2022
200 million structures
The predicted-structure database closed a half-century folding problem and became shared infrastructure for biology.
Branches · SCIENCE
Autonomous laboratories
The model proposes a molecule, a robot immediately runs the assay, the result flows straight back into the model — and the loop repeats without people. What took months now takes days.
A closed discovery loop. For the first time experiments run at software speed, and the data is machine-readable from the start.
Status
TypeBranches
Marker
What exists today8
Researched
2024
AlphaFold 3
Predicting complexes rather than lone proteins: protein with DNA, ligand, ion.
2026
2.2M experiments per week
Autonomous labs compress the hypothesis cycle from months to days.
2026
AI-discovered drugs in Phase II
Xaira, Generate Biomedicines and Isomorphic Labs moved from pipeline-filling to clinical candidates.
In progress
сейчас
In silico first
Target discovery starts on a computer and only then moves to the wet lab — the reverse of classical pharma.
сейчас
Climate models 25× faster
Generative methods fused with physics-based data run climate models about 25× faster.
Planned
в планах
Clinical proof
No AI-discovered drug has reached a pharmacy yet. Phase III and approval are the real test of this branch.
Distant horizons
впереди
Materials on demand
The same loop for alloys, catalysts and batteries: specify properties, receive a recipe.
Branches off
To the line