Fabless startup Netrasemi has announced the launch of its flagship AI system on chip (SOC) A2000. The chip has now achieves successful silicon Bringup and getting ready for commercial production.
The SoC chip with smart vision and real-time video analytics capabilities enables on-device AI for smart cameras and Edge AI platforms. The chip fabricated at TSMC’s 12nm advanced process node targets high-TOPS AI performance, computer vision, Video streaming, secure boot and significant IO capabilities such as PCIe. A2000 is one of India’s first AI/ML chips and it features Netrasemi’s in-house Neural processor (NPU), Vision cores (VPU), Image Signal Processor(ISP), Crypto-Engines and other hardware acceleration IP cores. It also features their patented Heterogeneous graph-stream parallel processing architecture.
The chip design will be shared with select OEMs for joint R&D to develop further commercial use cases. The chip commercial production is likely to begin next year at TSMC, Taiwan.
Jyothis Indirabhai, Co-founder CEO, Netrasemi, says, “Our SoCs goes beyond conventional AI/ML integration by combining proprietary hardware acceleration IPs with domain-specific optimizations tailored for high-performance, real-time edge AI. The architecture is designed specifically for compact, power- and cost-sensitive edge devices. We are currently working with several leading OEMs to facilitate early sample evaluations, co-development, and advanced R&D initiatives”
Govt of India’s (MeitY) Semiconductor Design Linked Incentives (DLI) and Design Infrastructure Support (DIS) played a key role in achieving this critical milestone. The company was one of the first 4 startups selected for 15Cr INR DLI support in 2023 by MeitY. Netrasemi has raised Rs 125 crore in total funding till date from its marquee investors including ZOHO and Unicorn India Ventures.
These highly secure and power efficient AI chips target applications like smart cameras for surveillance, Edge AI boxes and Intelligent video gateways. The chip will go to volume production in 2027 and it is expected to address challenges of global platform developers in bringing advanced AI algorithms to small edge devices.
