How a former DeepMind researcher raised at a $300M pre-seed valuation before launching a product

Andrew Dai spent over a decade at DeepMind contributing to some of the more consequential AI research of the past few years - work he says informed the development of systems like ChatGPT. That pedigree appears to have been enough to convince investors to back his new venture at a $300 million pre-seed valuation, before any product has been released or even publicly detailed.

The raise is a striking example of a dynamic that has become increasingly common in AI: valuations driven less by traction or revenue and more by the perceived credibility of the founding team. For investors, betting early on researchers with deep technical backgrounds and proven track records has started to feel like the closest thing available to a low-risk position in a fast-moving field.

Dai's focus is on visual AI - the capacity for AI systems to understand, interpret, and generate visual information in sophisticated ways. He has described this as one of the next major frontiers in the broader AI landscape. Generative image and video models have advanced considerably over the past two years, but the harder problems of visual reasoning, spatial understanding, and real-world perception remain largely unsolved, and that gap represents a large surface area for new research and product development.

What Dai's company will actually build has not been fully disclosed, but the fundraise itself signals that at least some major investors believe the visual AI space has room for a well-resourced new entrant. Whether a $300 million pre-seed valuation can be justified by whatever product eventually ships is a question that remains open - but for now, the bet is on the person, not the product.