CEO interview: InvenSense’s Abdi on expanding MEMS horizons
Abdi wants InvenSense (San Jose, Calif) to push forward on all fronts from hardware technology, through software to business models and in application sectors. There are a lot of opportunities for the companies that can move quickly and execute accurately, Abdi said.
Although founded in 2003, InvenSense was sufficiently successful with its original inertial MEMS components to go public in 2011. This success came mainly on being designed into consumer and mobile products such as smartphones. However, the IPO also brought about a change of management with the appointment of Behrooz Abdi as CEO replacing the founder and previous CEO Steve Nasiri.
Abdi has now been in place as CEO for a few years but surprised some people with a recent presentation that asserted that the MEMS component business is becoming a difficult place to do business (see Not enough money in MEMS, own the data, says InvenSense CEO).
Plans to move up the supply chain and start providing data services might seem like a radical change and run the risk of alienating component customers. At the same time InvenSense is a licensor of the intellectual property around its wafer-to-wafer CMOS-MEMS manufacturing process, making for a multi-faceted business model. We asked Abdi to say a little more about this.
Abdi acknowledge the complexity and said that a company has to focus. The various business models express, to an extent, where the company has come from, where it is and where it is going. But they also fit together, he stressed.
Right now InvenSense is a vendor of components-plus-software, Abdi said. It aspires to be a sensor solutions company offering end-to-end solutions but this is likely to be a long journey. “The company started with a process, which it took the company the first few years to perfect. We’ve now turned that process capability into a platform that is scalable.”
“The execution strategy was to be fabless which meant transferring the process to foundries,” Abdi said, but added that while InvenSense allows and encourages companies to use its process through TSMC and Globalfoundries, it is not open to all comers. “If you want to use the process, you have to deal with us,” he said.
So IP licensing is also part of R&D and business development. “We run [mulitproject wafer] shuttles and a bunch of universities and small companies have made use of this. We make sure it is beneficial to us rather than competitive. But we do want to promote innovation on the platform,” said Abdi.
Next: Data services
Expand on data services
We asked Abdi to expand on his idea that InvenSense would be migrating to being become a data services provider.
“There’s a lot of value in such things as exact location, activity and context. And inertial sensors used for dead reckoning can help with that. So take the example of cargo: where is it and is being dropped or banged? GPS is accurate with regard to location but consumes a lot of energy. So inertial navigation can provide location and additional context and we can help with the analysis because it needs to be done locally to minimize bandwidth and power consumption sending raw data up to the cloud.”
“It will probably be a business-to-business proposition. It probably wouldn’t make sense to produce a custom solution for one company. But if there are 20 companies all doing the same or a similar thing, that could make sense,” said Abdi.
A second example could be a wearables platform, he said. “Again we don’t want to sell wearables or services to end users but if there are 50 different companies all trying to decide what analysis they want and when to upload it to the smartphone or to the cloud it could make sense to provide a wearables local data service platform.”
However, it has been the case that inertial navigation via step counting has had issues over accuracy (see Has indoor navigation lost its way?). “That’s one of the reasons we have to get in there and do more of the heavy lifting; because it’s not very accurate. It is a challenge but there is a lot of performance to be extracted. It is a game of probabilities and statistics,” said Abdi.
Next: Back to hardware
What about the hardware?
But does this mean InvenSense is abandoning its hardware roots and focusing on middleware and software? No, says Abdi.
InvenSense started its business with inertial gyros and followed up with accelerometers on the same piece of silicon, then added pressure sensors and is now sampling microphones.
“There is more value to be added in core sensors. We started with gyroscopes for gaming on mobile phones. Next there is the gyro for image stabilization, which requires more precision and lower noise. And that higher precision can feed back into higher precision for location and navigation. Then there is the 7-axis sensor where pressure sensing is integrated with the 6 axes of accelerometer and gyroscope,” said Abdi outlining immediate steps the company can take.
What about more fundamental developments such as the use of piezoelectric sensing rather than capacitive or moving into other sensing domains beyond pressure and microphones?
“We are looking at every technology but we think that capacitive sensing has a lot of capability ahead of it,” said Abdi. “We’re not limited to inertial sensors and microphones but if you look at something like image sensors its too capital intensive. However, imaging for machines could be based on other technologies such as infrared and ultrasonics,” said Abdi adding that such options remain open for the company that wants to “sense everything” according to its logo.
Next Applications and Geographies
Applications and geographies
The success of InvenSense has been based on its involvement in smartphones, a market that is starting to plateau. Can InvenSense’s success be replicated in other consumer products and in other sectors?
“You need the consumer business to bring scale, to reduce cost structure and stimulate fast innovation. About 85 percent of our business is consumer, mobile, gaming. We’re not walking away from that. But now we can take of lot of those products and apply them to automotive and industrial applications,” said Abdi.
And InvenSense’s hardware expertise is not just focused on laying out sensors. The company has developed a series of its own software programmable processors to complement Cortex-M series microcontrollers that it includes in inertial measurement units (IMUs).
These include the DMP3 and DMP4 (see InvenSense adds multicore CPU to motion sensor). Abdi explained how they fit into a multi-tiered approach to software
“What we’ve designed in-house is optimized for power and low memory footprint but it is not open. On the other hand the Cortex-M0 is off-the-shelf and open, but not optimized. So we have a hybrid approach This means that if a new sensor fusion algorithm comes along it can be done in software running on the Cortex-M0 for time to market reasons before being migrated to the DMP.
Abdi added that hardware that supports neuromorphic computing for pattern recognition and learning behavior is on his radar but that for now it seems more suited to the large data sets generated by image sensors and machine vision.
Next: China and consolidation
We asked Abdi about China and its place in the MEMS world. “Right now China is a market rather than a source of competition. In the MEMS area it is similar to how China was in analog and RF 15 years ago. They are doing a great job in that area now. So MEMS will happen for Chinese companies some time but right now we see a lot of opportunities.”
We also asked Abdi about the wave of mergers and acquisition going through the industry in 2015 and could MEMS companies be prey to it?
“The semiconductor industry is going through a wave of consolidation and much of it is warranted because of a lack of scale, but it is mainly in the SoC world. InvenSense, because it is fabless, is capital-efficient. We do look at acquiring other companies or business lines from time to time; we acquired the microphones business from Analog Devices. We can do that but it is not something we have to do. In the MEMS sector it is more about partnership. And we do have partnerships with multiple microcontroller and processor companies; the likes of Atmel, Qualcomm and MediaTek.”
Atmel is set to be acquired by Dialog Semiconductor, which in turn has formed a partnership with leading MEMS company Bosch Sensortec, so it seems that industry consolidation will inevitably impact InvenSense, but nonetheless InvenSense seems to be in good company on its journey towards being an end-to-end sensor solutions company.
Related links and articles:
News articles:
London Calling: What’s next for Mr. MEMS
iPhone 6 chips discussed: InvenSense in, ST out
InvenSense adds multicore CPU to motion sensor
Has indoor navigation lost its way
Not enough money in MEMS, own the data, says InvenSense CEO
Audience, InvenSense buy up sensor fusion software firms
InvenSense buys ADI’s microphone business unit
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
