# Netrasemi Brings Up A2000 AI Chip, Begins Customer Evaluation Phase > Source: > Published: 2026-06-04 22:00:00+00:00 Netrasemi announced it has brought up its A2000 edge AI chip and has begun supplying engineering samples along with the development platform to selected customers. The chip, manufactured on TSMC’s 12-nm process, is the first full-feature SoC developed by the startup and forms the foundation of a broader product roadmap that includes the R1000 AI microcontroller and the future R4000 chiplet-based processor. The company is currently working with early-stage customers, including OEMs and ODMs, to validate the platform and develop reference designs before moving toward production next year. In an exclusive conversation with EE Times, Jyothis Indirabhai, CEO and co-founder of Netrasemi, said, “This is a very important milestone for Netrasemi. The A2000 is our flagship chip because it contains the complete set of in-house IPs that we have developed. The successful bring-up gives us confidence in taking our entire family of SoCs to production.” [View All](https://www.eetimes.com/category/sponsored-content/) You can watch the full interview below: The company is currently supplying evaluation kits (EVKs) and engineering samples to customers rather than selling production silicon to focus on validation, customer feedback, and ecosystem development. “Our objective is to ensure that the chip is tested properly and that the reference designs are ready,” he said. “Without reference designs, it is difficult to engage with OEMs and ODMs.” The company expects initial revenue by the end of next year as production begins. **Netrasemi’s heterogeneous graph stream architecture** Indirabhai described the A2000 as the company’s flagship product because it incorporates the full set of in-house hardware acceleration IP developed by Netrasemi. The chip delivers up to 12 TOPS of AI performance and integrates a complete video pipeline, vision processing capabilities, and a neural processor. The company said it has been designed for applications requiring AI, vision, and vector processing while maintaining power efficiency and cost targets suitable for products such as surveillance cameras. The A2000 is built on what Netrasemi calls its heterogeneous graph stream architecture. The architecture combines multiple in-house IP blocks, including neural processing units (NPUs), graphics processing units (GPUs), vector engines, general-purpose vector engines, direct memory access (DMA) engines, and encryption engines. “We developed what we call the heterogeneous graph stream architecture, which enables multiple heterogeneous algorithms to run efficiently together,” Indirabhai said. “We built the architecture and the IP portfolio first.” The company has also developed a software development environment called Netra Edge Studio, which includes sample applications, compiler tools, drivers, and supporting software. “We have built the studio to simplify development and reduce the effort required from customers,” he said. Indirabhai said the architecture allows multiple AI models to run simultaneously with minimal overhead. “Take the example of an in-cabin driver monitoring system that may require around six AI models operating together,” he explained. “The architecture is designed to minimize inefficiencies caused by context switching and resource allocation.” The company selected the 12-nm process node because it offers what Indirabhai described as a balance between performance and manufacturing cost. “We are targeting edge and embedded applications,” he said. “At 12 nm, we do not lose much performance while the manufacturing cost remains manageable.” Netrasemi licenses interface IPs such as double data rate (DDR), USB 3.0, and Peripheral Component Interconnect Express (PCIe), while developing its core acceleration and processing IPs internally. The company also uses Arm processors under the Arm Flexible Access program and has adopted the Shakti RISC-V core as the basis for its RISC-V development efforts. “We selected Shakti because we wanted complete control over the IP,” Indirabhai said. “We are modifying and extending it.” **Future product roadmap** The A2000 forms part of a larger family of processors sharing a common architecture. Alongside the A2000, Netrasemi has taped out the R1000, a real-time processing microcontroller with built-in AI capabilities aimed at IoT and smart sensor applications. The R1000 was taped out in the last week of April and is expected to return from fabrication in August. The startup described the R1000 as an ultra-low-power AI microcontroller targeting AIoT applications. “Both the R1000 and A2000 are targeted for production next year,” Indirabhai said. The company is also developing the R4000, a higher-performance chiplet-based processor built on the same architectural foundation. This chip is being positioned as an edge AI server-class chip by Netrasemi. “The neural processor architecture we developed can scale both up and down,” he said. “That is why we were able to create both the R1000 and A2000 using the same architecture. The next evolution of that architecture is the R4000.” The R4000 uses a chiplet architecture and is being developed as a two-die solution. Although the processors target different performance levels, they share common architectural components. The A2000 uses an Arm Cortex-A processor, the R1000 uses a RISC-V R25 core, and the R4000 also adopts a Cortex-A-based SoC architecture. The company’s strategy is to create a portfolio of scalable IPs and products rather than building standalone chips. **Avenues for market growth** Netrasemi initially targets surveillance, in-cabin monitoring, selected automotive applications, and drone-based surveillance. “These are sectors where smart vision is a primary requirement,” Indirabhai said. He added that the underlying technology can support many more applications, including media analytics, audio processing, and millimeter-wave radar processing. “We believe it can support more than 100 applications,” he said. “However, as a startup, we cannot address every opportunity simultaneously.” Netrasemi has therefore focused its interfaces, software ecosystem, and reference designs on a limited number of high-volume markets. Although the startup sees opportunities globally, India remains an important market because of local procurement advantages and growing demand for surveillance and vision-based systems. “India is an attractive market because we are a local company, and there are favorable procurement policies,” Indirabhai said. “Even in some of our market studies, India is emerging as the second-largest market opportunity for both the R1000 and R2000.” Potential customers include companies developing single-board computers, cameras, and other embedded systems, along with integrators serving industrial, automotive, commercial, and residential markets. “Camera deployments in India are growing rapidly,” he said. “If local companies do not address this market, a large volume of chips will continue to be imported.” The company currently packages its multi-project wafer (MPW) chips through overseas outsourced semiconductor assembly and test (OSAT) providers, including Kyocera, but is evaluating Indian OSAT providers for future production. Netrasemi also credits government support for helping the company reach its current stage. The company is one of the beneficiaries of the Design Linked Incentive (DLI) scheme under the Ministry of Electronics and Information Technology (MeitY). “Officials from MeitY have been involved throughout the process, reviewing progress and supporting the program,” Indirabhai said. Over the next year, the company plans to continue customer sampling, build reference designs, and gather product feedback. Depending on the results, it may undertake another multi-project wafer cycle before entering full production. “Customer feedback is more important than orders at this stage because it helps us avoid mistakes and refine the product,” he said. Netrasemi expects to secure sufficient customer commitments by the second quarter of next year to support mass production and plans to move to full-mask production by the middle of next year. Netrasemi has raised about $15 million in two funding rounds and expects to raise additional funds to support manufacturing. ##### See also: [Indian Startup Builds Full-Stack Edge AI Chips Using In-House IP](https://www.eetimes.com/indian-startup-builds-full-stack-edge-ai-chips-using-in-house-ip/) [VLSI Design Conference 2026: Experts Advocate Product-Led Chip Push for India](https://www.eetimes.com/vlsi-design-conference-2026-experts-advocate-product-led-chip-push-for-india/)