Zenithon raises $10M to build world models for extreme physics

UK AI startup Zenithon has raised $10 million to build what it describes as the first world models for extreme physics and accelerate the development of technologies that will define the 21st century. Its backers include BACKED, Lunar, Seraphim, MMC, and SOSV, alongside founders and hyperscaler directors.

Zenithon is developing “world models” — AI models that learn how complex physical systems behave and can rapidly predict the outcomes of different designs — for technologies including fusion reactors, rockets, and semiconductor manufacturing.

The company argues that these technologies will help shape the 21st century alongside AI, but that developing them increasingly depends on understanding and controlling highly complex physics. According to the company:

“Building the next generation of these technologies requires engineers to understand and control increasingly complex physics.

However, existing tools struggle to keep pace: when a single simulation or experiment takes days, even the best engineering teams can explore only a fraction of the possible designs.”

Zenithon says its models will be able to explore a million design points in the time it takes to simulate one, while also learning directly from real-world experiments.

The company says this could enable engineers to search beyond familiar designs, discover new levels of performance, and continually improve how their machines operate.

Co-founded by Alex Higginbottom and Abetharan Antony, Zenithon brings together researchers behind foundational breakthroughs in machine learning for physics, along with a broader team drawn from leading laboratories worldwide.

Zenithon’s advisers include Charlie Songhurst, Board Director at Meta and former Head of Strategy at Microsoft; Dan Brunner, co-founder and former CTO of Commonwealth Fusion Systems, which has raised around $3 billion, and a former research scientist at MIT; and Zongyi Li, Assistant Professor at NYU.

The funding will be split between recruitment and compute, enabling Zenithon to release new models every three months and expand its team in San Francisco and across the US.