From Language Models to Physical AI: A New Investment Magnet
Physical AI funding refers to capital flowing into AI startups that build systems able to understand, simulate, and act on the physical world, going beyond text or image generation to encompass design, manufacturing, robotics, and autonomous decision-making in real environments. After years of hype around language models, investors are turning to embodied AI systems and world models AI that connect software intelligence to hardware, factories, and machines. This shift is reflected in record AI startup valuation figures for companies promising infrastructure for real-world interaction. Prometheus is building tools to design and manufacture complex objects, while Odyssey develops models that predict how people, objects, and environments behave over time. Together, they point to an emerging consensus: the next wave of AI value is likely to come from systems that not only talk about the world, but also model it and help control it.
Prometheus: An “Artificial General Engineer” with Record-Scale Capital
Prometheus has raised USD 12 billion (approx. RM55.2 billion) in Series B funding at a USD 41 billion (approx. RM188.6 billion) valuation, in what co-CEO Jeff Bezos calls a “capital-intensive startup” due to compute and specialized training data needs. The round includes JPMorgan, BlackRock, Goldman Sachs, DST Global and Arch Venture Partners, with Bezos again participating after being the largest backer of the earlier USD 6.2 billion (approx. RM28.5 billion) Series A. Prometheus is building an “artificial general engineer” aimed at compressing the journey from design to manufacturing for physical objects, from chips to bridges. Co-CEO Vik Bajaj describes tasks like jet engine design—normally a decade-long effort for large teams—as now possible to “formulate… as an end-to-end AI problem.” This focus squarely places Prometheus at the core of embodied AI systems, where software guides the full lifecycle of real-world engineering.
Odyssey and the Race to Build General World Models
Odyssey’s USD 310 million (approx. RM1.42 billion) Series B at a USD 1.45 billion (approx. RM6.67 billion) valuation underlines rising investor belief that world models AI is a new foundation layer for robotics AI development. Led by Natural Capital, with participation from Amazon, AMD Ventures, GV, EQT and others, the company develops models that simulate how people, objects and environments interact over time. CEO Oliver Cameron argues that “world models represent a new class of foundation model — AI that can understand and simulate the world itself,” aiming for a “GPT-3 moment” in this domain. Strategic alignment with Amazon Web Services, including use of Trainium AI chips, highlights how compute providers see these workloads as among the most demanding in AI. Odyssey’s projects—from physics-based simulation in Odyssey-2 Max to multi-agent environments in Agora-1—show a roadmap toward generalized simulation engines for robots and autonomous systems.
Why Investors Are Betting Big on Embodied and Physical AI
These funding rounds signal a strong pivot from pure software interfaces toward AI systems that plan, design and act within the physical world. With Prometheus, investors are backing a platform that could automate and accelerate high-stakes engineering workflows, potentially reshaping how factories, infrastructure, and hardware are conceived and built. Odyssey’s world models aim to give robots, autonomous vehicles and scientific tools a persistent understanding of cause and effect, enabling safer decisions and more reliable autonomy. Top-tier participation—from large asset managers in Prometheus to strategic investors like Amazon and AMD Ventures in Odyssey—shows that physical AI funding is seen as critical infrastructure, not a niche bet. The shared message is clear: whoever controls the foundational models that tie AI to real machines and environments could sit at the center of the next decade of AI startup valuation growth.
