The Great Chip Shakeup

It’s too early to predict who will win and why, but there are huge new opportunities everywhere.

popularity

Facebook, Alibaba, Google, Apple and Samsung are all designing their own chips. So are Cisco and Huawei. So what exactly does this mean for big chipmakers and the semiconductor ecosystem?

While your first impulse might be to draw a straight line between Qualcomm’s decision to cut 1,500 jobs and reports about giant systems companies developing chips in-house, it’s not clear there is any correlation between these events. The semiconductor industry is moving into uncharted territory with autonomous vehicles, AI/ML/DL, IIoT, IoT and cloud-based services, and it remains to be seen where and how technology developed for each of these markets overlaps.

Qualcomm was the big winner in the smartphone era, but that market is flattening. In fiscal Q2 2018, ended March 25, the company’s revenue grew 5% to $5.3 billion, compared with $5.0 billion in the same period in 2017. Net income was $400 million in the latest quarter, a steep drop from the $700 million it earned last year, which may set off some alarm bells. But Qualcomm’s next big opportunity is tied to 5G, which isn’t expected to start picking up steam for at least a year or two. The company is retrenching to prepare for the next wave, while trying to carve out new territory in sectors such as automotive. That helps explain why revenue is up and net income is down. The pending acquisition of NXP, which acquired Freescale’s automotive IC business in 2015, would give Qualcomm a significant boost in that market.

On the flip side, STMicroelectronics has been struggling to regain its footing in recent years, but its earnings surged last quarter due to a pickup in the automotive, industrial and IoT markets. These are markets that several years ago relied on low-margin commodity, off-the-shelf components. Much has changed since then. ST’s 2018 Q1 net income was $239 million, compared to $108 million in the same period last year. In Q1 2016, the company lost $41 million, and in the same period in 2015 it lost $22 million, according to company reports.

Overall, the chip industry is doing well. According to the Semiconductor Industry Association, total chip sales grew 21.6% in 2017, topping $400 billion for the first time. And the growth numbers were even higher in January and February of 2018.

Meanwhile, there is a war being waged by big systems companies is on the inferencing side of AI/ML/DL, moving advanced computing capabilities to the edge for a variety of capabilities that don’t exist today. The challenge there is to develop systems that can pre-process or fully process enormous quantities of data rather than sending it all to the cloud, and much of that has to be done on a single charge of a battery. The only way to make that happen is to tighten the interaction between algorithms and hardware. And because the system companies own most of the data—and because they’re reluctant to share that with outside companies—the obvious path to improving compute efficiency and speed is to develop hardware based upon their particular type of data.

This is a new market jam-packed with startups and well funded by VCs and the systems companies themselves. Growth here ultimately may impact existing chip companies, but it’s too early to tell. What is clear, though is that for the chip ecosystem—EDA companies, IP companies, foundries, equipment makers—this is like selling shovels in a gold rush. And the story for those companies gets even better. Unlike in the past, the race by systems companies to build chips in-house moves all of this infrastructure one step closer to the end markets they serve, potentially opening up new opportunities that reach well beyond the chip.

So how all of the pieces fit together, and who will be the winners and losers, is unknown. But old correlations no longer apply because these are entirely new markets that never existed in the past. Projections and conclusions based upon those numbers need to be viewed in the context of insufficient data and fluid market dynamics, and should be viewed with caution.

Related Stories
More Nodes, New Problems
Acceleration of advanced processes, skyrocketing complexity and cost, and concerns about IP availability are raising some difficult questions.
Transistor Options Beyond 3nm
Complicated and expensive technologies are being planned all the way to 2030, but it’s not clear how far the scaling roadmap will really go.
Design Rule Complexity Rising
Total number of rules is exploding, but that’s only part of the problem.



2 comments

Dr. Dev Gupta says:

remember how things used to be in the days of mainframes and mini computers, a handful of companies ea. w/ their own processor designs, ea. selling a relatively small numbers and then came the PC deluge that buried them all. Microprocessors of general purpose design ( therefore ea. chip had to have more transistors ) made at in house Fabs that operated at higher yields even at smaller nodes, So gradually the specialized design machines faded. this cycle may repeat again for AI processors specially those for IoT / edge computing / inferencing where the volumes will be high. Established fabless design houses will have to compete against start – ups who too can go to a Foundry. Time for IDMs with well thought out strategies to establish new standards

Cliff Keller says:

How many of these companies will be using FPGAs for their in-house designs as opposed to foundry services?

Leave a Reply