MEMS market decline in 2020 includes options for long-term growth

MEMS market decline in 2020 includes options for long-term growth
Market news |
The impact of the Covid-19 pandemic on transportation markets – automotive and aerospace – will cause the global market for MEMS components to fall by 5.2 percent in 2020, according to Yole Developpement.
By Peter Clarke


The market analyst says the market will be worth US$10.9 billion in 2020, down from US$11.5 billion in 2019. The weakest application sectors will be automotive, down 27.5 percent and defense/aerospace down 20.5 percent. While industrial and medical applications will both achieve double-digit percentage growth.

Yole is predicting a recovery to pre-Covid, 2019 market values in 2021 and growth to continue thereafter.

MEMS market forecasts by application sector, 2019 to 2025. Source: Yole Developpement.

Over the period 2019 to 2025 predicts an compound annual growth rate of 7.4 percent to reach US$17.7 billion in 2025. The consumer electronics market will enjoy robust growth to grow as the largest market although the needs of 5G telecommunications will be reflected in that sectors CAGR of 33 percent.

One of the reasons Yole marks industrial MEMS up so strongly is that thermal imagers and detectors for elevated body temperature detection are included in that category. Medical MEMS are following the same growth trend with pressure, flowmeter and microfluidics devices boosted by the covid-19 demand for ventilators and diagnostic tests.

MEMS for RF applications is supporting growth in consumer (smartphone) and telecommunications sectors. This is partly due to the expansion of 5G cellular and sub-6GHz bands for Wi-Fi, which precipitates the need for BAW filters.

Including relevant RF MEMS in the consumer market means it will only contract by 2.6 percent in 2020 but without RF MEMS it is poised to slump by 16 percent.

Next: Three paths to glory

Yole outlines three ways MEMS vendors are trying to adapt to escape the commoditization of their market and acquire more value from sensors.

One way is to open up new applications and use cases for sensors. Yole gives the example of augmented reality/virtual reality headsets and controllers with the value being in additional sales.

A second method is by the aggregation of sensors and improving existing use cases via novel algorithms and software. Here additional value lies in the chip/ASIC running the algorithms and software.

The third way is by adding processing and computing “at the edge” to perform artificial intelligence and machine learning type functions so that sensors can output useful information rather than large datasets. This increases the value in both the hardware and software. More silicon die area as an extra ASIC/MCU is added, but this also adds more functionality and could reverse the long-standing decreasing price curve of MEMS due to commoditization.

“Each player has its own strategy. Knowles, for example, managed to increase its value from the Google Pixel 3 to the Google Pixel 4 smartphone by adding an extra audio processor for DSP,” said Eric Mounier, senior analyst with Yole. “The acquisition of Audience some years back was critical for reaching this step. While selling MEMS microphones as usual, by adding the processing function, Knowles increased the value of the silicon sold to Google.”

To boost component functionality Bosch is collaborating with Qualcomm, while ST has added a machine learning core in its inertial sensors.

AI on the edge alongside the sensor is alluring and numerous startups are working on it, including such firms as Imerai, Aspinity, Syntiant and Cartesiam.

Related links and articles:

News articles:

Broadcom, Bosch increase lead in 2018 MEMS vendor ranking

ST embeds machine learning in 6DoF motion sensor

Bosch Sensortec adds software to Qualcomm BOM

3D echo-location startup raises some money

Infineon gets ‘always-on’ with Aspinity deal

Syntiant, Sensory deliver multi-language speech recognition

Sony plans to turn image sensors into a subscription platform

Development environment eases machine learning on to microcontrollers

Linked Articles
eeNews Analog