Most startups guard their best work. The three researchers behind Boltz posted theirs online for anyone to download, and the world’s biggest drugmakers picked it up.

Now that generosity has investors lining up. Boltz announced a $110 million Series A on November 18, 2026, led by Bessemer Venture Partners. A Series A is usually the first large round a startup raises from professional investors after its seed money, and this one is almost four times the $28 million Boltz raised in January.

Andreessen Horowitz came back for more. So did Amplify Partners, Zetta Venture Partners and Obvious Ventures. The new names are Greycroft, F-Prime Capital, Hummingbird and Benchling, the lab software company.

Three researchers in Regina Barzilay’s lab

The story starts at MIT. Gabriele Corso, Jeremy Wohlwend and Saro Passaro worked in the group of Regina Barzilay, a professor who studies AI and health, inside the university’s computer science lab and its Jameel Clinic.

In November 2024 they released Boltz-1, a model that predicts how proteins and other molecules fold and fit together. It was a foundation model, meaning a large AI system trained on huge amounts of data that other tools can be built on top of. The team said it matched the accuracy of Google DeepMind’s AlphaFold 3.

And they gave it away. Open source, free for anyone.

Boltz-2 came next and added something drug hunters care about a lot: a prediction of how tightly a molecule grips its target. BoltzGen followed for designing proteins. The models have now passed 5 million downloads and are used by more than 100,000 scientists, according to Bessemer’s announcement.

Boltz isn’t alone in thinking software can design biology. Biomatter raised an $18M Series A for AI-designed enzymes and spent it on a lab of its own.

The January decision to leave MIT

Code on GitHub doesn’t answer support tickets or sign contracts with Pfizer.

So on January 8, 2026, the trio turned their academic project into a company, with Corso as CEO. They set it up as a public benefit corporation, a for-profit business legally bound to pursue a stated mission alongside making money. Boltz’s mission is to put better molecular design tools in the hands of every scientist.

The launch came with a $28 million seed round led by Amplify, a16z and Zetta, with Hugging Face CEO Clem Delangue among the backers. It also brought Boltz Lab, a platform for running the models, and a multi-year partnership with Pfizer.

June 16 and the token meter

The open models stayed free. What Boltz started selling on June 16, 2026 was something else.

That day it released two proprietary models, BoltzMol-1 for finding small-molecule drug candidates and BoltzProt-1 for designing proteins, plus an API that lets other software call them directly, including AI coding agents like Claude Code, OpenAI’s Codex and Google’s Gemini. Customers pay in “molecular tokens,” buying AI the way a lab buys reagents, a little at a time as each experiment needs it. The lab results gave buyers a reason to try. In drug discovery, a hit is a compound that shows real activity against a disease target in testing, and finding one usually means screening a mountain of molecules. BoltzMol-1 went after 10 difficult targets and landed confirmed hits on six, with only 28 to 51 compounds tested for each. BoltzProt-1 nearly tripled BoltzGen’s hit rate on nanobody design. Then the meter started running. By August, token use was about eight times the June level. It passed 25 times in late October and reached 50 times on November 8.

Six of the world’s 20 largest drug companies are now partners, Pfizer, GSK, Takeda and Bristol Myers Squibb among them. That pace is why paid usage jumping 50-fold in five months became the headline of the raise.

Wait, giving it away was the business plan?

Pretty much. And look at the order.

Boltz spent the better part of two years becoming a tool scientists already trusted before it asked anyone for money. When the paid models showed up in June, the buyers were already in the building. No cold calls required.

The free models are still free, too. That’s the part I keep coming back to. The company didn’t lock up what made it popular. It built

Laura is a business historian and researcher who writes about how companies began and grew. At Companies History, she works primarily from original sources such as annual reports, regulatory filings, patents, and archived news coverage. Her articles often correct popular myths about famous founding stories.