At SOSV HAX, we believe the next generation of AI breakthroughs won’t come from ingesting more internet text. Instead, they will come from bringing unique data from the physical world into foundational models.
Plasma physics is one of the most exciting examples: it’s an area genuinely ready for AI modeling, but doing it right requires deep domain experience in extreme physics paired with rich, specialized data…not just more compute. Plasma is the fourth state of matter, and is a powerful technology that underpins critical infrastructure of the future. It is the key to enabling fusion energy and fabrication of advanced semiconductor chips that power data centers and are key unlocks for Physical AI. Scientists’ ability to control plasma in fusion and semiconductor applications used to be a very Edisonian, trial-and-error process. But over the past three to four years, AI and machine learning have evolved to allow researchers to move from guessing and testing everything one by one to a world where product developers are modeling, calculating, predicting, and controlling behavior of plasma. These breakthroughs lower the cost and improve the success of each experiment and are fundamentally speeding up the pace of innovation tremendously.
That’s why we’re thrilled to back Zenithon’s $10M Seed round, alongside BACKED, Lunar Ventures, Seraphim Space, MMC, and other funds and angels.
Zenithon is building the first world models for extreme physics, and specifically for plasma physics. By creating spatial and temporal representations of complex physical systems, their models run one million times faster than existing simulations on a single GPU. Nothing like this exists for the kinds of physics that they are tackling across the fusion energy, semiconductor manufacturing, and aerospace industries.
Today, design cycles across these areas remain bottlenecked by slow, expensive simulation and experimentation. As the demand for Physical AI accelerates, infrastructure like Zenithon’s is a key enabler of it, including for fabricating advanced chips using plasma, with lower latency, higher bandwidth, and lower energy consumption than today’s processes allow. Zenithon’s world models offer a step-change: exploring a million design points in the time it currently takes to explore one.
This investment is a natural extension of what we’re building at the HAX Plasma Forge alongside our partnership with the Princeton Plasma Physics Laboratory (PPPL). We’re excited to collaborate with Zenithon to leverage existing data on plasma — both low- and high-temperature regimes — and generate new, targeted data where gaps remain.
Co-founders Alex Higginbottom (CEO and PhD in Artificial Intelligence) and Abetharan Antony (CTPO and PhD in Plasma and High-Temperature Physics) have spent their careers at the intersection of AI for complex systems, fusion energy, and commercialization, and couldn’t be more well suited to build this foundational layer for the technologies that will define the next century.
Read more about Zenithon in their recent post and featured in Sifted: