing, not retreat.
The pattern is global: national pharma champions from Denmark to the US are ceding early-stage discovery work to a small cluster of AI-native infrastructure providers, most incorporated in the US and UK. Target identification, once run inside company R&D departments regardless of country, is migrating to platforms licensed across borders.
For international drugmakers, the calculus is shifting from building proprietary discovery stacks to renting access to whichever foundation-model layer proves most predictive. NVIDIA's expanding partnerships and MindWalk's patent filings suggest that layer is consolidating fast — and that the geography of pharma R&D is becoming less about where a company is headquartered and more about which AI platform it licenses.
