US-based artificial intelligence firm Luma AI has secured a $900 million Series C funding round led by Saudi-backed HUMAIN, the AI venture owned by the kingdom’s sovereign wealth fund, the Public Investment Fund. Participation also came from AMD Ventures and existing investors including Andreessen Horowitz, Amplify Partners and Matrix Partners. The funding is part of a broader strategy to advance multimodal artificial general intelligence capable of generating, interpreting and interacting with the physical world.
Luma AI’s valuation following the round is estimated at over $4 billion, underscoring the intense competition among frontier AI firms. The company specialises in “world models” that integrate video, audio, image and language to drive applications in robotics, simulation, advertising, gaming and personalised education. CEO Amit Jain described the aim as training systems on “a quadrillion tokens” of multimedia data to move beyond current large-language-model architectures.
HUMAIN, having positioned itself as a global full-stack AI player, will use the funding partnership to anchor its compute infrastructure ambitions. The joint roadmap includes the rollout of what HUMAIN terms “Project Halo” — a super-cluster targeting up to 2 gigawatts of AI compute capacity in the kingdom. Luma AI will become a key customer of this infrastructure.
The funding and infrastructure build-out reflect a strategic shift in Saudi Arabia’s economic diversification efforts. Under the leadership of Mohammed bin Salman, the PIF is deepening commitments to AI and high-performance computing as pillars of its wider Vision 2030 agenda. Luma AI’s partnership becomes a vehicle for exportable AI capabilities and for strengthening the kingdom’s position in the global tech ecosystem.
From Luma AI’s perspective, the capital infusion and access to sovereign-scale compute represent a strategic lever to accelerate development of next-generation models. In entertainment and creative industries the company already claims traction: its Ray3 reasoning video model has been embedded within major platforms. The new funding will allow expansion into simulation, robotics and immersive media. Jain asserted that only by integrating systems and infrastructure at scale can models “understand and simulate the universe”.
For HUMAIN the deal accomplishes several objectives: securing a marquee US AI partner, anchoring compute commitments, and beginning commercial deployments of its planned AI infrastructure. The first phase of the accompanying joint venture with Cisco Systems and AMD will deliver a 100-megawatt data-centre cluster in Saudi Arabia, fully subscribed by Luma AI, with construction scheduled to begin in 2026 and renewable energy powering the facility.
The global implications are significant. With Asia, Europe, India, the Middle East and Africa identified as target markets-worth an estimated 4.5 billion people-the infrastructure build-out promises to reshape the geography of AI training and inference. HUMAIN’s model seeks to challenge the dominance of US- and China-based AI supply chains.
Challenges remain. Developing multimodal AGI is widely acknowledged as highly complex, with technical risks, high compute costs and uncertain commercial timelines. Critics highlight concerns about geopolitical dependencies in AI infrastructure and the ethical implications of deploying models trained on vast multimedia datasets. The Saudi-US dimension adds further scrutiny given strategic sensitivities around data, export controls and digital sovereignty.
