Aureka Closes $100M Series B, Betting the Wet Lab Belongs Inside the Model

Aureka Biotechnologies closed a $100M Series B on August 10, 2026, capital aimed at expanding the biological foundation models and experimental systems behind its AI-driven drug-discovery platform.

The round was assembled in tranches rather than one clean check. Granite Asia funded the first exclusively. An unnamed strategic investor led a later tranche, with HighLight Capital joining alongside follow-on investment from returning shareholders MPCi and NRL Capital. Aureka did not disclose its valuation or the identity of the strategic investor, leaving a substantial cross-border syndicate and several important terms private.

Aureka sits at the intersection of two normally separate operations: models and laboratories. Its platform joins AI agents, digital biology, proprietary single-cell screening, high-throughput validation and drug-development workflows into one feedback system. Models propose molecules and scientific hypotheses, experiments test them, and the resulting functional data returns to training and project-specific optimization. In that architecture, the lab is not a cleanup crew waiting after inference — it is part of the learning apparatus.

That is the business argument the round is underwriting. Public biological datasets can train capable models at many companies, but proprietary experimental feedback is harder to replicate. The stated technical priorities for the new capital are research and large-scale training for the next generation of biological foundation models, spanning de novo molecular design, biological structure modeling and function prediction, plus an expansion of the company's Lab-in-the-Loop system.

Part of the thesis is publicly inspectable. Aureka's proprietary foundation model is AuraIDE; its open-source counterpart, OpenDDE, is described in its repository as an all-atom biomolecular foundation model for structure prediction, design and optimization — with the caveat that it remains a preview whose interfaces and checkpoints may change. Independent testing reported by Tamarind Bio put OpenDDE v1 at 76.1% DockQ success on FoldBench v1, against 47.9% for previously published AlphaFold 3 results on identical inputs and setup. The benchmark covers antibody-antigen structure prediction, so it supports a claim of technical competitiveness on that task and nothing further: it cannot establish clinical safety, efficacy, manufacturability or regulatory success.

Founder and CEO Weian Zhao comes from both academic science and commercial translation. UC Irvine lists him as a professor of pharmaceutical sciences whose work includes diagnostics, biosensors, microtechnology and cell therapies, with chemistry degrees from Shandong University and McMaster University. Aureka's origin is tied to the gap between promising scientific results and medicines that reach patients.