Kevin Weil launches Warp Sciences with $215 million to build robot chemistry labs in the US

What Warp Sciences announced

Kevin Weil, who until recently led OpenAI's science efforts, has launched a start-up called Warp Sciences with $215 million in funding. The company emerged from stealth on 22 October 2026 with a plan to build a US-based chemistry contract research organisation, or CRO, that combines expert chemists, AI and robotics to make the molecules drug researchers need.

Bloomberg Law reported that Lightspeed Venture Partners led the round and that it values Warp at $865 million after the new money, according to the company. Lightspeed describes it as Warp's Series A. IVP, Premji Invest, Lux, Mayfield and Menlo Ventures are among more than a dozen other investors.

Warp is careful to say what it is not. It has no proprietary drug pipeline. "We're not developing drugs of our own. We're building the infrastructure for the people who do," said DJ Patil, Warp's chief operating officer.

What a chemistry CRO does

Drug discovery runs on a loop. Scientists design a molecule, chemists make it, biologists test it, and the team uses the results to design the next one. Finding one promising drug candidate can take years and thousands of compounds, and each turn of the loop depends on someone physically synthesising new molecules.

Much of that synthesis is outsourced to contract research organisations. Biotech start-ups in particular often lack large chemistry teams, so they send designs to a CRO and get compounds back weeks later. Warp's website argues that this foundational chemistry has largely moved overseas and still depends heavily on manual work.

A faster, more reliable CRO shortens the loop for every customer that uses it. That is the business Warp wants to build, rather than discovering medicines itself.

The first service: analog synthesis

Warp's initial service is lead optimisation analog synthesis. Once researchers have a promising molecule, called a lead, they make many variations of it, called analogs, to see which changes improve potency, safety or how the body handles the drug. It is repetitive, high-volume work that is well suited to standardisation.

The company says its goal is to shorten the path from molecular design to testing while giving customers more control over where sensitive research takes place. For a start-up whose value rests on a handful of molecules, knowing exactly who is handling its chemistry and where can matter as much as speed.

Chemists first, robots later

Warp is not claiming a fully automated lab on day one. Its launch materials say experienced chemists will guide experiments as the lab is increasingly automated by AI and robotics. Lightspeed, in its own post, says expert chemists will do much of the work early on, with AI tools helping them plan how to make each compound and pick the next experiment.

Over time, robots are meant to take over repetitive, high-volume steps such as setting up reactions and purifying samples, leaving chemists to work on the hardest problems. Warp's careers page lists openings for chemists, robotics hardware engineers, robot operators and AI agent engineers.

TechCrunch reported that Warp's human chemists will wear cameras and sensor gloves so their work can be used to train robots to perform lab tasks, and that the aim is a lab that can run around the clock. Early hires, according to the company, include robotics staff from Genesis, Skild AI and 1X, and scientists from Genentech and Lawrence Livermore National Laboratory.

The data argument

The bet underneath is about data. Warp says its ambition is to make discovery compound: experimental data helps train better AI, better AI powers more capable labs, and scientists gain the capacity to test more ideas. Mano Paluri, the chief technology officer, said every experiment should make the next one smarter.

Lightspeed spells this out. It says Warp records every step of every run in a form a model can learn from, including the steps that fail. That data is used to run the lab better, by showing which steps are slow or unreliable, and to choose what to make next, by training models that predict how to make a compound and which compounds are worth making.

Failed reactions are the interesting part. Published chemistry mostly records what worked, so models trained on the literature see few failures. Because Warp's own robots run the experiments, it keeps the negative results most labs never capture. AI-for-science work has been moving towards this kind of closed loop; OpenAI recently bought Orchestra Bio to add drug-programme agents to Rosalind, its life-sciences model.

Bringing the work back from China

Warp is also pitching itself on geography. TechCrunch notes that biotech companies have for years outsourced much of their labour-intensive molecule synthesis to contractors in China such as WuXi AppTec, and that Weil wants to reshore that work and reduce exposure to geopolitical supply-chain risk.

Warp's release cites a 2024 Biotechnology Innovation Organization survey, reported by Reuters, in which 74% of respondents said they had contracts with Chinese companies for preclinical and clinical services. It also points to the BIOSECURE Act, signed into law in December 2025, which set up a framework for restricting federal procurement and funding involving designated biotechnology companies of concern.

That policy backdrop gives US drug developers a reason to look for domestic options, even if Chinese CROs remain cheaper and far larger today. Warp's argument is that automation can close the cost gap that pushed the work abroad in the first place.

Who is behind it

Weil was most recently OpenAI's chief product officer and then vice president of science, where he led the company's push into AI for research. Warp's other founders are Sol Reisberg, the chief scientific officer, who held medicinal chemistry leadership roles at Terremoto Biosciences; Manohar Paluri, a former vice president at Meta Superintelligence Labs and a founding member of Meta's FAIR research group; and DJ Patil, a founding executive at Devoted Health and the first US chief data scientist in the White House.

The angel list reads like a who's who of AI. Warp's site and release name, among others, OpenAI chief financial officer Sarah Friar, OpenAI chief research officer Mark Chen, Anthropic Labs co-lead Mike Krieger, Coinbase chief executive Brian Armstrong, Figma chief executive Dylan Field, Databricks co-founder Ion Stoica, Skild AI chief executive Deepak Pathak, former Tesla robotics vice president Milan Kovac and former Johnson & Johnson chief executive Alex Gorsky.

What we do not know yet

Warp has not said where its labs are, how many chemists it employs, when it will take its first customers or what it will charge. It has not named customers or published turnaround times, success rates or any comparison with existing CROs.

Lab automation has a long history of promising more than it delivered, because chemistry involves solids, sticky liquids, air-sensitive reagents and messy purifications that are hard for robots to handle. Warp's choice to start with human chemists suggests it knows this. How quickly robots can take on real work, and whether the data loop produces better models, are the questions that will decide whether an $865 million valuation holds up.

Why it matters

AI companies have spent the past year arguing that models can speed up science, from OpenAI running thousands of GPT-Rosalind agents over a brain atlas to drugmakers such as Lilly signing AI deals for drug delivery. The limit, increasingly, is the physical experiment. A model can propose a thousand molecules in minutes, but someone still has to make and test them.

Warp is betting that the lab, not the model, is now the bottleneck, and that whoever owns automated wet labs will also own the data needed to train the next generation of scientific AI. With $215 million and one of the best-connected founding teams in AI, it now has to show that robots and chemists in the US can make molecules faster and more reliably than the established overseas labs.