The technology sector in recent months has faced significant challenges caused by major constraints in the supply of semiconductors. This has had the knock-on effect of hampering the production of consumer items such as cars, TVs, phones and more.
While the cause of the shortage is not the topic of this piece, the implications are both tactical and strategic in nature. Procurement teams are under significant pressure to find and secure everything they can lay their hands on. However, it also has forced development teams to think about ways to shield their company when shortages inevitably strike again.
Given the ever-increasing silicon content in a modern vehicle, driven by significant innovation in the occupant user experience, advanced driver assistance systems and of course the dramatic increase in electric vehicle demand, what options are available to development teams to protect against recent events?
One option is to take on development of semiconductor products in-house. The lure is obvious as it puts one in better control of their own destiny. But such an undertaking should be carefully considered. There are a number of key factors that must be addressed such as product definition, design options, schedule and, of course, cost. Let’s look at each of these.
Product definition can be a complex process especially when trying to scale across multiple applications or use cases. This often leads to a compromise with the trade-off being a reduced number of different semiconductor devices able to be developed.
These decisions encompass performance, architecture, connectivity, peripherals, power requirements and physical constraints as well as the software strategy to interface with the device within the finished product. These are all vectors that any development team should have a good understanding of.
Next, we have the design of the device itself – probably the most complex task. Designing a highly integrated “system-on-chip” (SoC) device is far from trivial and is not a skill set that most auto companies possess.
However, there is help available from a growing network of design houses that specialize in full turnkey services from design-to-supply. These companies have silicon design experience and based on customer input, can source the required IP, design the integrated devices themselves and work with fabs to manage and deliver product volume.
Schedule is always a challenge. Developing a complex semiconductor from project start to volume production averages 18 months. That can contribute heavily to a system design project schedule, not forgetting that the SoC sits in the critical path.
Finally, and probably most important, is cost. The development of a single complex multicore silicon device can easily run in to the tens of millions of dollars.
This cost breaks down into three fundamental, pre-production chunks. First, licensing of IP is significant. This is directly dependent upon the device feature set with high-performance CPUs and GPU cores commanding top dollar.
There are market dynamics at play that help reduce cost with new back-end-loaded licensing models from popular IP vendors as well as the emergence of the license-free RISC-V architecture that is gaining popularity.
Labor cost is the next vector and will be directly proportional to the number of engineering hours spent designing the device. More complexity equals more hours.
Finally, there are the mask sets that are required for mass production of the device. Again, these costs are proportional to device complexity and are measured by gate count and coupled with the process geometry with the latest sub-10-nm nodes being considered premium.
So, is it really worth developing your own in-house silicon? Unfortunately, it’s not a simple yes or no answer. The industry has certainly taken steps to make this approach more accessible to the masses.
However, it remains a significant adventure in both time and money that ultimately can bring greater control over the availability of critical components.
Yet there is one fundamental consideration above all else: The global silicon fab capacity is finite, so even taking this approach is not a guarantee of supply when the next constraint cycle kicks in. You will, unfortunately, still be competing for capacity and relying heavily on the supply contracts you negotiated.
Neil Stroud (pictured, above left) is vice president of marketing and business development for CoreAVI, a developer and provider of safety critical graphics and compute software drivers and libraries, embedded system-on-chip and discrete graphics processor components.
