Seven AI-native drug discovery platforms launched within months of each other in early-to-mid 2026, all built on NVIDIA GPU infrastructure supplied to partners including Thermo Fisher and Eli Lilly.1 BioNeMo, MindWalk's HYFT and ReefIQ, Boltz Lab, Owkin, Natera, Basecamp Research and Edison Scientific each run GPU-heavy models against biological data at scale.
MindWalk gave the first public demonstration of ReefIQ in July 2026.1 The platform runs on HYFT Technology, built from roughly 660 million biological patterns encoding relationships between sequence, structure and function.3 The company's bet: accurately representing biology, not running the models, is the harder problem worldwide — and the place to build a durable edge.3
MindWalk backed that bet with a new patent. EP26187897.9, filed in July 2026, adds a computational layer to its foundational HYFT patent (EP3881326A1), organizing biological data for reuse across infrastructure and customer programs globally.2 The company calls it an extension, not a re-filing.2
The shared target across these platforms is data fragmentation, a problem that spans labs on every continent. Biological data from drug discovery sits scattered across incompatible files, formats and systems.1 Before AI can reason about a biological question, MindWalk argues, that data must be reconnected into one governed, queryable structure — the task now driving demand for NVIDIA-class GPU capacity from Copenhagen to Boston to Shanghai.
Established pharmaceutical players are responding by moving capability outward. Denmark's Novo Nordisk posted strong first-quarter 2026 earnings and a rising share price while restructuring its own R&D — closing its internal cell therapy unit and licensing the asset to Cellular Intelligence, an AI-native platform, rather than continuing the work in-house.4 The move suggests even top-tier global pharma now prefers external, GPU-powered AI specialists over maintaining internal discovery pipelines.
