Powering the Future of Physical AI: SiMa.ai Modalix in Production

At SiMa.ai, we’re on a mission to scale Physical AI—bringing intelligence to the real world by delivering performant, power-efficient AI at the edge. Today, we’re proud to celebrate a key enabler in this journey: the launch of our breakthrough MLSoC, Modalix, which is built on TSMC’s N6 process technology.

Why Physical AI Requires a New Class of Silicon

Building chips for Physical AI isn’t just about accelerating neural networks—it’s about engineering systems that can interpret, react, and adapt to the real world. This poses unique challenges and distinct design priorities that differ sharply from traditional cloud or consumer AI applications:

• Real-Time Responsiveness: Physical AI systems must respond to dynamic environments—immediately and reliably. Latency is not just a performance metric; it’s often a safety or functionality requirement.
• Multimodal Processing: Unlike pure vision or language models in the cloud, edge applications must simultaneously process sensor fusion, video streams, control loops, and decision trees—all within tight thermal and power envelopes.
• Deterministic Compute: Applications in robotics, automotive, and defense demand predictable performance and deterministic execution, often with hard real-time constraints.
• Rugged Power Efficiency: Most edge environments—factories, vehicles, or remote locations—lack the power and cooling infrastructure of a datacenter. Silicon must perform within a strict energy budget, without sacrificing throughput.
• Security and Reliability: Physical AI systems often operate in critical infrastructure or untrusted environments. Chips must be designed with security, isolation, and fail-safe mechanisms from the ground up.

Designing Modalix to meet these needs required more than just brute-force compute—it demanded a sophisticated, tightly integrated system built on robust, efficient, and scalable process technology. That’s also the reason we chose to work with TSMC to build the new product based on the TSMC N6 process, which provides a significant enhancement of TSMC’s industry-leading N7 technology and offers customers a highly competitive performance-to-cost advantage as well as fast time-to-market with direct migration from N7-based designs.

The Physics of AI, Made Possible by Silicon

Physical AI is not just about neural networks—it’s about grounding intelligence in the physical world. This requires silicon that can handle a mix of modalities: vision, control, fusion, and decision-making—all in real time and at low power.

With Modalix on TSMC N6, we’ve built a chip that’s not only AI-capable but also physics-aware—enabling machines to perceive, interpret, and respond to the world with precision and reliability. From smart factories and autonomous drones to surgical robots and retail automation, Modalix brings intelligence to the edge, where it matters most.

SiMa.ai’s vision is to push the limits of what’s possible with AI at the edge.

As we continue to scale SiMa.ai’s ONE Platform across industries, we will continue  to deliver full-stack, production-grade AI solutions that are fast to deploy and easy to use.

Together with our industry partners, we’re not just building chips—we’re building the future of intelligent systems in the real world.