What Mythic does
Mythic is an AI chip company developing analog “compute-in-memory” processors aimed at energy- and cost-constrained inference at the edge and in data centers. Its core architectural premise is to reduce (and in some cases eliminate) the power cost of moving AI tensors between separate memory and compute blocks by computing matrix multiplications where model weights are stored, using analog compute-in-memory.
More
On its website, Mythic positions its solution as an alternative to GPU-based inference for workloads such as robotics/physical AI, automotive/ADAS, defense/sensing, and enterprise LLM inference. The company’s product line includes the M1 Analog Processing Unit (APU) for deployed edge use, along with higher-integration/hybrid “platform” roadmaps such as Vanguard (a next-generation APU described as a hybrid analog die + digital die platform) and longer-term products described as in development, such as Mead (a 3D NAND-based LLM-in-a-chip concept). Mythic also describes a software workflow/SDK for model optimization, quantization, compilation, and framework compatibility (including ONNX, PyTorch, TensorFlow, and NVIDIA TensorRT integration via supported toolchains). A notable strategic step is Mythic’s acquisition of Videantis (announced May 19, 2026; described on Mythic’s site as in January 2026) to add a unified digital processing/IP layer that Mythic says complements its analog compute plane, supporting orchestration of additional AI workloads beyond dense matrix operations.