TLDR: SiMa.ai has launched its next-generation platform, featuring the Modalix Machine Learning System on a Chip (MLSoC) and a new System-on-Module (SoM), designed to run reasoning-based Large Language Models (LLMs) and other generative AI workloads on-device with less than 10 watts of power. This innovation aims to accelerate the adoption of Physical AI across various industries, including robotics, automotive, and healthcare.
SiMa.ai, a US-based AI chip company, has introduced its latest platform, spearheaded by the Modalix Machine Learning System on a Chip (MLSoC), alongside a new System-on-Module (SoM) and development kits. This release is poised to meet the escalating demand for Physical AI applications, enabling advanced AI capabilities directly on devices while maintaining remarkable power efficiency.
The Modalix MLSoC, a second-generation offering, is engineered to deliver high performance without compromising power efficiency, operating at less than 10 watts. Krishna Rangasayee, founder and CEO of SiMa.ai, stated, “The era of Physical AI is here. With Modalix now in production, we’re accelerating its global adoption.” The chip’s flexible, Arm-based architecture facilitates real-time decision-making, natural language interaction, and seamless support for a wide range of AI workloads, including Large Language Models (LLMs), transformers, convolutional neural networks (CNNs), and generative AI (GenAI).
Further enhancing its appeal, SiMa.ai has also unveiled the Modalix SoM, which boasts pin-compatibility with leading GPU SoMs. This feature significantly simplifies integration into existing systems, making it an attractive solution for developers. The platform is comprehensive, including MIPI, memory, and other essential I/O components, designed for the rapid scaling of Physical AI systems. Complementing the hardware, the company’s LLiMa software framework further streamlines the deployment of LLMs and the integration of GenAI capabilities into Physical AI applications.
SiMa.ai’s commitment to innovation is underscored by its strategic collaborations. Ami Badani, chief marketing officer of Arm, commented, “Modalix showcases the scale of innovation possible with Arm’s flexible, power-efficient compute platform. SiMa.ai is enabling smarter and more sustainable systems across industries.” The development process was accelerated through collaboration with Synopsys, leveraging their AI-powered design tools. Furthermore, TSMC’s advanced N6 process was instrumental in ensuring Modalix met stringent power, thermal, and reliability standards. Sajiv Dalal, president of TSMC North America, affirmed, “This collaboration underscores our commitment to driving energy-efficient chip innovations that are redefining the future of AI.”
Also Read:
- NVIDIA Advances Robotics with New AI Models for Human-Like Reasoning
- Alif Semiconductor Unveils Groundbreaking Benchmark Results for GenAI-Enabled Edge MCUs, Solidifying Market Leadership
The new Modalix SoM and DevKit are now commercially available. Pricing for the 8GB SoM begins at $349, with the 32GB version priced at $599. SiMa.ai anticipates strong demand for its products, particularly in sectors such as robotics, automotive, and healthcare, as it aims to accelerate the global adoption of Physical AI.
