Zenithon builds machine learning models that learn how extreme physical systems behave, from the plasma inside a fusion reactor to the shocks around a hypersonic vehicle and the conditions in a semiconductor fab. Engineers in these fields depend on simulations that can take days per run, so they only ever test a small slice of the designs open to them.
Its world models are trained on a mix of simulation output and real experimental data, and the company says they can explore a million design points in the time a conventional solver handles one. Zenithon builds the whole stack itself, from data generation and model architecture through training, inference and delivery, and plans a new generation of models every three months.
The company was formally founded in July 2025 by Alex Higginbottom (CEO) and Abetharan Antony (CTO), and came through the first cohort of King's College Cambridge's SPARK incubator. In September 2026 it raised £7.5M ($10M) in a seed round led by BACKED, with Lunar, Seraphim, MMC and SOSV taking part. It has a team of 11, with research in London and a delivery team being built in San Francisco.