Global Expansion and Talent Acquisition
Video generation platform Luma AI, a startup valued at $4 billion and supported by Nvidia, is gearing up for a significant expansion in London. The company intends to hire approximately 200 new employees, representing about 40% of its current workforce. Operations at the new London base, which will encompass research, strategic development, partnerships, and engineering, are slated to commence by early 2027.
Luma AI has identified the UK as a strategic starting point for its expansion, citing its access to substantial talent pools. The artificial intelligence startup is leveraging its advanced video models to target the marketing, entertainment, media, and advertising sectors. Currently, Luma AI offers these video models as part of a content creation suite accessible via an Application Programming Interface (API).
The company is also actively developing "world models" designed to learn from diverse data inputs, including video, images, text, and audio. This approach is similar to the large language models (LLMs) that power prominent AI systems like Google's Gemini and OpenAI's ChatGPT.
Spreading World-Scale AI Globally
Amit Jain, co-founder and CEO of Luma AI, expressed confidence in the company's capital and capacity to deliver world-scale AI to creatives worldwide. He stated that launching in Europe and the Middle East represents a logical progression following the company's $900 million Series C funding round, led by Humain, and its ongoing 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 highlighted London's exceptional research talent, attributing it to the presence of leading universities and institutions like DeepMind. He views London as a crucial entry point for accessing European markets.
According to Jain, visual models are currently one to one and a half years behind LLMs, but he anticipates that world models will soon become the primary interface for most daily AI interactions. He stressed the significance of world models in the pursuit of Artificial General Intelligence (AGI), while also acknowledging their current developmental stage compared to LLMs.
Major technology firms such as Nvidia, Google, and Meta are also actively developing world models for a variety of applications. In September, Luma AI released its Ray3 model, which Jain claims is comparable to Google's Veo 3 and superior to OpenAI's Sora.
Funding to Support Scaling and Development
Jain emphasized that the recent funding will be instrumental in scaling and accelerating the startup's efforts in training and deploying world models. He asserted that these world models offer greater effectiveness than LLMs in practical applications within the physical world.
In parallel, Luma AI and Humain are collaborating on "Project Halo," a 2-gigawatt AI supercluster planned for Saudi Arabia. Jain indicated that this project will represent one of the largest GPU deployments globally, aligning Luma AI with the strategic investments made by tech giants in training large AI models using supercomputing infrastructure.
Tareq Amin, CEO of Humain, stated that the company's investment in Luma AI and the development of the 2-gigawatt supercluster will enable the training, deployment, and scaling of multimodal intelligence to its full potential. Amin further commented that this partnership establishes a new benchmark for the convergence of compute power, capital, and capability.
The collaboration also includes Humain Create, an initiative focused on developing sovereign AI models trained using Arabic language data and regional information. Jain added that Luma's models and capabilities will be made available to businesses in the Middle East, alongside efforts to build the first-ever Arabic video model.
Jain pointed out that countries outside the U.S. and Asia are often underrepresented in AI-generated content, as most models are trained on data scraped from the internet. He believes it is crucial to integrate diverse cultures and their representations into Luma's models.