Basecamp Research has raised $140 million to expand biological AI models trained on genetic material collected from rainforests, oceans, volcanic ground and other underrepresented habitats. S32 led the round, joined by Nvidia, Anthropic’s Anthology Fund, the NATO Innovation Fund, Redalpine and others.
Public genome repositories are heavily concentrated: Basecamp says five species provide about 68% of sequence volume, with humans alone accounting for roughly 54%. Its network works with local researchers across more than 30 countries and records environmental context alongside long DNA sequences. The resulting data trains EDEN models intended to design antibiotics and enzymes that insert therapeutic genes.
Early results remain preclinical. In a non-peer-reviewed paper, 97% of a small, curated group of generated antimicrobial peptides showed laboratory activity. The company says one antibiotic candidate performed about as well as a last-resort drug in infected mice, while half of tested gene-insertion enzymes were active in human cells. These results do not establish safety or effectiveness in patients.
None of Basecamp’s six published therapy programs has moved beyond lead optimization, which precedes preclinical studies and human trials. The financing will support cell therapies and other candidates, but delivery, toxicity and immune responses remain unresolved. The next evidence must come from increasingly demanding experiments, not model benchmarks alone.