Dealmaking & Partnerships: Who's Actually Writing The Checks in AI x Pharma

Author: Chris Leiter (moderator), Jonah Cool, Sarah Nam, Eli Berlin, Stacy Feld Collection: UNLOCK-2026-Sessions

Summary

This panel brings together five leaders to examine how pharma-AI deals are actually being structured in 2026, covering perspectives from a frontier AI lab (Anthropic), big pharma business development (AbbVie, Johnson & Johnson), and an AI-first biotech (Terra Therapeutics). The moderator is Chris Leiter, Founder & General Partner at Atria. Panelists are Jonah Cool (Life Sciences, Anthropic), Sarah Nam (VP, Global Head of Partnerships AI, AbbVie), Eli Berlin (Co-Founder, Terra Therapeutics), and Stacy Feld (Global Head, External Scientific Innovation, J&J).

The panel opens with the provocative question of whether Anthropic is becoming a pharma company, given its Coefficient Bio acquisition and the addition of Vas (from Novartis) to its PBC board. Jonah Cool clarifies that Anthropic's goal remains pushing Claude's limits through deep partnerships rather than becoming a drug developer. Sarah Nam describes how AbbVie stood up an entirely new AI-focused business development team roughly one year before the panel, now independently assessing 400–500 companies per year across the full pharma value chain — from AI drug discovery biotechs to hyperscaler partnerships to physical AI and manufacturing. She explains that AbbVie's governance for AI deals is materially different: a centralized "AI External Opportunities" committee chaired by the Chief Business Officer, with representation from Chief R&D, Chief Commercial, and Chief Operating Officers, to provide momentum and speed for deals that traditional functional teams would not greenlight.

Eli Berlin describes Terra Therapeutics' evolution over its 7.5 years: originally founded on proprietary high-throughput chemistry data (now 16 billion precise target-ligand binding measurements, growing ~1 billion per quarter), pharma partners historically engaged only on targets with no existing chemical equity. In the last 6–12 months, the paradigm has reversed: pharma now wants to leverage Terra's foundation model of chemistry (trained on 1 billion+ data points) across programs where they do have existing data. Stacy Feld describes J&J's "External Scientific Innovation" function, which operates in four global regions, embeds ML engineers and computational chemists alongside traditional BD transaction leaders, and has an active protein design innovation challenge with 18 finalists in partnership with Nvidia.

The discussion covers emerging deal archetypes (model co-development, fine-tuning, data sharing, tech enablement, platform licensing), limitations of current AI (long-horizon agentic planning, causal reasoning, low-data regimes), and the non-negotiable role of human trust and relationship-building in partnerships.