Trouble Spots And Optimism For 2015

A survey of top execs points to a solid upcoming year, but significant changes are underway as the industry grapples with thorny business and technology issues.


Most top executives in the semiconductor industry are bullish about 2015 and even beyond, particularly as the begins to drive new markets and market mash-ups, and as more semiconductors find their way into markets such as automotive, health-care and manufacturing. But it’s not an entirely rosy picture, and top executives point to potential trouble spots on both the business and technology fronts.

Unlike previous years, the concerns are less about overall demand for chips and new designs. There are enough new markets, along with increased demand for chips in existing markets, that demand isn’t likely to be an issue next year. The bigger issues involve such things as outside investment, consolidation, market shifts, and the rigors of moving to new materials and technologies.

Consolidation and new investment
Perhaps the biggest unknown on the business side involves consolidation. The proposed acquisition of Spansion by Cypress Semiconductor, the folding of LSI Logic into Avago, widespread restructuring in Japan—not to mention the scaling back of Broadcom’s baseband cellular chip businesses and Texas Instruments’ embedded chip business—have narrowed the field of companies that will be buying tools and developing new chips.

“The amount of money being spent on tools historically has been proportional to the R&D of the semiconductor industry, which in theory could make it hard to grow EDA,” said Wally Rhines, chairman and CEO of Mentor Graphics. “But realistically the R&D of EDA grows with the revenue of the semiconductor industry, which is predicted to be 7% for the year. Given that, we should see a good year next year.”

Consolidation is a natural force in business, but consolidation coupled with the end of Moore’s Law could have some interesting repercussions.

“The pricing curve has changed permanently in terms of the rate of decline,” said Greg Waters, president and CEO of Integrated Device Technology (IDT). “We’ve been on a super aggressive price curve, but in a year or two you will see that change. That’s good news because you’re going to start seeing some new innovation coming into this market.”

Lip-Bu Tan, president and CEO of Cadence, pointed to the same consolidation trend, which he said would continue among semiconductor companies. But he said a counter trend is China’s funding of new companies, both inside and outside of China.

“There will be more companies that will become known outside of China next year,” Tan said. “There also will be more customer consolidation.”

Money is flowing into and out of China. China’s Ministry of Commerce said that outbound foreign direct investment would be $120 billion in 2014, a growing percentage of which involves technology, such as Lenovo’s $2.1 billion purchase of IBM’s low-end server business, according to Rhodium Group. Meanwhile, Intel will invest up to $1.5 billion in Spreadtrum and RDA Microelectronics in conjunction with China’s Tsinghua Unigroup.

The Chinese government also is investing between $3 billion and $5 billion per year inside its borders. What makes this different from previous efforts, according to Rhines, is that it is being done through private equity companies. “The government is investing alongside private investors. About half of that is in manufacturing, but the rest includes fabless and other companies.”

Done right, this could open massive new opportunities to outside companies doing business in China. Done wrong, it could limit the number of players inside of China, as well as the innovation. At this point, numerous executives say they don’t know which way the tide ultimately will go.

New and aging markets
Another big unknown involves the Internet of Things, a grand vision that likely will remain poorly defined and executed in 2015. That doesn’t mean it will be any less important, though.

“We’ll see continued strength in the industry and new markets opening up,” said Scott McGregor, president and CEO of Broadcom. “We’ll also see growth in some of the same markets, so there will be demand for large data center build-outs. If you look at HDVC (high-definition visual communication) and ultra HD in the home, that requires twice the bandwidth of digital TV. That requires an infrastructure build-out. We also will need much more content and more diverse content.”

McGregor noted that while the definition of the IoT is still a bit fuzzy, the opportunity appears to be very real. “We certainly need more products—and not just toys. Many of the announcements we’re seeing are somewhat lame, but some are real. If you look at fitness and health care, people can relate to those.”

There also is likely to be a shift to lower-end smart phones as markets in more established economic regions hit saturation.

“Mobility is starting to slow,” said Charles Janac, chairman and CEO of Arteris. “There is still plenty of innovation left to be done, but we’re already seeing carrier consolidation and everyone who is going to buy a new phone already has one. There is still a huge Chinese white box market, but the market that’s really looking exciting is the automotive market. Tesla is forcing car companies to be more innovative with downloads, apps on smart phones, proactive service, and integration with smart phones and cars. These all require a lot more silicon and software. There are 60 million cars produced every year, and each new car has 10 to 15 SoCs. That’s a lot of electronics. It’s a very large opportunity.”

And while the various markets that will ultimately comprise the IoT continue to ramp—automotive, health care and consumer, among others—there is a much faster ramp under way for the Industrial Internet of Things, or IIoT.

“Most of the early applications that have gotten attention are consumer-related, but with the customers we talk to the real interest is on the industrial side,” said Grant Pierce, president and CEO of Sonics. “It’s all about what data is being collected and what’s being done with it. That leads to concerns about how much power is being consumed by devices and security around all of the data so we can make sure it isn’t coopted by individuals.”

