Elorian AI raises $55M seed

Elorian AI has raised a $55 million seed round as it comes out of stealth to build artificial intelligence models that reason about images and video. (elorian.ai) (bloomberg.com) The Palo Alto startup said Striker Venture Partners, Menlo Ventures, and Altimeter backed the round, with participation from Nvidia, 49 Palms, and AI researcher Jeff Dean. Bloomberg reported the company at a $300 million valuation. (elorian.ai) (bloomberg.com) Elorian was co-founded by Andrew Dai, a former Google DeepMind researcher, with co-founders Yinfei Yang, who worked at Google and Apple, and Seth Neel, a former Harvard professor. Bloomberg reported the company has hired more than a dozen people and is not yet generating revenue. (bloomberg.com) The company is chasing a specific problem in artificial intelligence: current systems are strong at text, but weaker at understanding the physical relationships inside images, diagrams, and video. Elorian says many models still convert visual input into words first and then reason from the text. (elorian.ai) Elorian argues that this two-step method breaks down on tasks that are hard to describe precisely in language, including spatial layouts, engineering constraints, and visual flaws. Its approach is to train models to work directly with visual representations instead of treating images as static prompts. (elorian.ai) (pitchbook.com) That pitch lines up with a broader push across the artificial intelligence industry to move beyond chatbots and into systems that can handle robotics, manufacturing, design, and scientific work. Bloomberg reported Dai said better visual reasoning could help in architecture, automotive work, and robotics. (bloomberg.com) (elorian.ai) PitchBook lists Elorian as founded in 2025 and describes its product as multimodal models that learn spatial relationships, objects, and constraints from images and video. The company says those models could be used in engineering, robotics, medicine, and science. (pitchbook.com) (elorian.ai) Bloomberg reported the seed round was raised in two tranches, with one portion priced at a $120 million valuation and a later portion at $300 million. Dai told Bloomberg the startup plans to release its first public reasoning model in about 12 months while it talks with potential customers. (bloomberg.com) For now, Elorian is selling a research agenda as much as a product: build models that can look at the world more like engineers and less like autocomplete. The $55 million seed gives it time to try. (elorian.ai) (bloomberg.com)