Usual business advice includes thinking big to win big. Some organizations create Big Hairy Audacious Goals. Others like to find new markets that are underserved and grow to be number one. The semiconductor industry has Moore’s Law – the premise that the minimum cost point is achieved by doubling the number of transistors per chip every two years – driving it forward for almost fifty years.
Janusz Bryzek set a dramatic and ambitious goal of $1 trillion sales for the microelectromechanical systems (MEMS) market in 2022. Even though the MEMS market is expected to have “only” $12 billion in revenue in 2012, he isn’t being called a fool. Having cofounded eight seminal Silicon Valley MEMS companies and currently the Vice President of MEMS Development at Fairchild Semiconductor (which recently acquired his last company), Janusz is taken quite seriously.
Yes, at last week’s MicroElectronics Packaging and Test Council (MEPTEC) 10th annual MEMS Technology Symposium there were some who doubted the market would grow this large. As each presenter quickly discovered, Janusz as conference chair asked as the first question if MEMS would reach either $1 T or 1 T units if not addressed directly in their presentation. In keeping with the conference’s theme “Sensors: A Foundation for Accelerated MEMS Market Growth to $1 Trillion” the presentations focused on the growth and application of MEMS sensors for a wide variety of applications.
In his opening keynote “Sensory Swarms”, Professor Kristofer Pister (Electrical Engineering and Computer Science at University of California, Berkeley and founder / CTO of Dust Networks) described both the possibilities and challenges of very large sensor networks. The market analysts were proven wrong about the adoption rate of wireless sensor networks, predicting in 2007 that there would be more sensors than cell phones. In 2012, instead of the predicted market of 600 M units only 20 M units were sold. Once again, the hockey stick revenue curve was not realistic.
Professor Pister described how sensor technology has crossed the chasm for use in industrial automation applications which does not have the extreme cost sensitivity found in consumer applications. Key attributes are the sensor must be reliable, usable anywhere (including harsh environments), and last for over a decade without requiring service (including changing batteries). They need to be easy to deploy without requiring an information technology (IT) group to support the network full-time. Hence the interest in wireless self-configuring networks that drop in to existing IT infrastructure and integrate with current business application systems.
Unfortunately, it has taken longer than anticipated to achieve these requirements. For example, until last month there were three competing and incompatible wireless protocol standards being used by different manufacturers. They were finally resolved as one Institute of Electrical and Electronics Engineers (IEEE) standard. And for the last fifty so years, companies have attempted to implement wireless sensing most often with poor results. As a result, potential users have viewed new technology with skepticism that requires proof of many years of rock solid performance to dispel.
One impressive example Professor Pister provided was a network of one million sensors in a single Wheeling-Pittsburgh Steel mill. Using real time measurements, the sensors improved overall mill efficiency more than 5% by moving the equipment from a strict scheduled based preventative maintenance regime to as-needed service (when predictive measurements crossed a certain limit) which better reflected actual usage. He claims this as one of the biggest performance improvements in the history of industrial steel making. The sensors operate in a very harsh environment which not only preclude wired installation but also may be inaccessible for regular servicing. Another example of a large but distributed network is those being built by Streetline which sense available street and garage parking spaces in major cities. Streetline uses these sensors to power their Parker smartphone application to help drivers find parking spots in realtime.
In order to achieve greater adoption, especially beyond the area of industrial sensors, Professor Pister believes the cost of the network “link” hardware (i.e. the microcontrollers and the radio electronics) for each sensor node need to drop from the current $2 to $20 each to less than $0.20 each. He does not believe the current market estimate of one billion sensor units in 2015. However, there is still a sizable market once costs come down. That said, when adoption time is factored in the predicted market growth rates for new applications is still too high. Yes, there will be more large sensor swarms but that won’t get us to one trillion units.
In “Accelerating MEMS Market to $ Trillion / Trillion Units”, Janusz further laid out the growth history and potential of the sensor and overall MEMS market. He highlighted two large sensor applications: Hewlett-Packard’s (HP) Central Nervous System for the Earth (CeNSE) which is proposing a network of a trillion nodes (described in a presentation by Rich Friedrich, Director of the CeNSE Program for HP) and Bosch’s view of 7 trillion devices serving a population of 7 billion in 2017 (i.e. 1000 devices per person on average). He identified three areas that are current challenges needing to be solved to enable future growth: no atomic level simulation to reduce the need to build physical devices to check both designs and fabrication processes, device packaging which tends to be unique to each family of designs, and cumbersome test methods that are specific to each design. Lastly he identified the need for greater process standardization to enable a foundry model similar to that of the semiconductor industry. Using standardized fabrication processes, companies would compete on their designs and not their process development ability.
