Electronic Specifier attended the briefing to find out more about how the company is approaching the emerging Edge AI market and refining its GaN portfolio.
New Edge AI MCU series
The new series of MCUs, the RA8P1 group, are specifically targeted at AI, ML, and real-time analytics applications. Thanks to TinyML models, devices that are more constrained both in terms of board space and power, are able to harness the advantages of Edge AI.
The launch of the RA8P1 group follows a roadmap of Renesas‘ where its first MCUs with Arm Cortex-M23 and M33 CPUs were launched in 2020, followed by its first MCU with Arm Cortex-M85, currently the highest performance Cortex-M processor today.
Renesas is launching products directed at the Edge AI market because of multiple underlying trends driving its growth, Daryl Khoo, Vice President, Embedded Processing said.
“We have IoT, AI, and 5G, three powerful technologies maturing at the same time,” said Khoo. “This convergence will be transformational.”
Where processing five or so years ago followed a “Cloud-centric model,” true intelligence has to be enabled at the Edge. There are trends driving demand for higher-performance AI-enabled MCUs, which include increased demand for low-power capabilities on Edge devices; the need for more integrated solutions that reduce cost; and the building of multi-modal applications that bring together voice, vision and real-time analytics.
“These trends in industry drive us to develop higher performance MCUs that can meet the needs of emerging applications and use cases,” noted Khoo. “MCU architectures are evolving to include AI capability and higher compute performance approaching that of lower end MPUs.
“Solutions are also involving smaller, highly optimised neural network models ideal for resource-constrained MCUs and AI optimised software tools and ecosystem solutions are becoming available.”
All of these micro and macro trends compelled Renesas to develop and launch the RA8P1 MCU family, designed for AIoT. It features the 1GHz Arm Cortex-M85 and 250MHz Cortex-M33 cores with the Ethos-U55 NPU. It utilises the Ethos-U55 to offload the CU for any compute intensive operations and depending on the neural network used, can provide 35x more inferences per second than the Cortex-M85 on its own.
What’s also new about the RA8P1 is that it has been manufactured on 22 nanometer, which has enabled the use of MRAM, offering faster write speeds, and complementing the adoption of other advanced processes like sub-22 nanometer technology, said Khoo.
Alongside the launch of the MCU family, Renesas also announced its RUHMI framework, designed to support AI application development by providing a comprehensive AI framework for MCUs and NPUs to generate and deploy neural network models. This allows model optimisation, quantisation, graph compilation and conversation, and generates efficient source code. It also provides support for ML AI frameworks such as TensorFlow Lite, Pytorch & ONNX.
Expanding GaN portfolio
In May 2025, Renesas halted the production of SiC chips for the time being, to focus on the GaN market. Its decision to double down on GaN – which is increasingly being viewed as an alternative to SiC, which some posit has hit its ceiling in performance – was reflected in the briefing, as Primit Parikh, Vice President, GaN Business Division, Power presented next to take attendees through its three latest solutions belonging to the Gen IV Plus product family.
“Renesas is doubling down on GaN and MOFSET, where we have key competitive strengths and see a very healthy, sustained demand to enable higher power density, high efficiency designs, given the market dynamics over the last two years,” explained Parikh, noting that this pivot in focus allowed the company to capitalise on key areas including data centres and power infrastructure.
GaN performs power conversion better than traditional silicon and SiC, and is one of the fastest growing segments, Parikh added, which attracted Renesas’ interest.
Therefore, the launch of its three high-voltage 650V GaN FETs targeting AI data centres and server power supply systems shows how the company is capitalising on this competitive advantage. The GaN FETs combine GaN technology with a silicon-compatible gate drive input, to reduce switching power loss and keep the operating simplicity associated with silicon FETs. “We see fast growth in demand for both high voltage and low voltage GaN products and solutions in the core markets, achieving a CAGR of well over 50%,” said Parikh. “We’re expanding the 650V high voltage GaN portfolio, plus adding 40 to 200 voltage GaN that will ramp in 2026.”
Although GaN market growth is focused in low-power chargers and adapters, it is the higher power required in AI and industrial environments, where Renesas’ approach will be to focus on the mid and high-power segments.
New and notable features of the GaN Gen IV Plus family include achieving a smaller die size, higher performance, lower loss, higher efficiency, and lower cost.
“The power densities and efficiencies that lead to lower system cost from GaN are becoming the necessity for large growing market areas of AI, servers, industrial, as well as e-mobility,” said Parikh. Responding to ongoing trends in the market, Renesas reflected how it is thinking deeply and responding to growth in both Edge AI and GaN; technologies that are anticipated to be transformative across an array of industries.
