Untether AI

About

Untether AI is a Canadian semiconductor innovator pioneering at-memory compute architecture for artificial intelligence inference acceleration. Founded in 2018 and headquartered in Toronto, the company has developed a revolutionary approach that fundamentally reimagines how AI computations are performed by bringing compute directly to where data resides in memory, eliminating the performance bottleneck of data movement that plagues traditional architectures. Their flagship tsunAImi inference accelerator chip delivers industry-leading performance per watt and performance per dollar for AI workloads in data centers and edge deployments. The tsunAImi architecture features a massively parallel array of processing elements tightly coupled with memory banks, enabling thousands of simultaneous compute operations without the energy-intensive data transfers required by conventional von Neumann designs. This innovative design achieves over 2 petaOPS of AI inference performance while maintaining exceptional energy efficiency, making it ideal for deploying large-scale AI models in cloud infrastructure, autonomous systems, and intelligent edge applications. Untether's runAI software platform provides a complete development environment supporting TensorFlow, PyTorch, and ONNX frameworks, enabling seamless migration of existing AI models to their hardware with minimal engineering effort. The technology addresses critical pain points in AI deployment including inference latency, power consumption, total cost of ownership, and scalability for production workloads. With over $150 million in funding from prominent venture capital and strategic investors, Untether AI is targeting high-performance computing markets where AI inference demands exceed capabilities of conventional GPU and CPU-based solutions, including autonomous vehicles, robotics, natural language processing, computer vision, and large-scale recommendation systems deployed by hyperscale cloud providers and enterprises.