Pierce said the big opportunity ahead is on the subsystem side to productize security, connectivity and low power.

Hardware, software and people
One of the big shifts that started several years ago is the emphasis on software as a way of differentiating products and adding new features. But making software more energy efficient and more effective at using what is developed in hardware is a relatively new challenge—one that was suggested years ago but never realized.

“The interaction between hardware and software increasingly will be the center of gravity,” said Aart de Geus, chairman and co-CEO of Synopsys. “This can be either an increasing cost or a way of differentiating yourself. Today, there are more software engineers than hardware engineers. All companies say they only charge for hardware, but the reality is they charge for both. So do you do less software or more software to differentiate? The answer is both.”

The question, though, is who will actually do that work. Cadence’s Tan observed that chipmakers and EDA are competing for engineering talent and graduating students against companies such as Google, Facebook and Apple, all of which offer much better stock options and myriad stories about the best way to get rich quickly.

“Innovation is a top priority, and we will need to innovate everywhere,” Tan said. “If you look at EUV and finFETs, those require massive investments on all sides. That will have to change. There is too much risk. If we don’t, we won’t have enough innovation.”

One possibility is that there simply will be fewer companies building hardware, an eventuality that semiconductor service companies—Open-Silicon, eSilicon such as ASE and Amkor—are preparing for.

“The problem is that companies are not getting a return on money spent in creating unique hardware,” said Taher Madraswala, president of Open-Silicon. “There is more money to be made in applications and software. And you can see where that’s coming from if you look at the bill for 28nm and 16nm hardware.”

Madraswala predicts a shift will begin starting next year, and continuing through 2017, where more companies will focus on software than hardware, which could help take care of a possible shortage of hardware engineers. “Not every company will need 800 to 1,000 VLSI engineers, but they do need hardware to run their software on. Even with big data, the algorithms doing the giga analysis still need hardware, but the chips, the capacitors and the boards—that whole design—will be outsourced.”

eSilicon, likewise has been developing an online database to allow customers to make tradeoffs on hardware and IP—as well as perform the chip-making. Jack Harding, eSilicon’s president and CEO, said the online configuration tool, which was rolled out this year, already is being used in 45 countries.

Nuts and bolts
Even finFETs are new technology. So far, the only company that is in volume production of finFETs is Intel. As the volume of chips produced at 16/14nm continues to grow, new problems will turn up. Mentor’s Rhines said will allow tool makers to fine-tune their test strategies, DRC, parasitics and PERC tools.

“This all was developed five or six years ago, but now that volume is beginning to ramp changes will need to be made,” he said. “Another challenge will be at 7nm. 2015 is a big year to finalize what will happen at 7nm, and right now it is not at a baseline that’s stable.”

What is stable is the technology at more established nodes such as 28nm. “The mystery is largely gone at 28nm and we’re starting to see a lot of development at that node,” said eSilicon’s Harding. “We’re also seeing that 16/14nm finFETs are just a check point. All the real action is going on at 10nm. So we largely ignored 20nm and have waited a lot of cycles. That means you have to leapfrog into IP development at the next node. The process may work, but I’m not sure the industry will have the right IP. At the very least, there is sufficient confusion about what to target, and that could make the market bumpy next year.”

All of this has helped fuel the market for 2.5D, Harding said. “In the last 90 to 120 days, 2.5D has become very real, and with it come the ability to facilitate innovation. For the OSATs, this could be very good news.”

The bigger picture
Taken as a whole, there are some significant issues that will continue to define the semiconductor industry in 2015 and for years to come. One is the growing cost of developing chips at the most advanced nodes. While some people insist that Moore’s Law is still alive and well, the fundamentals that made Moore’s Law possible in the first place have changed. There are reasons why some companies will continue to shrink feature sizes and cram more onto a die, but the economic underpinnings of cutting costs per transistor at each new process geometry are gone. EUV is not likely to be ready until 7nm, which will require double patterning even with EUV, and even then multipatterning, more layers and more complexity will push planar shrinks into the realm of custom silicon.

It’s possible that die stacking, particularly with 2.5D configurations and possibly monolithic 3D-ICs, will help propel the industry forward, but for the immediate future the real activity seems to be at older nodes, including 28nm where there are a raft of process changes under way to extend that node.

“In 2015, we’re going to see the first real step up for FD-SOI,” said Paul Boudre, chief operating officer at Soitec. ” Samsung and GlobalFoundries are both behind it, and next year you’re going to see customers talking about their applications and products using FD-SOI. And in the next two to three years we’re also going to see FD-SOI finFETs for high performance and low power.”

No matter what changes happen on the process or architecture side, though, the demand for chips is rising. IDT’s Waters notes that even though economies are slowing in China, Russia and Europe, the chip content in almost everything is increasing. And barring any economic catastrophe, that should continue unabated through 2015.

As Synopsys’ de Geus observed, “The market says give us more stuff. And there are many more opportunities because the stuff is becoming smarter stuff.”

Leave a Reply

(Note: This name will be displayed publicly)