Skymizer is a Taipei-based AI technology company founded in 2013, specializing in compiler and intellectual-property design that makes AI inference efficient on edge hardware. Originally rooted in compiler engineering, Skymizer evolved into a software-hardware co-designed IP company aimed at empowering the next generation of AI-powered devices, particularly for edge inference where power and memory budgets are tight.
The company's technology centers on its Language Instruction Set Architecture (LISA), a design that enables accelerators to execute modern AI workloads, including language and vision models, efficiently. By co-designing the instruction set with the compiler stack, Skymizer aims to extract more performance per watt and per dollar from edge silicon than general-purpose approaches, positioning it as critical IP for chip designers and device makers building AI capabilities into constrained hardware.
In May 2025 Skymizer launched HyperThought, a platform powered by its third-generation LISA v3. HyperThought is designed for edge AI and supports efficient execution of text, vision, and instruction-based models, concurrent multi-model processing, and agent-based workflows. The company has also highlighted breakthrough architecture work enabling ultra-large LLM inference on a single card, underscoring its ambition to push the limits of efficient on-device and edge-server inference as model sizes grow.
Skymizer has raised roughly $18.3 million from investors including TGVest Capital, Harbinger Venture Management, Sunsino Venture Group, and the Qualcomm Innovate in Taiwan Challenge, with backing from the Orion Opportunity Fund. As demand surges for running capable AI models outside hyperscale data centers, Skymizer's compiler-and-IP strategy positions it to supply the foundational software and architecture that make efficient edge and post-GPU inference practical.