Overview
Mythic is a semiconductor company pioneering analog compute-in-memory architecture for AI inference. Instead of shuttling data between memory and digital processors, Mythic's Analog Processing Units (APUs) store neural-network weights directly in flash memory and perform computation in place, dramatically reducing power consumption. The approach targets edge AI applications in cameras, drones, robotics, automotive and defense systems where energy and cost budgets are tight.
Funding and Partnerships
Founded in 2012 as a University of Michigan spinout, Mythic has raised roughly $260 million across multiple rounds, including a $125 million Series D led by DCVC with participation from SoftBank, Honda and Lockheed Martin. The company has announced a joint development agreement with Honda for software-defined vehicle chips and is led by CEO Taner Ozcelik, formerly of NVIDIA.
Why It Matters
As AI moves from cloud data centers to the edge, the power and cost of running models locally become the key constraint. Mythic's analog approach promises order-of-magnitude gains in energy efficiency, making always-on AI feasible in battery- and thermally-constrained devices.