AI for Science & Engineering Platforms
- Exponential Industry
- Technology Area
- 6 Milestones
Cloud platforms that run long scientific and engineering AI loops—simulation, materials or drug discovery, agentic EDA, and closed-loop wet labs. A typical stack couples a frontier model to a robotic lab or HPC solver and iterates faster than a human campaign. Examples include DeepMind AlphaFold, OpenAI×Ginkgo autonomous CFPS, and NVIDIA/Synopsys agentic EDA on cloud GPUs.
Hidden Factory
AI for Science & Engineering Platforms
Synopsys–AMD–Microsoft Agentic AI EDA Collaboration
Announcement of autonomous agentic AI chip design workflows developed with Microsoft, available on Microsoft Discovery; AMD evaluating for next-gen products. First EDA applications for evaluation on Discovery; builds on prior agentic specification-to-RTL workflow. Showcased at 2026 DAC.
OpenAI and Ginkgo announce GPT-5 autonomous-lab CFPS result
GPT-5 closed-loop with Ginkgo cloud lab (RAC + Catalyst); >36,000 CFPS compositions, 580 plates, 40% protein-cost reduction ($422/g vs $698/g sfGFP). Closed-loop GPT-5 + Ginkgo cloud lab tested >36,000 CFPS compositions across 580 plates; $422/g sfGFP vs $698/g prior SOTA (40% reduction; 57% reagent-cost improvement per OpenAI) /OpenAI GPT-5 lowers the cost of cell-free protein synthesis/ /PR Newswire / Ginkgo Bioworks Ginkgo Bioworks' Autonomous Laboratory Driven by OpenAI's GPT-5 Achieves 40% Improvement Over State-of-the-Art Scientific Benchmark/.
Nobel Prize in Chemistry 2024 for AlphaFold2 protein structure prediction
half the prize jointly to Demis Hassabis and John Jumper for protein structure prediction; AlphaFold2 used to predict virtually all ~200 million identified proteins /Nobel Prize Outreach / KVA Press release: The Nobel Prize in Chemistry 2024/.
AlphaFold Protein Structure Database expanded to 200 million structures
AFDB expanded to over 200 million predicted structures with EMBL-EBI /Google DeepMind AlphaFold — Google DeepMind/.
AlphaFold methods published in Nature and code open-sourced
Nature paper 2021-07-15 and open-source code release /Google DeepMind AlphaFold — Google DeepMind/ /Nature Highly accurate protein structure prediction with AlphaFold/.
AlphaFold2 recognised as solution to protein-folding problem (CASP14)
2020-11-30 CASP14 — AlphaFold2 predicted structures to atomic accuracy (median RMSD95 < 1 Å), three times more accurate than the next-best system /Google DeepMind AlphaFold — Google DeepMind_/.