OpenAI GPT-5 Automates Science Experiments, Cuts Protein Costs 40% in 3 Runs (Ginkgo Bioworks)

OpenAI and Ginkgo Bioworks have integrated GPT-5 with an automated cloud lab to design, execute, and analyze experiments in cell-free protein synthesis (CFPS), reducing production costs by 40% and reagent costs by 57% in three iterations. The system, operating as a closed-loop AI research pipeline, ran over 36,000 reactions across 580 plates in about two months, completing the optimization in roughly 90 days. The collaboration leverages GPT-5’s data analysis, hypothesis generation, and design capabilities, iteratively refined by high-throughput automation. This marks a significant acceleration in scientific discovery and optimization in biotechnology and materials science, with implications for drug development, diagnostics, and industrial enzymes. Global governments and enterprises are racing to scale AI for Science: the U.S. launched the Genesis Initiative under the Office of Science and Security Bipartisan AI Platform, and China has prioritized AI+-driven scientific innovation in its 14th Five-Year Plan and local policy frameworks. Venture funding for AI for Science startups reached over $8 billion in the past year, with major investments from NVIDIA and other industry leaders.