Basecamp Research raises $140m in a Series C led by S32, with new and existing tech and pharma investors, after its $60m Series B led by Singular at the end of 2024.
Co-founder Oliver Vince says Basecamp's Eden models are GPT-4-scale, the largest biological models ever trained, and that an antibiotic designed with Anthropic's Claude works in the lab and early preclinical studies.
Vince expects Basecamp to be treating patients within a couple of years, while building its own infrastructure and partnering with pharma where manufacturing and hospitals already exist.
Basecamp Research raises $140m to teach AI the language of biology
Basecamp Research is betting that AI can learn biology the way language models learn language, given enough data and scale. Instead of building a separate model for each drug discovery step, it trains one large model on proprietary genomic data and checks its outputs in the lab. The 97% lab success rate on designed antibiotics is the early evidence, and cancer and genetic medicines are the harder goals it points toward.
Takeaways
Basecamp raised $140m Series C led by S32. The round follows a $60m Series B led by Singular at the end of 2024. New and existing investors across tech and pharma joined, including NVIDIA and Anthropic through their investment arms. Roche vice chairman Andre Hoffmann is among the longest supporters.
One general biology model beat stage-by-stage approaches. Rivals such as Isomorphic Labs and EvolutionaryScale build separate models for each drug discovery stage. Basecamp trained one large model on biology, showed it pathogen and antibiotic pairs, and its designed antibiotics worked in the lab 97% of the time.
Eden models are the largest biological models trained. Vince says the Eden models are GPT-4-scale. With Anthropic, Claude works with Eden to design an antibiotic from a patient's microbiology report. It works in the lab and early preclinical studies, though hospital deployment faces barriers.
Proprietary data from global expeditions is the foundation. Data comes from partnerships in around 30 countries and more than 200 locations, across all seven continents. The Trillion Gene Atlas with NVIDIA, Anthropic, PacBio and Ultimate Genomics will expand the dataset about 100 times, on track for the end of 2027.
Strategic investors fill gaps in the technical chain. NVIDIA and Anthropic let Basecamp apply their expertise to the middle technical stage instead of relearning it. Basecamp focuses on data collection and on the pharma side of making medicines. Vince wants investors who are both realistic and optimistic.
Cancer is the long-term target, but not yet reached. Each tumour is individual, complex and a moving target, so only an algorithm that devises its own solutions will reach many patients. The aim is to teach the immune system to shut down a DNA mistake. The full patient-to-safety pipeline is not yet built.
Remaining risks are tactical and patients may come within years. Vince calls the risks mostly tactical, such as affording the scale needed and how big a model must be to write 20, 30, then 40 DNA letters. He expects to treat patients within a couple of years, with no promised timelines.
Tap a name to go there · tap a line to see why they are linked