Samsung Ventures and Applied Ventures Back Flexcompute’s Physics World Model
The funding will support research, GPU capacity and development of physics-based simulation tools for advanced engineering and semiconductor design.
Flexcompute, the physics company, announced a strategic investment from Samsung Ventures and Applied Ventures, LLC, the venture capital arm of Applied Materials, Inc., among other investors. The funding will accelerate Flexcompute's work on the Physics World Model: AI that predicts how any design will behave in the real world by learning the laws of physics themselves, from a chip package warping under heat to a spacecraft reentering the atmosphere. Details of the investment were not disclosed.
Today's AI learns from the past: text, images, simulation results. Flexcompute takes a different approach. Its team of more than 70 PhDs has built CFD, electromagnetic, structural and thermal solvers from the ground up on GPUs. Through Tokenised Physics, Flexcompute builds both neural networks and first-principles physics directly into the Physics World Model, so every prediction obeys the same conservation laws as the physical world and stays accurate on designs, even ones the model has never seen.
"AI is changing how products are designed, but those products still have to obey the laws of physics," said the representative of Samsung Ventures. "Flexcompute is building AI that understands those laws, not just the data. We believe the Physics World Model will shape how the next generation of hardware is developed."
The approach is already solving problems that traditional simulation cannot reach. For a full advanced semiconductor package, with more than 800,000 solder joints, an ASIC, and 32 optical engines, Flexcompute solved a 5-billion degree-of-freedom (DOF) warpage model within 2% of commercial finite-element results in minutes.
"Every industry is racing to put AI to work on real engineering problems, and the limit is how quickly each one can be validated against the laws of physics," said Yu Cao, Chief AI Officer of Applied Materials. "Flexcompute has built something unique: simulation fast enough to keep pace with AI and rigorous enough for production decisions. In advanced manufacturing, that means more of the design cycle happens digitally, saving time and cost."
Customers access the Physics World Model today through Physics Twins: models of their own products that engineers and AI agents can question, test and optimise before anything is built. Engineers set the objective and approve the result, agents run the iterations, and physics governs what works.
"Language models learned the world from text. Hardware is governed by laws, not text, and no amount of data substitutes for them," said Vera Yang, Co-Founder and President of Flexcompute. "We are building the verification layer between bits and atoms: a Physics World Model any engineer or AI agent can ask whether a design will work, before it is built. Samsung and Applied see what we see: the next leap in AI comes from understanding the physical world."
Flexcompute will use the investment to grow its research team, scale GPU compute, and extend the Physics World Model across fluids, electromagnetics, structures, heat, and multiphysics couplings.
