What Mythic does
Mythic is an AI hardware company building analog compute-in-memory silicon and an associated software toolchain to run neural-network inference at the edge and in other power-constrained settings. The company’s core architectural approach is to collapse compute and memory into a single hardware plane: Mythic stores neural-network weights in on-chip flash memory and performs the bulk matrix-multiplication operations in analog inside the memory array, reducing or eliminating the energy cost of shuttling data between separate processor and memory systems.
Mythic sells (and co-develops with partners) purpose-built Analog Processing Units (APUs) for inference workloads. Its flagship generation, the Mythic M1, is positioned as an analog APU delivering 25 TOPS of AI inference at low single-digit watt and without requiring external memory; it is packaged in multiple edge form factors (chip and M.2 cards). For next-generation scaling, Mythic is developing Vanguard (available in 2027), described as a hybrid analog-plus-digital APU platform to extend the architecture toward LLMs, computer vision, and enterprise AI workloads.
Strategically, Mythic has been expanding from analog inference compute toward system-level platforms and production readiness. In 2026 the company acquired Videantis GmbH to add a “unified digital processor” architecture and production-proven software stack, positioning Mythic to offer a hybrid analog+digital compute platform more directly across automotive/robotics, defense, and data center environments. In parallel, Mythic announced a joint development agreement with Honda to co-develop a next-generation analog AI chip for future vehicles—an effort Mythic frames as validating its compute approach in demanding, production-oriented environments.