Ex-OpenAI Product Chief Kevin Weil Seeks $150M at $750M+ for AI Science Startup

Kevin Weil — the former chief product officer of OpenAI who later led the company’s “OpenAI for Science” initiative — is in discussions to raise $150 million for a new AI-for-science startup at a valuation of at least $750 million, according to multiple reports published on August 11–12, 2026. The fundraising, first reported by Business Insider and subsequently confirmed by Dealroom, Yahoo Finance, and Quartz, marks one of the most ambitious seed-stage capital raises of the year and signals that the intersection of AI and scientific research has become one of the most hotly contested arenas in the technology landscape.

From ChatGPT to the Scientific Method

Weil’s trajectory to this moment is itself a story of how AI’s center of gravity has shifted. He joined OpenAI in June 2024 as chief product officer, bringing a resume that included senior product roles at Instagram and Twitter (now X). In that capacity, he oversaw the consumer-facing explosion of ChatGPT and became one of the company’s most visible public advocates. In late 2025, he transitioned to lead OpenAI for Science, an initiative he described as “building the next great scientific instrument: an AI-powered platform that accelerates discovery.”

The vision was ambitious: Weil argued that 2026 would be for AI in science what 2025 was for AI in software engineering — a year when the technology moved from novelty to transformation. He predicted autonomous AI research assistants by the fall, and publicly pushed for OpenAI to shift its competitive focus from math and coding benchmarks toward genuine scientific advancement.

But on April 17, 2026, Weil departed OpenAI in a dramatic leadership shakeup that also saw Sora lead Bill Peebles and enterprise apps chief Srinivas Narayanan exit the same day. OpenAI folded its science team into Codex and shut down the Sora video project entirely, a move TechCrunch characterized as the company “continuing to shed side-quests.” For Weil, it was the end of one chapter and the beginning of another.

The Startup: Scientific Data as the Bottleneck

According to sources cited by Business Insider and Yahoo Finance, the exact details of Weil’s new venture “could not be learned,” but multiple people familiar with the pitch say he has positioned it as “a way to gather scientific data for AI models.” This framing addresses what many researchers consider the single most critical bottleneck in applying AI to science: not compute, not algorithms, but data.

Frontier AI models have become extraordinarily capable at processing and reasoning over text, code, and images. But scientific domains — chemistry, materials science, biology, physics — suffer from a chronic shortage of structured, machine-readable data. Experimental results are locked in PDFs, scattered across proprietary databases, or simply never digitized. A startup that could systematically collect, clean, and structure scientific data would be providing the raw material that every AI-for-science effort desperately needs.

The concept bears more than a passing resemblance to Discovery Loop, an AI-for-science startup whose founding Weil was associated with during his time outside of OpenAI, though the precise relationship between Discovery Loop and Weil’s current venture remains unclear. Discovery Loop’s stated mission is to “automate the entire discovery process, from generating hypotheses and designing experiments to analyzing results and publishing findings” — effectively building AI systems designed to act as autonomous researchers rather than mere tools for human scientists.

A Crowded and Well-Funded Field

Weil is entering a market that has already attracted enormous capital. The AI-for-science sector has produced some of the largest seed rounds in venture capital history:

• Periodic Labs, an OpenAI spin-out, raised a record $300 million seed round in October 2025, backed by Andreessen Horowitz, DST Global, Nvidia, Accel, Elad Gil, Jeff Dean, and Eric Schmidt. By March 2026, the company was reportedly in deal talks at a $7 billion valuation.
• Lila Sciences, born inside Flagship Pioneering, raised a $235 million Series A in September 2025, later expanded to $350 million, bringing total funding to $550 million and pushing its valuation past $1.3 billion. Lila describes its mission as building “scientific superintelligence” and operates “AI science factories” — autonomous lab platforms that conduct experiments without human intervention.
• FutureHouse, a nonprofit backed by Eric Schmidt, has released AI tools it claims can accelerate literature review, hypothesis generation, and experimental design in biology and adjacent fields.

At a $750 million valuation for a seed round of $150 million, Weil’s venture would immediately rank among the most valuable pre-launch companies in the AI-for-science space — behind Periodic Labs’ rumored $7 billion and Lila’s $1.3 billion, but ahead of most other entrants. Notably, the 43-year-old Weil has yet to publicly name the company or disclose its product, making the valuation a powerful statement about investor confidence in the founder himself.

Why This Matters

The AI-for-science thesis rests on a simple but profound argument: scientific progress is fundamentally constrained by the human bottleneck. A single PhD-trained researcher can run a handful of experiments per week, read dozens of papers, and synthesize findings over months. An AI system with access to the right data could, in principle, run thousands of experiments in parallel, ingest the entire scientific literature, and generate and test hypotheses at a pace orders of magnitude faster.

Weil’s bet on data infrastructure rather than model development or autonomous lab hardware represents a distinct strategic angle. While Lila Sciences builds physical “AI science factories” and Periodic Labs pursues an end-to-end AI scientist, Weil appears to be targeting the foundational layer — the data supply chain that makes all of these approaches possible. If scientific data is the new oil, Weil is positioning his company to be the refinery.

This approach could prove prescient. The companies building AI models and autonomous labs all face the same data scarcity problem, and a platform that solves it could become indispensable infrastructure — the “pickaxes and shovels” of the AI-for-science gold rush. Alternatively, it could face existential questions about whether scientific data can be gathered and structured at sufficient scale to justify a $750 million valuation, especially when much of that data is generated by the very labs that competitor startups are building.

What Comes Next

The fundraising is still in progress, and terms could shift. But with Weil’s pedigree — he has been a board member at Cisco, a global board member of The Nature Conservancy, and one of the most recognizable product leaders in consumer AI — investor interest is likely to be intense. The question now is whether the unnamed startup can deliver a compelling enough product narrative to justify its valuation before the AI-for-science hype cycle peaks.

For OpenAI, the departure underscores a strategic tension: by folding its science team into Codex, the company signaled that its near-term focus remains on software and coding. If Weil’s venture succeeds, it may prove that the biggest opportunities in AI for science belong not to the frontier model labs, but to specialized startups willing to go deep on data, domain expertise, and the unglamorous infrastructure work that makes discovery possible.