The Integration Ladder

Most AI drug discovery companies run one or two parts of a discovery engine well. This is the interactive from my essay Cheap to Run, Expensive to Be Right. Filter the 111 companies by who runs their own lab, tap a bar for the count, and open the table.

The scores come from public evidence only, as of October 14, 2026, and no company is named.

Research note

Of 111 AI drug discovery companies, one passes all four tests.

I scored each company on four tests of an integrated engine, from public evidence. Most pass one or two.

Its own dataGenerates its own data in house
Human biologyLearns from primary human cells, patient samples or genetics
A working loopLab results feed back into the models, with benchmarks
Clinical proofA candidate from the engine has reached the clinic

Tests passed, of 4

Tap or hover a bar.

A company passes a test at level 2 or higher on a 0 to 3 scale. Sources: company sites, papers, filings, trial registries and press, as of 10.14.2026, scored by AI models against a written rubric, with source checks on disputed and top scores. Public disclosure only: a capability a company hasn’t disclosed counts as missing. Not a ranking of companies and not investment advice.

Read the essay: Cheap to Run, Expensive to Be Right