Arena Physica Collaborates with AMD to Train Next Generation of Heaviside, the First Foundation Model for Electromagnetism

Arena Physica will scale Heaviside on the Lux AI Supercomputer, powered by AMD hardware

NEW YORK, Nov. 12, 2026 /PRNewswire/ -- Today, Arena Physica announced it will work with AMD to train the next generation of Heaviside, its foundation model for electromagnetism, on the Lux AI supercomputer at Oak Ridge National Laboratory.

Heaviside-1, released in September, is a 350 million parameter model trained on 250,000 3D field simulations spanning 500 billion individual field samples. The first foundation model for electromagnetism, Heaviside learns from full-wave electromagnetic simulations. Each sample solves Maxwell's equations across a detailed 3D model of a structure. Scaling model performance requires a significant amount of compute, both for generating this simulation data and for training the model itself, a challenge Arena Physica researchers wrote about in Practical Scaling Laws.

"As chips move to 3D packaging and SerDes speeds climb from 224 to 448 gigabits per second, electromagnetic behavior has become one of the hardest problems in semiconductor design," said Kumaran Siva, Corporate Vice President of AI Enterprise Strategy at AMD. "Heaviside is exactly the type of ambitious work Lux was built for. With AMD Instinct GPUs and an open ecosystem, companies like Arena Physica can bring a new class of model to systems like Lux, and we're excited to support them as they scale the first foundation model for electromagnetism."

The next generations of Heaviside will extend to far-field radiation patterns, more complex geometries, and wider frequency ranges, scaling up to larger components and full systems. Applications include antenna design for 5G and 6G devices and satellite communications, RF sensing in automotive systems, and packaging and signal integrity in semiconductors, including high-speed SerDes design. Across these domains, Heaviside is designed to predict electromagnetic behavior orders of magnitude faster and cheaper than traditional solvers, and enables autonomous inverse design.

"Heaviside is a new class of foundation model that directly predicts electromagnetic fields. These fields are highly sensitive to the details of a design, from small local features to the structure as a whole. Capturing that complexity takes a lot of high-fidelity data and a purpose-built model architecture," said Michael Frei, Chief Research Officer at Arena Physica. "That's why we are excited to collaborate with AMD to train Heaviside with Lux. Lux brings physics, AI, and compute together, enabling us to push the frontier of physics foundation models and train at a far larger scale than was possible before."

In the months ahead, Arena Physica will share results as training on Lux progresses.

About Arena Physica

Arena Physica is the electromagnetic superintelligence company building Heaviside, the first foundation model for electromagnetism. Founded in 2019 and based in New York City, Arena Physica works with the world's most advanced hardware companies including AMD, Anduril, Bausch + Lomb, Mercedes-Benz, and Sivers Semiconductor. To learn more, visit arenaphysica.com.

SOURCE Arena Physica