Elorian AI, a new lab founded by former Google DeepMind researcher Andrew Dai, has raised a $55 million seed round at a $300 million valuation despite having no product on the market, betting that models built for native visual reasoning can outpace today's vision-language systems.
Seed Round · Elorian AI
$55M raised at a $300M valuation — before shipping a single product
Founded by ex-DeepMind researcher Andrew Dai, the lab is betting that models built for native visual reasoning can outpace today's vision-language systems — closing the round on pedigree and vision alone.
$300M
Pre-product valuation
~5 mo
From leaving Google to closing the deal
Valued at 5.5× the money in — on vision alone
No product · no revenue · no benchmarks. The valuation is built entirely on founder pedigree and thesis.
The thesis: images as things to reason over, not describe
Today's vision-language models translate images into text first. Elorian wants to skip that step.
Today's VLMs
Image → text → reason
vs
Elorian's "visual AGI"
Reason directly on the image
Targeted at hard-to-describe tasks — spatial relationships, physical constraints, design intent — across robotics, engineering, medicine, agriculture and satellite imagery .
Why investors bit
~12 years across Google Brain & DeepMind
Led Gemini data efforts; co-led PaLM 2 pretraining
Backed by Striker, Menlo, Altimeter — plus NVIDIA and Jeff Dean personally
What's untested
No product, no user or developer feedback
No disclosed models, benchmarks or pricing
Native-visual-reasoning claims unproven vs working systems
Investors are willing to bankroll deep research bets on multimodal AI well ahead of commercialization — wagering the gap between language and visual reasoning is both real and lucrative to close .
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…