Warp Sciences Raises $215M to Automate Drug Discovery Experiments

Warp Sciences, founded by Kevin Weil, former head of science at OpenAI, has officially launched after securing $215 million (approximately 280 billion won) in Series A funding. According to a Bloomberg report dated October 22, 2026 (local time), the company was valued at $865 million (approximately 1.14 trillion won) following the investment. Lightspeed Venture Partners led the round, with participation from IVP, Mayfield, Lux Capital, Premji Invest, and Menlo Ventures. The co-founding team includes Manohar Paluri, former vice president at Meta's superintelligence research organization; DJ Patil, the White House's first Chief Data Scientist; and Solomon Reisberg, a former executive at Teramoto Biosciences.

What Is Warp Sciences Automating?

Rather than developing its own drugs, Warp Sciences provides research services that synthesize and test substances needed by pharmaceutical and biotech companies. Its first business line is 'lead optimization'—improving the drug-like properties of a promising molecule once it has been identified. This involves synthesizing and testing hundreds to thousands of variant compounds to boost a molecule's efficacy and reduce side effects in the human body. Pharmaceutical companies have long outsourced this work to Chinese firms, and Warp Sciences aims to bring it back to the U.S. by automating the process with AI and robotics.

This market has been dominated by Chinese firms such as WuXi AppTec, which offer low costs and fast turnaround. Weil noted that reliance on Chinese vendors has become a risk factor amid U.S.-China tensions, and that some pharmaceutical companies are uncomfortable sharing sensitive research data and intellectual property with Chinese contractors. To bring this work back to the U.S., the company determined it needed to achieve efficiency capable of competing with Chinese firms, and set out to build a fully automated laboratory.

Where Does the Automation Begin?

Warp Sciences launched with roughly 20 employees and has already secured related hardware, including robotic arms and AI chips. However, not all experiments will be handed over to robots from the outset. The process is divided into stages:

1. Early StageExpert chemists perform the bulk of the work, with AI assisting in determining compound synthesis pathways and the next experiments to run.
2. Subsequent StageRepetitive tasks such as reaction preparation and sample purification will be the first to be automated with robots.
3. Long TermThe goal is to build an environment where robots can run experiments around the clock, reducing the time and cost required for compound synthesis and validation.

The data used to train the robots comes from human work. Researchers will wear devices such as wearable cameras and gloves equipped with pressure sensors during experiments, collecting data from skilled chemists' work to train the company's own AI models—and ultimately training robots to perform a variety of experimental tasks based on that data.

Weil explained that while AI has achieved remarkable results in software development and mathematics, it has yet to do the same in scientific fields where physical experimentation is critical, and that models need exposure to real experimental data. This is why the company is scaling up experiments through automation to gather data, which AI then uses to design the next experiments and automate even more tasks—creating a virtuous cycle. Weil stated, 'The lab advances AI, and AI advances the lab.' In a podcast released in February 2026, he said the goal is to achieve, by 2030, scientific breakthroughs that would otherwise not be expected until 2050.

Implications for AX at Korean Companies

What stands out in this case is not the scale of investment but the sequence of automation. Despite being billed as a 'fully automated lab,' the starting point is a structure in which AI assists expert judgment, the first tasks handed to robots are repetitive ones like reaction preparation and sample purification, and even the data robots learn from is gathered from skilled chemists' work. For manufacturing and research organizations that rely on on-site expertise and are considering automation, the right sequence is to first check whether the work of skilled experts is currently being captured as data—rather than jumping straight to selecting which processes to replace.

A more directly relevant point for Korean companies concerns the relationship between outsourcing and data. The issues Weil raised were the risk of dependency created by U.S.-China tensions, and the discomfort some pharmaceutical companies feel about sharing sensitive research data and intellectual property with Chinese contractors. If a company outsources part of its research, design, or testing work, it would be worth adding to its contract review checklist—not just cost and delivery time, but also where the data generated in that process ends up, and whether that data can be reused to train the company's own AI.

Model selection raises the same question. Warp Sciences plans to use both cutting-edge AI models and its own customized open-weight models, with Weil saying that, being still in an early stage, he would like to use open-weight models developed in the U.S. If a company intends to train models on its own experimental or process data, it needs to decide not only where that data will be stored but also which models to use and where to operate them.

This week, try writing down one step in your most skilled staff member's work that is repetitive yet currently leaves no data trail. Deciding what to record before bringing in robots is the sequence this case illustrates.

Source: Former OpenAI Chief Scientist Sets Out to Build a 'Fully Automated Lab' for Drug Discovery