Boltz PBC Launches with $28M to Democratize AI Platforms for Drug Discovery
Boltz, a public benefit corporation spun out of MIT CSAIL, has raised $28 million to turn the Boltz series of biomolecular AI models into a practical platform for scientists. Instead of building a drug pipeline, the company is focused on infrastructure-giving researchers end-to-end tools to move from hypothesis to a human-ready molecule on a laptop.
The founders-Gabriele Corso, PhD (CEO), Jeremy Wohlwend, PhD, and Saro Passaro-come from the lab of Regina Barzilay, PhD, and are carrying an open science mindset into the company's product strategy. The goal: make high-accuracy structure prediction, binding affinity estimation, and therapeutic design accessible and usable inside real workflows.
What the Boltz models have achieved so far
In late 2024, Boltz-1 reached AlphaFold 3-level accuracy in predicting 3D structures of biomolecular complexes. That momentum continued with Boltz-2 for binding affinity prediction and BoltzGen for therapeutic design across multiple modalities. Today, Boltz reports more than 100,000 scientists across thousands of biotechs are using the models to speed up discovery.
If you've followed structure prediction progress, you know why this matters. Accuracy that rivals leading systems like AlphaFold 3 is now showing up inside tools built for day-to-day use, not just benchmarks.
From open research to product: why form a company
The team realized that pushing biomolecular AI forward at the frontier needs sustained investment in compute, data access, and talent-beyond what academia can consistently support. They also saw that publishing models isn't enough if adoption stalls at setup, cost, or usability.
So Boltz is taking the open research it's known for and turning it into reliable, supported products that plug into existing scientific workflows. It's a bet that broader access-not exclusivity-creates the most value for the field.
Introducing Boltz Lab
Alongside the launch, Boltz previewed Boltz Lab plus initial agents for small-molecule discovery and protein design. The emphasis is on speed to insight and operational simplicity.
- Lower compute cost with scalable infrastructure
- Interfaces that fit how scientists actually work (and collaborate)
- End-to-end flow: from therapeutic hypothesis to a candidate that's ready for downstream testing
For teams tired of stitching together disconnected tools, this approach is a relief. The promise is fewer engineering detours and more time on decision-making.
Collaboration with Pfizer
Boltz has signed a multi-year collaboration with Pfizer. Using Pfizer data, Boltz will build exclusive models for structure prediction, small-molecule affinity, and biologics design.
Boltz scientists will work with Pfizer's discovery teams to develop custom models and workflows across selected programs to improve preclinical decisions. It's a practical testbed for the platform's scalability and impact.
Funding and who's backing it
The $28 million seed round is led by Amplify, a16z, and Zetta Venture Partners, with angel investors including Clement Delangue (Hugging Face), Factorial Capital, and Obvious Ventures. Zetta's Dylan Reid underscored their focus on AI-native tools and infrastructure-and the idea that the best tools are the ones that become widely used.
Why the open science stance matters
Many bio startups inherit a closed, asset-first culture. Boltz is taking a different path-one closer to software and machine learning norms, where open releases accelerate shared progress and downstream customization.
Expect foundation models to remain central, but the practical value will come from end-to-end systems that are reliable, scalable, and easy to run. Most scientists won't want raw models; they'll want dependable tools that fit into their lab stack.
What this means for researchers and R&D leaders
- Benchmark on tasks that reflect your pipeline: complex formation accuracy, affinity ranking, and cross-modality design.
- Check data provenance, fine-tuning controls, and versioning for regulated environments.
- Evaluate throughput, queue times, and cost predictability-especially if you're running batch projects.
- Look for integrations with ELN/LIMS, audit logging, and collaboration features your team will actually use.
- Plan for model drift and continuous evaluation with held-out assays or prospective tests.
Quick takeaways
- Boltz PBC launches with $28M to productize the Boltz model family for real-world discovery workflows.
- Boltz-1, Boltz-2, and BoltzGen have already seen broad adoption and strong accuracy signals.
- Boltz Lab aims to reduce compute friction, make collaboration easier, and compress time from idea to candidate.
- A multi-year Pfizer deal will pressure-test exclusive models and custom workflows in active programs.
- Open science remains core, but the focus is shifting from raw models to end-to-end systems scientists can trust.
If you want to explore the academic roots behind the team, start with MIT CSAIL. And if you're upskilling your team on applied AI for research, you can browse practical course maps by role here: Complete AI Training.
Your membership also unlocks:
