The Risks of Custom Silicon

Ever since Apple has gone all in developing custom silicon chips like their A-series processors as well as companion chips like connectivity modules, an image sensor, display modules, and several others they don’t talk about publicly, other companies have followed suit.

I often get asked why companies feel the need to start making their chips when there are often great options from companies like Qualcomm, Intel, NVIDIA, and AMD, and many other semiconductor companies who have deep expertise in this area. The short answer is more and more companies are going to want more control over specific experiences with their products and therefore will want to have control over how those experiences are differentiated.

While it is true that many semiconductor companies will offer solutions that seem like a more logical choice to be used in a particular product the reality is many companies will continue to invest in their differentiation which will lead to more investments in proprietary hardware and silicon in many cases.

Google has been going deep down this road with their TPU efforts, and now with their TPU 3.0, their chipset strategy has been refined and focused on machine learning acceleration. Similarly, Tesla has now shown their cards on their custom silicon development stating they are preparing to shift away from NVIDIA and go all in with their chips dedicated to their autonomous systems.

While I expect this trend to continue, with other big companies starting to enhance and strengthen their differentiation with custom hardware, I’ve been thinking lately about the risks associated with this strategy for many companies.

What’s at Stake
By going down the proprietary silicon route, companies are betting their differentiation on this strategy. I say it this way because of one common theme with companies efforts in custom silicon is how focused it is on a specific area of differentiation. If one did an in-depth analysis competitively, you would likely find where these companies believe they are differentiated by where they try to build custom silicon to enhance or deepen that differentiation.

The risk, however, in this scenario is their efforts in custom silicon are not able to sustain their differentiation as companies like NVIDIA, Qualcomm, Intel, AMD, etc., design better, more powerful solutions that enable said companies competitors to have an even better offering. To elaborate, I’m going to pick on Tesla.

Tesla dumping NVIDIA seems like a fairly significant risk. Especially, since Elon Musk and company are saying they are going to build a better autonomous system, including all the extremely complicated set of chips needed to create self-driving cars. Essentially, Tesla is saying they can make a better solution than NVIDIA, and I find that hard to believe.

With Tesla, the risk may be more multi-faceted than it is with other companies because human safety is a factor as well. Do we believe that Tesla has the experienced semiconductor engineering teams that can out-innovate NVIDIA a company with decades of experience and some of the best silicon engineers on the planet (many focused solely on autonomous systems)? Alternatively, to make the point another way, would you trust a self-driving car running unproven Tesla silicon or proven and vetted NVIDIA technology?

In this scenario, Tesla will be up against NVIDIA who will likely be powering the vast majority of Tesla’s competition in the automotive industry. While Tesla will focus their efforts on some specific areas like AI and machine learning, I have a feeling they will be challenged to keep up with NVIDIA in the more fast feature set around automotive. There is a significant risk Tesla is not as competitive not just in fully autonomous systems but the broader landscape of technology invading many corners of the automotive experience.

Tesla’s risk is not unlike Apple’s in that Tesla has decided to invest in the total solution as they want to develop a CPU/GPU and additional sensors needed for autonomous systems. Similarly, Apple is becoming increasingly less dependent on third parties for their silicon, but the difference here is Apple has not only purchased core expertise by buying semiconductor companies, but they also have hired/built one of the best teams on the planet poaching many of the best silicon engineers from companies like Intel, Qualcomm, NVIDIA, and AMD. Tesla has built no such group, and I’m skeptical they can.

While Google is at risk as well, they are focusing just on a few areas with their chips. They aren’t trying to eliminate NVIDIA, AMD, or Intel but rather develop semiconductors that accelerate or compliment those companies technologies. This makes their risk slightly less, but the proof will be in the execution and performance. Google’s cloud platform benchmarks in these competitive areas will need to yield on-par or better results than companies like Amazon or Microsoft who are using the total solutions from third parties, for now.

If Google’s performance benchmarks don’t remain competitive and companies like Intel, AMD, and NVIDIA out-innovate Google with their total solutions then companies who rely on the above names then their investments in silicon become a disadvantage instead of an advantage. Here, like Tesla, it all comes down to talent. Here again, the best silicon designers are not at Google, and I’m skeptical they will go there.

While I appreciate the strategy and efforts of companies investing in silicon, I also know how steep this road will be for many of them. Time will tell if this strategy becomes a strength or a weakness competitively, but there will be more pressure and more spotlight and scrutiny on the Google’s and Tesla’s of the world as they embark on this path to prove their efforts will pay off.

Published by

Ben Bajarin

Ben Bajarin is a Principal Analyst and the head of primary research at Creative Strategies, Inc - An industry analysis, market intelligence and research firm located in Silicon Valley. His primary focus is consumer technology and market trend research and he is responsible for studying over 30 countries. Full Bio

322 thoughts on “The Risks of Custom Silicon”

Leave a Reply

Your email address will not be published. Required fields are marked *