Untether AI focuses on improving AI inference efficiency and reducing energy consumption, developing the SpeedAI accelerator cards for enhanced performance in AI applications.
Key Focus Area
Untether AI is concentrated on enhancing AI inference performance for various applications through the creation of advanced hardware solutions. Their primary aim is to facilitate the deployment of artificial intelligence (AI) from the cloud to the edge across multiple sectors, such as AgTech, automotive, government, and financial services. They focus on accelerating AI inference tasks, promoting performance, energy efficiency, and cost-effectiveness.
Unique Value Proposition and Strategic Advantage
Untether AI’s unique value proposition lies in its novel at-memory compute architecture. This architecture positions the processing element directly next to memory cells, thereby reducing the power consumption associated with data movement, which traditionally dominates AI workloads. This setup diminishes the need for larger data transfers, leading to significant operational efficiency and exceptionally high compute density, making it particularly appealing for high-performance, power-sensitive AI applications. Their products offer energy-efficient AI operations, which claim to outperform traditional systems by providing higher throughput and lower power usage.
Delivery on the Value Proposition
To achieve their value proposition, Untether AI delivers a suite of products comprising integrated circuits (ICs), accelerator cards, and software development kits:
AI Inference Accelerator ICs: These include the runAI200 and the speedAI240. The ICs exploit an at-memory architecture for optimal compute density and low power consumption. They support a variety of AI workloads, from vision-based networks to transformer networks, catering to needs across automotive, finance, agriculture, and government sectors.
AI Accelerator Cards: Untether’s cards, such as the tsunAImi tsn200 and tsunAImi tsn800, are designed for high-demand inference tasks, delivering substantial performance in a power-efficient manner. They incorporate features suitable for both server-grade and edge applications, making AI deployment more broadly feasible.
Software Development Kit (imAIgine SDK): This SDK enables seamless model transition from popular frameworks like TensorFlow and PyTorch to Untether AI’s hardware. It offers a Model Garden and custom development flows that support the rapid deployment of AI solutions, emphasizing ease of integration and operational flexibility.
Target Markets and Applications: Untether AI focuses on sectors like agricultural technology (AgTech), automotive, financial services, and government applications, where AI can deliver substantial improvements in operational efficiency, safety, and sustainability. For example, in AgTech, AI assists in precision farming and crop monitoring, while in automotive, it enhances advanced driver-assistance systems (ADAS).
Overall, Untether AI emphasizes an innovative approach to overcoming AI deployment challenges, delivering higher performance and efficiency through a unique hardware design that benefits from reduced power consumption and increased density of compute resources. Their strategic partnerships and compliance with industry standards further support their market positioning in enabling AI across diverse operational environments.