Kevin Weil has seen the inside of both the AI industry's biggest labs and the government's biggest data problems. Now he's betting $215 million that the next frontier for AI isn't a bigger model — it's a robot arm in a chemistry lab.
Weil, who ran OpenAI's science push until April, launched Warp Sciences on Thursday with a round led by Lightspeed Venture Partners, valuing the company at $865 million before it has shipped a public product. IVP, Premji Invest, and Lux Capital joined, and the angel list reads like an industry group chat: OpenAI CFO Sarah Friar, LinkedIn co-founder Reid Hoffman, Coinbase CEO Brian Armstrong, and Anthropic executive Mike Krieger, according to Bloomberg's reporting. Weil announced the startup and funding on X, saying the Series A will "bring frontier AI out of the browser and into the lab."
The target is specific and unglamorous, which is exactly the point. After researchers identify a promising molecule, someone has to synthesize and test hundreds or thousands of chemical variants to improve its activity and cut side effects. Biotech companies have outsourced that slog to contract labs in China for years — WuXi AppTec became the go-to on speed and cost. Warp plans to reshore it to the United States with a fully automated lab, where human chemists wearing cameras and pressure-sensing gloves generate the training data for the company's own models. The roughly 20-person team has already bought robotic arms and AI chips, Bloomberg reports.
Weil's pitch to Bloomberg is that this is where the AI-for-science bet was always heading: drug chemistry is "still an area where frontier AI models are not yet changing the world" the way they changed coding and math, "because the field is so much more experimentally driven. The models need to see that data." That, in a sentence, is the whole physical-AI thesis — the next leap comes from instrumenting the real world, not scaling another training run.
The timing isn't accidental either. The BIOSECURE Act, signed into law last December, set up restrictions on federal procurement involving designated biotechnology companies of concern, making reliance on Chinese contractors a board-level liability rather than just a supply-chain convenience, per the company's announcement. Warp is selling geopolitics as a feature.
The founding lineup is engineered for exactly this hybrid. Weil was OpenAI's chief product officer — the teams behind ChatGPT and Codex — then head of OpenAI for Science, the unit that built GPT-Rosalind for chemistry and genomics work, before leaving in April when OpenAI folded its science teams into the core org. His co-founders: Solomon "Sol" Reisberg, who led medicinal chemistry at Terremoto Biosciences; Manohar "Mano" Paluri, a VP at Meta Superintelligence Labs and a founding member of FAIR; and DJ Patil, the first U.S. chief data scientist, as COO.
The open question is the only one that matters: can American robots do chemistry cheaper and faster than WuXi? An $865 million valuation for a company with no customers is investors answering "yes" on faith — faith in Weil, in the data flywheel (more experiments, better models, more experiments), and in geopolitics doing the sales work. But wet labs are where grand automation promises go to die; chemistry is messy, equipment breaks, and every edge case is a new molecule. Watch the first customer contracts — when a pharma company trusts a robot with its pipeline compounds, the valuation stops being a faith statement.
— via TechCrunch