Here is a prediction: MCUs with AI functions will flourish everywhere~~

okhxyyo

Here is a prediction: MCUs with AI functions will flourish everywhere~~ [Copy link]

I was browsing in the group a few days ago and happened to see netizens talking about AI. I felt that there is really no AI and no terminal now. Now AI sinking to the edge has become a general trend. In the group chat, a netizen asked: Will more and more MCUs have AI functions in the future?

I think the answer is yes. I would like to make a bold prediction that in the next few years, MCUs with AI functions will flourish everywhere.

Call a forecast data found on the Internet:

By integrating AI functions into MCUs, AI algorithms can be processed and responded to locally in real time without relying on the cloud or other remote servers, enabling devices to make decisions and react more quickly, and to achieve efficient AI computing under low power conditions. It is predicted that by 2025, 75% of data will be processed on the edge side, and the end-side AI MCU market has huge potential.

▲ Total shipments of devices with edge machine learning capabilities in major global market sectors from 2021 to 2026

One thing that needs to be mentioned is that if you want to develop AI applications, you need to improve the hardware computing power. This requires that the MCU that can drive AI must have an NPU. And the MCU manufacturers have already started to lay out.

for example:

Last year, NXP launched the MCX N microcontroller product;

In 2022, ST launched the STM32N6, a general-purpose microcontroller with an NPU neural network hardware processing unit;

Infineon Technologies also recently launched the PSoC Edge series of microcontrollers, which provide high performance and security for edge devices that operate machine learning;

Last year, Renesas launched the RZ/V2MA series of products with built-in visual AI accelerators, which enable accurate image recognition and multi-camera image support.

ADI's MAX7800X series of edge AI solutions consists of two microcontroller cores (ARM Cortex M4F and RISC-V) plus a convolutional neural network (CNN) accelerator;

In recent years, Arm has continuously released series such as MicroNPU Ethos U55 and U65 with computing power cores.

. . . . . . . . .

Do you think Guan Guan’s prediction makes sense?

What else do you think MCU can integrate?

wangerxian

MCU with AI function mainly provides accelerators for some AI algorithms

秦天qintian0303

Without a model foundation, it's all nonsense. At most, you have the computing power, but you can't save it.

freebsder

Qintianqintian0303 posted on 2023-12-18 21:45 Without a model foundation, it’s all nonsense. At most, you have the computing power, but you can’t save it.

Most of them are model parameter adjusters.

coderHH

Indeed, edge computing has been very popular in recent years.