How Autonomous Labs Will Transform Scientific Research: Ginkgo Bioworks’ Jason Kelly

Ginkgo Bioworks CEO Jason Kelly argues AI reasoning models plus robotic labs beat human scientists on a 40% benchmark, and will restructure how all experimental science is funded and done.

• Ginkgo’s OpenAI collaboration ran 30,000 cell-free protein synthesis experiments and beat the Stanford/Jewett lab benchmark by 40% after six iteration rounds.
• The AI won not through superior intelligence but by running experiments 24/7 and sharing raw data across 100 parallel hypotheses daily, which human labs never do.
• Less than 5% of current NIH and biopharma R&D spend goes to reagents; the rest is overhead — autonomous labs could flip that ratio and deliver ~10x more data per dollar.
• China went from under 5% to over 40% of drug discovery deals acquired by big pharma in just three years, driven by more experimentalists at lower cost.
• Ginkgo sold 97 autonomous lab racks to the Department of Energy for Project Genesis, which targets doubling the speed of scientific discovery at national labs.
• Drug development costs have risen every year for 25 years because science is still manual labor; Kelly calls this Baumol’s cost disease applied to the bench.
• Ginkgo scientists now write plain-language protocols that Claude Code or Codex converts into robot instructions, replacing visual programming tools like LabVIEW.
• Kelly believes democratizing lab access — experiments starting at $39 via Ginkgo’s cloud lab — could turn mass consumer curiosity into distributed scientific discovery.

2026-03-24 · Watch on YouTube