Robert Haak (Vice President Asia/Pacific of MANCEF) followed with “Implementing the Trillion Dollar MEMS Roadmap.” Mr. Haak described a similar set of challenges identified by Janusz along with other road mapping efforts. In contrast to the International Technology Roadmap for Semiconductors (ITRS) which focuses on the technology advances required to keep parity with Moore’s Law, he proposed a Market Goal roadmap focused on what is needed to drive the market to the $1 T per year level. Three areas were identified that need to be driven in a coordinated manner: “sensor technology itself, technology of data transfer from the sensor, and technology behind the equipment used to process the data.” Similar to the ITRS, the proposed timeline should run fifteen years out with a review cycle of five years. In order to achieve rapid progress, Mr. Haak suggested companies need to move from competition to coopetition possibly with shared pre-competitive research similar to how the Semiconductor Research Corporation (SRC) operates for the semiconductor industry. He outlined next steps to advance the idea of a MEMS roadmap including raising funding to permit MANCEF to organize the roadmap.
Several of the presentations took a quantative approach to the question of potential market size. Gregory Galvin (President and CEO of Kionix) reviewed a brief history of different types of MEMS sensors (accelerometers, gyroscopes, and magnetometers) and future potential applications (enhanced location based sensing, augmented reality, health monitoring, etc.). From their experience, MEMS sensors are extremely price elastic and lower prices clearly increase adoption rate and numbers of units sold. For example, some of their common sensors are $1 each today but they would be used in many more devices if these prices dropped to $0.10 to $0.20 each.
In terms of the one trillion dollar market, Mr. Galvin calculates since MEMS sensors on average are 2% of the end cost of the products this would imply a $50 trillion market. This is, of course, unrealistic since this is roughly 80% of the world’s current gross domestic product (GDP) of $63 trillion (2010 data). He argues the industrial end markets which could use MEMS devices today have a GDP of $7 T. And if you apply a 2% MEMS content metric assuming a MEMS device in every industrial application this results in a reasonable $140 B available market today. At a 15% compounded annual growth rate (CAGR) this could grow to roughly a $0.5 T market in 10 years. And with unit average selling prices (ASPs) below $0.50 total, unit volume could reach over 1 T units.
Jérémie Bouchaud (Director and Senior Principal Analyst MEMS and Sensors, IHS iSuppli) in “The MEMS Revolution: from Billions to Trillions?” reviewed market trends and forecasts. Unfortunately the historical growth sectors for MEMS sensors have already saturated and gaming controllers in particular has begun to show a decline in unit volume. The mobile “stars” – smartphones and tablets – has gone from explosive growth of 60-90% year over year to estimated growth rates of below 25% moving forward. I would note many industries would be ecstatic with a growth rate of over 20% per year.
He also sees the “central nervous system for the earth” including oil exploration, asset tracking, earthquake monitoring, structural health, etc. as being extremely price elastic but of a limited market potential of a few million units per year. Lastly, Jérémie does not believe vehicle sensors will grow significantly. Electric vehicles require fewer sensors and car manufacturers are learning how to improve software to either use existing sensors or reduce the number of sensors per vehicle to obtain the required data.
In “Integration of the Accelerometer — the First Step of the MEMS Revolution”, Jean-Christophe (JC) Eloy (President and CEO of Yole Développement) similarly reviewed their analysts’ overall market trends. They see gyroscopes and accelerometers reaching market saturation and the move to combined navigational devices lowering total unit counts and revenue per function. As the MEMS market matures there is a greater stratification between the top four MEMS manufacturers and the rest of the market in terms of revenue. Currently a very small number of MEMS companies are fabless. But longer term, there will need to be a switch to this model to take advantage of economies of scale available to only the largest manufacturers. Especially as they are estimating the cross over from 150 mm wafers to 200 mm wafer volumes (with significantly lower cost per area) to occur in 2012. These economies of scale should result in a significant increase in the number of fabless MEMS companies. He also does not see a $1T market but other opportunities to add value beyond the MEMS sensors that will grow over time.
In the afternoon keynote “Motion Interface the Next Large Market Opportunity”, Steve Nasir (Founder, President, and CEO of InvenSense) also provided market overviews of sensors by application including cost trends. One very important point that Mr. Nasir made that should be kept in mind when reviewing market projections is that five years ago no market analyst predicted any substantial penetration of gyroscopes in mobile devices. This was based upon input from both phone manufacturers and end customers indicating no interest. Personally, he had given up going to Asia to meet with phone manufacturers to discuss his products since they had no plans to include gyroscopes. The entire situation changed when Apple launched the iPhone in 2007 and every other mobile device manufacturer madly scrambled to include gyroscopes. As Mr. Nasir pointed out all it takes is one leader to create a substantial market. I would add the corollary that analysts and customers don’t always know what they don’t know so markets can change quickly.
In Part 2, the challenges to achieving a trillion dollar market along with example applications will be discussed along with key takeaways.
Just for fun, here is an illustration of what one trillion dollars looks like.