Bunsen AI Co-Scientist: Schrödinger's New Drug Discovery Engine

How Bunsen Combines LLMs with Physics-Based Simulation

Bunsen pairs large language models (LLMs) with Schrödinger's core physics-based molecular modeling — including FEP+, docking, and molecular dynamics (MD) simulations . The LLM layer interprets natural-language scientific goals and plans workflows, while the physics engine provides rigorous, experimentally validated predictions of molecular properties. As Dr. Karen Akinsanya described, this pairing unlocks previously unexplored biological "target space" by combining AI reasoning with first-principles computation .

The Four Core Functions

Bunsen's workflow centers on a closed-loop cycle:

1. Understand — Interprets the researcher's scientific objective in natural language.
2. Strategize — Develops a computational strategy, selecting the right physics-based methods for the problem.
3. Execute — Runs the molecular discovery workflow autonomously (docking, free-energy perturbation, ADME prediction, etc.).
4. Iterate — Analyzes results and refines the approach, closing the loop between AI reasoning and physics-based outcomes .

This design allows researchers to direct multi-step computational campaigns using natural language, with Bunsen handling the workflow automation and physics-based computation.

Key Competitors in the AI-for-Science Space

Google DeepMind — Co-Scientist

Co-Scientist is a multi-agent AI system built on Gemini 2.0, designed for structured hypothesis generation and scientific reasoning . Published in Nature in May 2026 , it uses seven specialized AI agents (Generation, Ranking, Evolution, Proximity, etc.) that debate and refine hypotheses. It has demonstrated the ability to propose novel therapeutic hypotheses — for example, drug repurposing for liver fibrosis — that were experimentally validated . The key difference from Bunsen is that Co-Scientist is a general biomedical hypothesis-generation engine; it does not natively integrate physics-based molecular simulation. It reads literature and generates ideas but relies on external tools or human researchers for computational chemistry execution .

FutureHouse — Robin

Robin is a multi-agent AI system for end-to-end biological discovery, also published in Nature in May 2026 . It automates the full cycle of hypothesis generation, experimental design, data analysis, and insight generation in one continuous workflow . Robin demonstrated its capabilities by autonomously identifying ripasudil (an existing glaucoma drug) as a promising therapy for dry age-related macular degeneration (dAMD) . It uses specialized sub-agents: Crow (literature Q&A), Falcon (deep synthesis), Owl (answering), and Phoenix (coding/analysis) . The key difference from Bunsen is that Robin is focused on experimental biology and drug repurposing, not computational chemistry, and lacks integrated physics-based molecular simulation engines .

Other Notable Competitors

Q1 2026 Financial Context

Schrödinger reported Q1 2026 results on May 5, 2026 with the following highlights:

• ACV (Annual Contract Value): $28.4 million in Q1, up 12% year-over-year; trailing four-quarter ACV was $201 million . Growth was driven by usage scale-ups with top-20 pharma customers and new product deployments .
• Total Revenue: $58.6 million, down 2% year-over-year (beating consensus by approximately 22%) .
• Software Revenue: $35.6 million, down 21% year-over-year, directly attributed to the accelerated shift from on-premise licenses to hosted (SaaS) delivery .
• Hosted Mix: Hosted software reached 34% of software revenue in Q1, up from 24% a year earlier . The company targets approximately 75% hosted by 2028 .
• Drug Discovery Revenue: $23 million, more than doubled (+124%) year-over-year .
• FY 2026 Guidance: Software ACV expected between $218–228 million (10–15% growth); drug discovery revenue between $55–65 million .

Management framed the hosted transition as a near-term headwind to recognized revenue but not to ACV or cash flow. Each 1% increase in hosted mix impacts reported revenue by roughly $2–3 million due to ASC 606 timing . The shift is intended to align revenue recognition with customer consumption and drive higher retention.

Cash Position

The available earnings summaries did not explicitly state a specific cash and equivalents figure for Q1 2026. However, Schrödinger has consistently maintained a strong balance sheet to fund ongoing platform development (including Bunsen) and its therapeutics pipeline. For precise Q1 2026 cash, cash equivalents, and marketable securities, the 10-Q filing would be the definitive source. In Q4 2025, Schrödinger reported robust liquidity, and management has expressed confidence that the hosted transition will not require additional capital .