FRIDAY, NOVEMBER 20, 2026 Tech news, minus the noise.
N News Cast
LATEST
Startups

Boltz Raised $110 Million to Sell Drug Discovery by the Token

Boltz, the MIT spinout behind open-source molecular AI models, raised a $110 million Series A led by Bessemer. Usage of its paid platform jumped 50-fold in four months, and six of the top 20 drugmakers are already customers.

Boltz Raised $110 Million to Sell Drug Discovery by the Token
Source: National Institutes of Health (public domain)

A $110 million bet on molecules as a metered service

Boltz raised a $110 million Series A this week, led by Bessemer Venture Partners with Andreessen Horowitz participating, and the round says as much about how AI companies sell their work as it does about drug discovery. Boltz is an MIT spinout building foundation models for biomolecular design, and it sells its models the way a cloud provider sells compute: by the token. Usage of its paid platform jumped roughly 50-fold in the last four months, after the release of its BoltzMol-1 and BoltzProt-1 models.

The company launched from stealth in January with a $28 million seed round co-led by Andreessen Horowitz, Amplify Partners, and Zetta Venture Partners. Less than a year later, it has paying customers at six of the top 20 pharmaceutical companies, including Pfizer, GSK, Takeda, and Bristol Myers Squibb, plus DSM Firmenich. Its open-source models, which came out of MIT's CSAIL and Jameel Clinic, have been downloaded more than 5 million times and are already in use at every top 20 pharma company, according to the investors.

What the Boltz Series A actually buys

Drug discovery is slow and expensive in a way that makes most software problems look trivial. Finding a molecule that binds to a disease target, then proving it is safe and manufacturable, takes years and can cost hundreds of millions of dollars per candidate. Boltz's pitch is that its models shrink the earliest, most failure-prone part of that process. The company says BoltzMol-1 cuts the time to a validated hit from months to weeks, and the cost from hundreds of thousands of dollars to as low as $10,000.

Those are company claims, not independent results, and drug discovery has heard big promises before. What makes this round different is the customer list. Pfizer signed a multi-year collaboration in January to train exclusive models on its proprietary data, covering structure prediction, binding affinity, and biologics design. When the largest drug companies in the world pay for a tool and keep using it, that is stronger evidence than any benchmark.

The Series A money goes into research, compute, and data to build more capable models, plus the commercial platform around them. Boltz is structured as a public benefit corporation, which fits a company whose founders came out of academic research and still publish open models alongside the paid product. The open models are the marketing and the recruiting tool; the paid platform, sold by molecular token, is the business.

Why investors are paying up for biotech AI

The funding context helps explain the size of the round. This month alone, Mirendil raised $1 billion for AI research agents and LangChain raised $300 million for agent infrastructure. Capital is flowing to the teams building the picks and shovels of the AI economy, and drug discovery is one of the few places where the payoff for better models is measured in billions of dollars and years of human life.

Boltz is also part of a smaller, more specific wave: open-source models that become real businesses. Its first model, Boltz-1, released in November 2024, was presented as the first fully commercially usable model to reach AlphaFold 3-level accuracy on 3D structure prediction. Giving the base model away built trust and distribution; charging for the hosted platform, the newest models, and the agents built on top is how that distribution turns into revenue. It is the same playbook as the open-source infrastructure companies, applied to biology.

The risks are real, and so is the signal

The honest risks are the same ones facing every AI-for-science company. Models that predict molecular structures well in benchmarks can still fail on the molecules a drug company actually cares about. The 50-fold usage jump is impressive, but it comes off a small base, and the company has not disclosed its valuation or its revenue. An SEC filing in October showed $97.4 million raised toward a $120 million offering, which suggests the round was assembled over several weeks rather than closed in a single day.

None of that changes the signal. A company that was a research project two years ago now has a $110 million war chest, the biggest drugmakers as customers, and a pricing model its users already understand. If Boltz can keep its models ahead of the open-source alternatives it helped create, the token meter keeps running. If it cannot, its own open models become the competition. Either way, the next year of drug discovery just got a lot more interesting, and a lot cheaper to start.

AS
News Cast Editorial Team

We cover AI, gadgets, startups and security — original reporting and explainers, written by humans, for humans. Got a tip? Reach us via the contact page.