What a difference a year makes. First the aesthetic: after two years of maneuvering around a construction zone at SEMICON West, the Moscone Center renovation is complete and looks fabulous. It’s also so much easier to navigate between halls, and to and from the keynote stage at the newly christened Blue Shield of California Theater.
Last year at SEMICON West, the semiconductor industry was flying high on soaring revenues, confident that technology megatrends like 5G, artificial intelligence (AI), machine learning (ML), quantum computing, the internet of things (IoT) and autonomous vehicles would propel us further because, after all, semiconductors provide the underpinnings of all these technologies.
This year, while still optimistic about the impact these same technology megatrends will have on semiconductor growth, the cloud of the geopolitical situation looms large, casting a shadow of uncertainty across the industry. SEMI reports downturn of 12% so far in 2019, caused largely by a 25% drop in revenue for the memory sector.
In addressing the SEMICON West press corps during the first annual lunch, SEMI’s Dave Anderson offered a silver lining: the downturn is expected to be short-lived, and we can all look forward to an uptick in 2020 as US-China relations stabilize. Despite it, the industry just experienced its third-best year ever. Ironically, equipment sales in China are expected to surge, and for the first time ever, China is the industry’s largest market.
Starting with this post, and over the course of the next few weeks, we’ll be looking at how the technology megatrends and the geopolitical situation is impacting the suppliers to the industry, based on interviews I conducted last week on the show floor.
The Technology Megatrends are All about the Data
Despite all the hype, there is a general consensus among industry experts that all of the technology megatrends are still very much in their infancy. Ultimately, they are needed to provide ways to process vast amounts of data into something that can be used. Artificial intelligence (AI) is at the heart of this. It’s going to take advancements in architectures, materials, and packaging to drive power efficiency and cost.
Industry 4.0 is happening, noted FRT’s Thomas Fries. There are lots of devices on equipment gathering data, for example, and we need AI to measure, analyze and evaluate the data, and then act on it. “We need more electronics, MEMS, and semiconductors to serve the growing population, manage resources and organize everything, optimize logistics and control production and the use of resources,” he said. This calls for advancements in all semiconductor devices, from transistors and finFETS to sensors and RF devices. Small structures for high frequencies and nanometer-scale electronics and more 200mm fabs are needed. “We need it all to manage everything. We can’t stop it (progress). This is a wonderful situation for FRT because as a small, independent company, we are nimble and can adjust to all of it. There is no downturn for us,” said Fries.
Kevin Crofton, CEO SPTS, is pragmatic about the megatrends and sees them more as feeders of the industry than drivers. In the long term, they will push the app aspect of our industry, he said. Crofton talked about the human data experience that is coming when everything we do will have data associated with it. For example, he sees great opportunity in the digitization of our personal health administration. “iPhones and smartwatches are the building blocks of the future to pass data to our HMOs,” he said. “But it has to be simple devices that monitor vital signs and blood chemistries that can be easily transmitted using RFID chips.”
Are Fully Autonomous Vehicles still a Technology Megatrend?
For the past few months, I’ve noticed a shift in the autonomous vehicle story, from the push to reach level five autonomy as quickly as possible, to a focus on advanced driver assist systems (ADAS) and electronic vehicles (EV), which together represent a huge market for the semiconductor industry.
Tim Kryman, Rudolph Technologies, noted that while the Tesla and Uber crashes certainly put the brakes on the race to full autonomy, with ADAS and EV, the increase of Si content in automotive applications hasn’t slowed. “The power space is going through the roof,” said Kryman. “Gallium Arsenide (GaAs) and Silicon Carbide (SiC) materials are changing how people approach power.”
Trymax’s Yann Guillou seconded that sentiment, noting that in addition to ADAS, infotainment and sensor content is increasing not only in high-end vehicles but are becoming standard in midrange cars as well.
“Automotive electronics use more complex chips, and standards are more stringent than other applications,” noted Raleigh Estrada, Zeiss. This has been a big motivator for the company’s investment in 3D X-ray imaging solutions for package failure analysis.
Crofton says he sees a different situation in the US, where he is from, then in the UK, where he lives and works. He believes Asia and Europe will adopt autonomous vehicles more quickly than in the US, partly due to logistics, and partly due to the litigious nature of the US. He also expects it will first be rolled out in fleet vehicles before personal use. Realistically, he says it will be 10-15 years before Level 3 autonomy is ubiquitous. A Level 3 vehicle can take full control and operate a vehicle when certain conditions are met, for instance on a freeway journey, excluding on and off-ramps, and city driving.
The 5G Race: Who will win?
Because a 5G infrastructure will require literally hundreds of micro-base stations placed in close proximity to work, the transition to 5G will vary according to geographical location. Most agree that the US will be the last to make the transition.
According to Bruce Kim, CEO of SurplusGLOBAL, Korea has already won the race to be first, with the launch of the commercial infrastructure. This is because Korea covers a relatively small area and is densely populated, providing the ideal environment to cost-efficiently build an infrastructure.
It stands to reason, then, that Samsung was the first device manufacturer to put a 5G ready handset on the market. There’s not much point in rolling out 5G phones if you can’t find service for them.
This is perhaps why Apple has held off on introducing a 5G handset until its 2020 model. Early adopters will get the phone, but 4G phones won’t be obsolete, so there’s no reason to panic and rush out to replace your phone unless you really need to download videos in seconds versus minutes.
Other applications promised to benefit from 5G aren’t here yet. Where will 5G make a difference? For anything that is time-critical and requires massive data transport, said Thomas Uhrmann, EV Group. For example, car-to-car communication for autonomous vehicles. AI isn’t mature enough, but 5G will help enable it.
Uhrmann is not fussed about 5G. He says it will experience a similar adoption as 5G. He explained that only the high data rate bands are different than 4G because they are further apart. All the other bands are the same. “5G needs mmWave technology for massive data transport, but for anything else, it’s really just a better 4G,” he said. The real issues are not coming from the first 5G module. First implementation of 5G is really just 4G plus. LTE handsets can support a 5GE signal. He calls it “5G lite”.
The real changes will come for EV Group when 6G arrives. Then we won’t be just working in Silicon. It will also require SOI, GaaN, and GaAs to handle frequency and power, he said.
A (final) few words about Moore’s Law
I can’t end this blog post without at least mentioning the opening question of the AI Design Forum’s CEO Roundtable, which, by the way, wasn’t a round table discussion at all. It was a panel with a moderator who asked all the questions. There wasn’t time for a Q&A from the media and analysts in the room – a bit of a disappointment, if you ask me.
Anyhow, the moderator, New York Times journalist, John Markoff, opened with the worn-out question, Is Moore’s Law dead? Why are we still belaboring this topic when there are so many more exciting discussions to have?
Let’s not re-package Gordon Moore’s intent in his 1965 paper so that we can say that Moore’s Law is “very much alive”, as Synopsys’ Aart De Geus did. Nor should we think about it as “the behavior of an exponential that has techonomic feedback on the exponential that drove a revolution of what mankind can do,” as he suggests, because what does that sentence even mean? (Thank you, SemiEngineering, for recording the round table discussion and posting the direct transcript here.)
Rather, as Lisa Su, AMD suggests, we should be focusing on how to keep the performance trend going for our industry and find the right computing environment to process the tons of data coming from 10s of billions of devices. A heterogeneous compute environment is the best choice for that. Suffice it to say that Moore’s law is out, heterogeneous integration is in. Enough said.
In our next post, we will look at the impact of the geopolitical situations on these companies and will conclude the series with a look at new capabilities from these suppliers. ~ FvT