SiMa.ai Raises $150 Million to Scale Physical AI Chips

SiMa.ai has raised $150 million as demand grows for AI computing that can run directly inside robots, vehicles and drones. The San Jose-based semiconductor company’s Series C round values it at $1.45 billion and takes its total funding to $500 million. 

The company is targeting a specific problem in Physical AI: running increasingly complex AI models on machines that cannot afford the power consumption, size or latency of data-center hardware. SiMa.ai’s approach combines purpose-built silicon with software, its current Modalix MLSoC is designed for workloads such as computer vision, sensor processing and on-device AI, with SiMa.ai listing performance of up to 50 TOPS at less than 10 watts.

The new funding will support the next generation of that platform. SiMa.ai says its planned 2028 hardware will target 1,000 dense TOPS and use a combination of machine-learning IP, chiplets and system-on-chip designs. The company is targeting applications ranging from humanoid robots and autonomous vehicles to advanced drones and AI-enabled automotive cockpits.

In June, SiMa.ai launched Palette Neat, an agentic development environment intended to reduce the work involved in porting and deploying Physical AI applications across its hardware. The company says the platform can reduce development cycles from months to days or hours and preserve much of the existing application code when moving between platforms.

Physical AI developers are not simply looking for more computers. They need systems that can process cameras and sensors, run AI models locally and respond quickly within tight power budgets.

SiMa.ai says its platform is already being used across robotics, automotive and drone applications, with customers and partners including Bosch, Emerson, Micron, TRUMPF, L&T Technology Services and others.

The funding gives SiMa.ai more room to push that strategy from specialised edge AI into a broader Physical AI computing platform. 

Image source: SiMa.ai