Elorian在未发布产品前以3亿美元估值融资5500万美元

做的是什么生意

A Palo Alto research lab building AI models that reason natively through images and physical space, aimed at robotics, architecture, automotive and design.

启动资金：$55M seed raised in two tranches, first at a $120M valuation and second at $300M, from Striker Venture Partners, Menlo Ventures and Altimeter Capital with participation from Nvidia and AI researcher Jeff Dean (Economic Times/Bloomberg).

起因

Andrew Dai, a 14-year Google DeepMind researcher who worked on Gemini, co-founded Elorian with Yinfei Yang (Google/Apple AI research) and Seth Neel (former Harvard professor). His thesis: frontier models reason about images 'like a 3-year-old', so he left to build models that think natively in the visual medium.

经过

Elorian emerged from stealth in April 2026 with $55M and a $300M valuation, a dozen-plus employees, customer conversations but no revenue, and a plan to release its first publicly available reasoning model within about 12 months. Dai said he picked a lower valuation to preserve employee equity upside, and described the two-tranche round structure to Bloomberg.

可借鉴之处

顶尖的研究背景可以在产品发布前获得3亿美元的估值，但只有当视觉推理成为真正的先锋领域时，这一赌注才会得到回报；较低的估值保持了员工的股权收益。