The repositioning of the semiconductor industry has started.
For the past couple of years we’ve been hearing about the coming onslaught of the IoT, difficulties in scaling device features and a shift left that will redefine verification, debug and software prototyping. It’s all happening.
Taken individually, these are noteworthy changes. Taken as a whole, they represent a massive repositioning of the semiconductor industry, from architectural exploration to design and all the way through manufacturing.
It’s easy enough to get lost in any piece of this shift. Industry consolidation is a big change in its own right. So is the prototyping of software earlier in the design cycle to make it more energy efficient and to get it out the door in working order. And the push to 7nm and the rise of advanced packaging and the introduction of high-bandwidth memory are significant developments.
But to see the big picture requires putting all of these pieces together across multiple regions and in multiple markets and extrapolating from there. When you do that, there are several key trends that show up.
First, semiconductors are becoming more important in more vertical markets and in more places. Going back even two years ago, SoCs inside of cars sounded like a PowerPoint slide about future growth projections. The IoT/IoE didn’t exist, except in analyst and corporate marketing presentations. And the industrial sector was just starting to grasp the potential benefits of what sensors inside of valves and on assembly lines could mean to their bottom line.
This isn’t several designs for billions of units, but it is tens of billions of designs for much smaller numbers of devices—sometimes in the hundreds or thousands of units. Moreover, many of these devices are much more complex than initially thought, and a good percentage of them will require the same kind of advanced tooling that mobile applications processors require. But it also will take some time for the semiconductor industry to adjust to this new world order. That will play out in more volatility in markets, more acquisitions before this process complete, and most likely more startups around the edges.
Second, security is rapidly emerging as the next big challenge for the industry, and a huge opportunity for EDA tools and IP. Connecting everything together opens the door to huge changes, including over-the-air updates for cars and other devices, but that door has to be closely guarded. Verification and debug tools will see a significant uptick in new applications, including a big push in formal verification and software debug.
In the past, most of the security focus was in software or networking, which were largely confined to the lower levels of the Open Systems Interconnect (OSI) stack. In the future, hardware and IP will be viewed as increasingly vulnerable—particularly as credit card theft becomes more difficult due to chipped cards and hackers look for new lucrative opportunities, and as more sophisticated state-sponsored cyber attacks begin ramping up side-channel attacks on memories and embedded code.
Third, increased reliability will become a requirement that will add time and cost to designs. One of the big changes ahead is that chips will need to last longer with no loss in performance or increase in power over time. That means redundancy and guard banding in some cases, but it also means more time spent on the front-end architecture to figure out how to minimize thermal problems and other physical effects, and on the back end for verification and debug.
Correct by design has been a mantra for some time, but for the most part there is little to show for it. That will have to change, and it likely will come in the form of proven subsystems, as well as advanced packaging and PCB design. Putting everything on one chip may be the least expensive approach in high volume. But at each new node that cost benefit will begin to erode because of the increased dynamic power density and, after 16/14nm, the return of leakage current and the damaging effects of heat. And if there are no markets that can warrant billions of the same chip, that economic benefit erodes even faster.
These are all interesting changes, and they each generate interesting problems that need to be solved and different ways of looking at problems in the first place. But there’s an upside to this, as well. While the changes on the macro level may look daunting—consolidation, rising complexity and the flattening of the mobile phone market—there are enough new challenges to keep everyone busy and fully employed for a very long time.