Nvidia-backed video generation startup Luma AI is joining a growing wave of U.S. tech companies launching operations in the U.K., with major plans for a London expansion revealed on Tuesday.
The Palo Alto-headquartered startup will look to hire around 200 employees — making up around 40% of its workforce — at its new London base by early 2027, across research, engineering, partnerships and strategic development.
The expansion comes two weeks on from Luma announcing a $900 million funding round led by Saudi Public Investment Fund-owned AI company Humain, which saw it hit a valuation upwards of $4 billion. The startup previously received backing from Nvidia.
Luma is building “world models,” a class of AI models that are able to learn from video, audio and images, alongside text, and which large language models (LLMs) like those powering OpenAI’s ChatGPT and Google’s Gemini use.
The startup is currently targeting marketing, advertising, media and entertainment sectors with its video models, which it sells via an application programming interface (API) and as part of a content creation suite.
“With this Series C raise and the upcoming build-out of global compute infrastructure, we have the capital and capacity to bring world-scale AI to creatives everywhere,” said Amit Jain, CEO and co-founder of Luma AI. “Launching across Europe and the Middle East is the logical next step in putting this power directly in the hands of storytellers, agencies and brands globally.”
The U.K. is the starting point of the expansion because of its access to talent, Jain told CNBC.
“London has some of the best people when it comes to research, given the universities here and institutions like DeepMind,” he said. “We also consider London to be the entry point to the European market.”
AI generated image created by Luma’s Ray3 model (Luma AI)
Luma AI
Luma is the latest in a wave of North American AI labs doubling down on the U.K. and Europe as they look to take advantage of talent pools and revenue opportunities.
In November, San Francisco-based Anthropic announced plans to open offices in Paris and Munich, months on from kicking off a hiring spree in London and Dublin. Canadian AI startup Cohere said it would open a Paris office to become its EMEA headquarters in September and OpenAI announced a new office in Munich in February.
While world models may not yet be as developed as LLMs, some researchers say they are as, if not more, crucial in the pursuit of achieving artificial general intelligence (AGI).
“These kinds of visual models are about a year to a year-and-a-half behind language models right now,” said Jain.
But world models will become the “natural interface” for AI for most day-to-day use in time, he predicted, pointing to the amount of time people spend watching video content each day.
Tech giants including Google, Meta and Nvidia are all developing world models for a range of use cases.
Luma released its latest model, Ray3, in September, which Jain told CNBC benchmarks higher than OpenAI’s Sora and at similar levels to Google’s Veo 3.