Luma AI Raises $900 Million to Build Supercluster

Luma AI has successfully raised nine hundred million dollars in its latest funding round, which was led by Humain, an AI company backed by Saudi Arabia's Public Investment Fund. The financing also saw participation from Advanced Micro Devices' venture arm, alongside existing investors such as Andreessen Horowitz, Amplify Partners, and Matrix Partners.

This funding round, announced during the U.S.-Saudi Investment Forum, has boosted Luma's valuation to over four billion dollars. CEO Amit Jain highlighted that Luma is focused on developing multimodal 'world models' that learn from text, video, audio, and images to simulate reality more effectively than conventional large language models.

With these funds, the company aims to scale its operations and accelerate the training and deployment of these advanced models. Luma recently launched Ray3, a pioneering reasoning video model capable of generating videos, images, and audio from prompts.

Jain stated that Ray3 currently outperforms OpenAI's Sora 2 and is competitive with Google's Veo 3. Humain, established in May, aims to enhance Saudi Arabia's standing as a global AI hub. Under the leadership of Tareq Amin, a veteran in the tech industry, Humain is set to partner with Luma to construct a two-gigawatt AI supercluster named Project Halo in Saudi Arabia.

This supercluster is expected to be one of the largest GPU deployments globally. Major players in the tech industry have been investing heavily in supercomputing infrastructure to train massive AI models.

For instance, Meta recently announced plans for its own one-gigawatt supercluster, Prometheus, while Microsoft has deployed a supercomputing cluster using Nvidia's GB300 NVL72 platform. Amin emphasized that this partnership between Luma and Humain combines capital, computational power, and capabilities to push the boundaries of multimodal intelligence.

Additionally, the collaboration will include the initiative Humain Create, aimed at developing sovereign AI models using Arabic and regional data. Jain underscored the importance of representing diverse cultures and identities in AI-generated content, especially since many existing models predominantly rely on data from the U.S. and Asia.

Luma's text-to-video platform, Dream Machine, had previously faced accusations of copying intellectual property, but Jain assured that the company has implemented safeguards to prevent misuse of its technology.

He noted that they are continuously enhancing their systems to detect and mitigate any attempts to exploit the platform's capabilities.