Video generation platform Luma AI is preparing for a major expansion in London

Video generation platform Luma AI, a $4 billion startup backed by Nvidia, is preparing for a major expansion in London with plans to hire about 200 workers (~40% of the workforce). The Palo Alto-headquartered budding tech firm will start operations across research, strategic development, partnerships, and engineering at its new base in London by early 2027.

Luma said the UK is just the starting point of the expansion due to its access to large talent pools. The AI startup is using its video models to target marketing, entertainment, media, and advertising sectors. It currently sells the video models as part of a content creation suite through an API (Application Programming Interface).

Luma AI also emphasized that it is building world models that can learn from video, images, text, and audio. Google’s Gemini and OpenAI’s ChatGPT are powered by similar large language models (LLMs). 

Luma seeks to spread world-scale AI everywhere

The co-founder and CEO of Luma AI, Amit Jain, touted that his company has the capital and capacity to bring world-scale AI to creatives worldwide. He added that launching in Europe and the Middle East is the next logical step after the Humain-led $900 million Series C funding round and the anticipated global compute infrastructure build-out.

“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.” 

–Amit Jain,  Co-Founder and CEO of Luma AI

Jain also added that London has the best researchers, given the Universities there and institutions like DeepMind. He noted that London is the entry point to European markets. 

The Luma executive said that these kinds of visual models are only one to one and a half years behind LLMs, but world models may soon become the natural interface for most day-to-day AI use. He emphasized the importance of world models in the pursuit of achieving AGI (Artificial General Intelligence), although he also acknowledged that they were not as developed as LLMs. 

Meanwhile, Nvidia, Google, and Meta are all developing world models for various use cases. Luma also released its Ray3 model in September, which Jain claims is almost at par with Google’s Veo 3 and ranks higher than OpenAI’s Sora. 

Jain says the latest funding will support scaling efforts

The Luma boss emphasized that the latest funding will be used to scale and accelerate the startup’s training and deployment of world models. He further claimed that these world models are more effective than LLMs in helping in the real physical world. 

Meanwhile, Luma and Humain will also collaborate to build a 2-GW AI supercluster called Project Halo in Saudi Arabia. Jain says the buildout will be among the largest deployments of GPUs globally. He added that his company is following in the footsteps of tech giants, which have invested heavily in training large AI models using supercomputers. 

The CEO of Humain, Tareq Amin, also stated that his company’s investment in Luma and its 2-gigawatt supercluster positions it to train, deploy, and scale multimodal intelligence to its full potential. Amin added that the partnership sets a new standard for how compute, capital, and capability merge. 

The Humain CEO said the partnership also includes Humain Create, which creates sovereign AI models trained using Arabic and the region’s data. Jain added that Luma models and capabilities will be deployed to businesses in the Middle East, along with building the first-ever Arabic video model. 

Jain noted that countries outside the U.S. and Asia are typically underrepresented in AI-generated content, as most models are trained using data scraped from the internet. However, he believes it is essential to incorporate these cultures and their diverse representations into Luma’s model.

If you're reading this, you’re already ahead. Stay there with our newsletter.