Warp Sciences Raises $215M Series A for robot wet labs that turn AI drug ideas into molecules
Kevin Weil, who led science at OpenAI, launched Warp Sciences on October 22 with a $215 million Series A at an $865 million valuation. Lightspeed led, with IVP, Premji Invest and Lux joining. The co-founders include DJ Patil, Manohar Paluri and Sol Reisberg. Weil announced it on X: https://x.com/kevinweil/status/2113294657673466060
The company is going after the slow, manual part of drug discovery: actually making and testing the thousands of compound variations that models and researchers propose. Biotech has outsourced most of that synthesis to contract labs in China such as WuXi AppTec for years. Warp wants to bring it back with robots that run around the clock. The training data comes from people: the startup's chemists work wearing cameras and pressure-sensing gloves, and the robots learn lab tasks from those recordings. TechCrunch: https://techcrunch.com/2026/10/22/former-openai-science-lead-launches-warp-science-to-automate-wet-labs-with-robots/
This is the bottleneck every AI-for-science story eventually runs into. Models can now propose molecules, materials and experiments faster than any lab can run them, so the speed limit has moved from the idea to the bench. Whoever owns a lab that can close the loop, propose, synthesize, measure and feed the result back to the model, gets the compounding. A pure software agent can't.
The round size tells you investors believe the physical loop is where the defensible value sits, not another model. It is also a reshoring bet with a geopolitics premium baked in. The open question is speed. Teaching robots chemistry from glove data is a long road, and contract labs in China are fast and cheap today. $215 million buys time to find out whether the automated bench can beat them.
← Back to all articles
The company is going after the slow, manual part of drug discovery: actually making and testing the thousands of compound variations that models and researchers propose. Biotech has outsourced most of that synthesis to contract labs in China such as WuXi AppTec for years. Warp wants to bring it back with robots that run around the clock. The training data comes from people: the startup's chemists work wearing cameras and pressure-sensing gloves, and the robots learn lab tasks from those recordings. TechCrunch: https://techcrunch.com/2026/10/22/former-openai-science-lead-launches-warp-science-to-automate-wet-labs-with-robots/
This is the bottleneck every AI-for-science story eventually runs into. Models can now propose molecules, materials and experiments faster than any lab can run them, so the speed limit has moved from the idea to the bench. Whoever owns a lab that can close the loop, propose, synthesize, measure and feed the result back to the model, gets the compounding. A pure software agent can't.
The round size tells you investors believe the physical loop is where the defensible value sits, not another model. It is also a reshoring bet with a geopolitics premium baked in. The open question is speed. Teaching robots chemistry from glove data is a long road, and contract labs in China are fast and cheap today. $215 million buys time to find out whether the automated bench can beat them.
Comments