Former OpenAI Chief Kevin Weil Seeks $150 Million for Science AI

Key Takeaways

• Kevin Weil is seeking $150 million for his new AI science startup at a valuation of at least $750 million, though terms could change.
• The startup is reportedly focused on gathering scientific data for AI models, extending Weil’s work on scientific applications at OpenAI.
• Periodic Labs and Discovery Loop show how competition is shifting toward autonomous experiments, research workflows, and proprietary scientific data.

Kevin Weil, the former chief product officer of OpenAI, is pursuing one of the more ambitious early-stage financing rounds in the emerging AI-for-science market.

Weil has sought a valuation of at least $750 million and aims to raise $150 million, according to three people with knowledge of the discussions. The financing details have not been finalized and could still change. Weil did not respond to a request for comment.

Little is publicly known about the product or commercial model. Sources said Weil has pitched the startup as a way to gather scientific data for AI models. That description points toward a potentially important part of the AI stack: connecting models with specialized, high-quality information generated through scientific work.

Scientific data is not interchangeable with the text, images, and code commonly used to train general-purpose models. Experimental results can be costly to produce, difficult to standardize, and constrained by intellectual property, privacy, or regulatory considerations. A startup that can systematically collect and structure such data could build a defensible asset, particularly if it also participates in designing or running experiments.

Weil brings a product background rather than a purely academic research profile. He departed OpenAI earlier this year after serving as chief product officer and later overseeing science initiatives. That work included Prism, a workspace intended to bring frontier models into scientists’ everyday workflows.

Before OpenAI, Weil held product leadership roles at Meta and Twitter. He also sits on several boards, including Cisco and Stoke Space. That mix of consumer product experience, enterprise exposure, and connections to aerospace could be relevant if the new startup needs to coordinate software, scientific users, and physical research infrastructure.

The proposed financing reflects investors’ continuing willingness to back experienced AI executives before all the details of a business become public. McKinsey found in 2024 that 65% of respondents said their organizations regularly used generative AI, a sharp increase from the prior year. Enterprise adoption helps explain why investors are looking beyond chat interfaces toward specialized applications with clearer operational value.

Science is an especially attractive, if difficult, target. Deloitte noted in 2024 that generative AI was being applied across drug discovery, trial design, and biomedical research. The commercial promise is substantial: models could help scientists identify candidates, plan experiments, analyze results, and refine subsequent work.

But what happens when an AI system proposes a costly or unsafe experiment? Governance becomes part of the product, not a policy document added later. The NIST AI Risk Management Framework offers one widely recognized structure for identifying, measuring, and managing AI risks. Startups selling into pharmaceutical, industrial, or government research environments will likely face demands for traceability, validation, access controls, and human review.

Competition is already taking shape. Startups such as Periodic Labs are developing AI scientists alongside autonomous laboratories, exploring ways to conduct physical experiments and generate new training data. Similarly, entrants like Discovery Loop are focusing on AI applications in science and engineering. These companies highlight a growing field where talent, computing resources, laboratory access, and proprietary datasets matter as much as model architecture.

The movement also follows a series of senior departures from OpenAI. Brad Lightcap, the company's longtime former chief operating officer, announced that he was leaving after eight years to start something new.

For Weil, the fundraising target sets a high bar before the startup’s strategy is fully visible. It also signals where experienced AI leaders think the next wave may form. Chatbots proved that models could reach mass audiences. Science AI now has to prove that those models can produce reliable knowledge, generate valuable data, and fit into research processes where mistakes carry real operational costs.