Ex-OpenAI, DeepMind Scientists Raise $300M for AI-Driven Science
Former AI Leaders Launch Scientific Automation Venture
Periodic Labs, a new startup founded by prominent artificial intelligence researchers, announced today it has raised $300 million in seed funding from top-tier investors including Andreessen Horowitz, NVIDIA, and tech luminaries like Jeff Bezos and Eric Schmidt.
The company was co-founded by Ekin Dogus Cubuk, former leader of Google Brain and DeepMind's materials science teams, and Liam Fedus, ex-OpenAI research VP who contributed to ChatGPT's development. Their team comprises veterans from major AI projects including Microsoft's MatterGen and OpenAI's Agent Operator.
Building the Lab of the Future
Periodic Labs plans to revolutionize scientific discovery through fully automated laboratories where AI-driven robots will:
- Design and execute physical experiments
- Continuously collect and analyze data
- Refine hypotheses through machine learning
The company's first major project focuses on developing advanced superconducting materials that could outperform current technologies while requiring less energy. "We're not just creating better materials," Cubuk explained. "We're building a system that generates fundamentally new scientific knowledge."
Data: The Next Frontier for AI Models
The founders argue that today's large language models have largely exhausted publicly available internet data. Periodic Labs aims to create a new category of training data through:
- Physical experiments conducted by robotic systems
- Structured scientific observations unavailable online
- Proprietary material discoveries
"Current AI learns from humanity's past discoveries," Fedus noted. "Our systems will generate tomorrow's discoveries while creating the training data for future models."
Competitive Landscape in Scientific AI
The push toward automated science has gained momentum since 2023, with several organizations exploring similar concepts:
- Tetsuwan Scientific (Japan)
- Future House (US)
- University of Toronto's Accelerate Alliance
However, Periodic Labs distinguishes itself through its founders' pedigree in both materials science (Cubuk developed DeepMind's GNoME system that discovered 2+ million new crystals) and large-scale AI systems (Fedus led OpenAI's trillion-parameter network project).
The $300 million seed round—exceptionally large for a pre-revenue science venture—reflects investor confidence in translating theoretical AI breakthroughs into tangible scientific progress.
Key Points:
✅ $300M seed funding secured from top tech investors
✅ Founders bring combined expertise from DeepMind & OpenAI
✅ Autonomous labs will research superconducting materials
✅ Aims to generate new scientific data for future AI training
✅ Part of growing movement toward automated scientific discovery
