Arzeda

Eighteen years in, Arzeda's flagship commercial win is a stevia sweetener measured in tens of tons — impressive as a proof point, modest as a business, and a reminder that de novo protein design has been slow to produce breakout commercial value despite enormous scientific prestige. The competitive ground has also shifted dramatically underneath the physics-based approach: the AI protein-design wave — Baker's own Institute for Protein Design outputs (RFdiffusion), EvolutionaryScale, Cradle, Profluent, Generate, and a dozen well-funded generative-model startups — has commoditized parts of the design workflow and attracts capital at a scale Arzeda's $38M round cannot match. Selling designed molecules as products means competing directly in the food and industrial-ingredients markets against giant incumbents (ingredient majors, flavor houses, and the fermentation-scale players) on cost and reliability, where a boutique design shop's advantage in the lab evaporates against commodity manufacturing economics. RebM specifically competes with extraction-based and fermentation-derived stevia from established suppliers, so Arzeda must win on taste and cost against entrenched supply. The DARPA award is validating but small and mission-specific. The deepest risk is strategic drift: a company that designs, validates, and manufactures across food, ag, and materials is spread thin, and eighteen years of runway with one signature product suggests the model converts scientific brilliance into commercial scale slowly — a liability now that faster, better-funded AI-native rivals are racing the same design space.