What Mythic does
Mythic (mythic.ai) is a private AI semiconductor company focused on analog compute-in-memory (CIM) architectures for efficient neural-network inference at the edge and in data centers. The company’s core premise is that conventional AI accelerators waste energy by shuttling data between separate memory and compute units; Mythic instead performs computation directly inside memory (via analog compute-in-memory processing), aiming to improve performance-per-watt and total cost of ownership for inference.
Mythic markets a chiplet-based Analog Processing Unit (APU) platform intended to scale from low-power edge deployments to higher-performance enterprise inference using the same underlying approach. Its flagship first-generation APU, Mythic M1, is positioned as an analog inference chip that stores up to 80 million neural-network weights on-chip and performs matrix multiplications in place, targeting low-watt operation in embedded/edge settings. Mythic also describes next-generation products: Vanguard (a 2027-available hybrid analog-die plus digital-die APU designed to scale toward LLM and broader enterprise workloads) and Mead (an in-development 3D NAND-based LLM compute platform targeting consumer-device-class sub-1W operation for large models).
In parallel with its compute roadmap, Mythic pursues vertical sensing/AI-in-sensor concepts such as Project Starlight (low-light correction running as a neural network at the sensor under low-power targets). The company’s recent strategic moves include (1) raising substantial new funding to scale engineering and product go-to-market, (2) announcing a joint development agreement with Honda for an automotive-grade AI SoC, and (3) acquiring Videantis GmbH to add production-proven digital processor IP/software into what Mythic describes as a hybrid analog+digital AI compute platform.
Business model-wise, Mythic appears to operate as a silicon provider and platform vendor: it offers its APUs and accompanying software toolchain/SDK intended to compile and run DNN workloads on Mythic hardware, then partners with OEMs and enterprise customers to integrate AI inference into target products (automotive/ADAS, robotics/physical AI, defense & sensing, and private/enterprise inference).