Six of the ten largest pharmaceutical companies already run parts of their clinical development on software from a San Diego startup most of their investors have never heard of. Faro AI raised $37.3 million on August 26 to push that position further.
Its edge is a proprietary clinical development data model that converts study intent into machine-readable structure AI agents can act on.
Six of the world's ten largest pharma companies already use the platform, which makes this a scale-up round anchored by paying customers.
The round was co-led by Merck Global Health Innovation Fund and S32. Existing investors General Catalyst, Northpond Ventures, Polaris Partners, PTX Capital, and Zetta returned, and Ankona Capital joined as a new backer. It will spend the proceeds expanding its agentic capabilities and accelerating deployments inside its customers' development organizations.
That phrase, development organizations, is where the story sits. Discovery attracts the headlines and the billion-dollar model launches. Development is where programs actually stall.
TIMELINE: Faro, from partnerships to Series B
─────────────────────────────────────────────────────────────
Jan 2025 Mar 2026 Jul 2026 Aug 2026
🔬 🧪 🏆 ◉ NOW
Recursion Bristol Fierce AI $37.3M
partnership Myers Squibb Innovation Series B
collaboration finalist
Sources: Faro AI newsroom, 2025-2026.
The bottleneck moved downstream
Clinical development is a chain of thousands of interconnected scientific, medical, regulatory, and operational decisions. Change one eligibility criterion and you disturb site operations, data collection, statistical analysis, regulatory documentation, and every downstream system that depends on them.
For decades the industry managed that complexity with people, spreadsheets, and documents. A protocol runs to a hundred pages. A study build is a manual translation of that artifact into a clinical trial management system. Each handoff loses context and adds weeks.
AI entered this space the obvious way first, as a drafting tool. Hand a language model the protocol template and the prior study, and it produces a faster first draft. That draft is useful, and increasingly commoditized. It is still a document. The work of turning it into an executable study remains.
Structured intent beats smarter documents
Faro's core claim is narrower and more technical than "AI for trials." The company spent years building proprietary clinical development data models that represent the concepts, relationships, constraints, and intent behind a study in a structured, machine-readable form. Agents that sit on that structure can reason across a program instead of answering one document-shaped question at a time.
Scott Chetham, Faro's co-founder and CEO, describes the shift as a change in what counts as the system of record. Documents stay important as outputs. They stop being the substrate that agents reason over.
Faro AI Series B
Co-led by Merck Global Health Innovation Fund and S32, with every existing investor returning and Ankona Capital joining. · Faro AI / Infor Capital, 2026
The practical effect shows up across the workflow. Study design, protocol authoring, study build, execution, data flows, and review all draw on the same underlying model. A change in one place can propagate with its context attached, rather than triggering a round of manual reconciliation.
This financing gives us the resources to accelerate what our customers are already asking us to do.— Scott Chetham, co-founder and CEO, Faro AI
What the money buys
The Series B is small by AI-infrastructure standards and large by clinical-software standards. It buys model development and customer deployment rather than a new category. The customer list is the asset that matters. A platform used by six of the world's ten largest pharmaceutical companies holds production data across therapeutic areas, study designs, and development workflows that a new entrant cannot rent.
Adoption among the largest pharma
The platform is used by six of the ten largest pharmaceutical companies. · Faro AI, 2026
What Faro automates, and what still needs a human
Study build and downstream system configuration can be derived rather than retyped.
Data flows and review steps can be monitored and reconciled by agents.
Oversight stays with development teams. The framing is automation with context, not autonomy without sign-off.
| Layer | Drug discovery | Clinical development |
|---|---|---|
| Primary AI use | Design molecules and proteins | Run documents and workflows |
| Core data structure | Sequences and structures | Study intent, constraints, decisions |
| Bottleneck | Finding viable candidates | Moving a program to approval |
| Buyer | Research leadership | Development operations |
Turning points
The company's path to this round ran through partnerships rather than one breakout product. In January 2025, Faro and Recursion partnered to apply AI to clinical trial design. In March 2026, Bristol Myers Squibb began working with Faro to scale autonomous AI across clinical development, including protocol design. In July 2026, it was named a finalist in the Fierce AI Innovation Awards. The Series B followed in August.
Mike Morgan, a principal at Merck Global Health Innovation Fund, framed the investment around reliability rather than novelty. Agentic AI, he said, has the potential to change how the industry conducts clinical development by saving time and improving quality, and Faro has built a foundational platform for applying it to increasingly complex workflows.
What could go wrong
Enterprise deployment inside large pharma is slow, and a handful of accounts carry the revenue.
The underlying language models are commoditizing, so defensibility has to live in the data model and the workflow integrations, not the model.
Regulated workflows invite validation scrutiny, and an AI error inside a clinical program is expensive.
The layer that owns the intent
If agentic AI compresses trial timelines, the question is where the value lands. Model providers capture attention. The company that owns the structured representation of a development program owns the context every agent needs, and context is the scarce input.
A nearby pattern makes the point. As we wrote in September, the first AI-engineered antibody reached Phase 3, a milestone that moved AI from molecule design into clinical reality. Faro is betting on the layer between that milestone and an approval.
Whether Faro turns six large pharma accounts into multi-program, multi-year contracts.
Whether incumbent clinical-software vendors ship comparable agent infrastructure. Together they decide if the structured-intent layer becomes a toll road or a feature.