Warp Sciences Charges Drug Makers Only When Robots Synthesize Molecules

Former OpenAI science vice president Kevin Weil has launched Warp Sciences, a chemistry research service intended to replace much human-led analog synthesis in pharmaceutical discovery with robotics, machine learning and automated laboratory infrastructure. The company emerged from stealth on Oct. 22, 2026, announcing a $215 million financing at an $865 million valuation, led by Lightspeed Venture Partners with about 18 other investors. OpenAI Chief Financial Officer Sarah Friar and Chief Research Officer Mark Chen participated as individual investors. The co-founders include Weil, former Terremoto Biosciences medicinal chemist Solomon Reisberg, former first US Chief Data Scientist DJ Patil, and former Meta FAIR and Superintelligence Labs vice president Manohar Paluri. Warp Sciences focuses on lead optimization for clinical small-molecule programs, a stage in which medicinal chemists repeatedly modify candidate compounds, synthesize variants, test them, and use the results to design the next batch. This design-make-test-analyze loop is difficult to accelerate because synthesis often requires multiple organic reactions and skilled human execution. Large pharmaceutical companies typically use internal teams or contract research organizations, while smaller biotechnology companies increasingly relied on outsourced chemistry, including lower-cost Asian service providers. Warp aims to offer an alternative with comparable economics, tighter collaboration, domestic control and quality assurance. The startup plans to charge customers per delivered compound rather than by full-time equivalent chemist hours, refusing payment for molecules it cannot make. Human chemists currently wear cameras and pressure-sensing gloves to record experiments, creating training data for robotic platforms. The company intends to combine advanced frontier models with customized open-weight models, preferring US-developed weights although it notes that many leading open models originate in China. Its pitch is a self-improving data loop: more experiments generate more training data, better models improve robotic reliability, and increasingly reliable automation accelerates synthesis and lowers costs. Weil’s thesis is that AI has advanced fastest in fields with cheap, rapid validation, such as software and mathematics, but scientific discovery in chemistry and biology is constrained by physical experiments with long feedback cycles. Public datasets often omit failed reactions, limiting model learning. Warp’s laboratory service is designed to produce proprietary closed-loop data and eventually enable scientists to specify desired molecular properties while AI designs experiments, schedules robotic work and iterates autonomously toward what the company calls scientific superintelligence. Warp enters a crowded AI-for-science landscape. Large model developers are moving into life-science infrastructure: OpenAI acquired Orchestra Bio and is expanding its Rosalind product line, while Anthropic acquired Coefficient Bio. Automated-laboratory efforts also include Eli Lilly’s collaboration with Nvidia and startups such as Lila Sciences and Chemify. Chinese and Singapore-based autonomous chemistry firms, including ChemLex, present cost and scale competition. US policy pressure on biotechnology supply chains, including the BIOSECURE Act and recent scrutiny of WuXi AppTec, may create near-term demand for domestic chemistry services, but durable advantage will likely depend on whether Warp can deliver molecules faster, more cheaply and more reliably than existing outsourcing networks.