Semiconductor Foundry Battles

As an analyst with a background in the semiconductor industry (my first tech job was at Cypress Semiconductor), and heavily covering the semiconductor industry when I first joined the Analyst community in 2001, I was beginning to think this knowledge and expertise was helpful to have but becoming less relevant. For a stretch, it seemed like the semiconductor industry was growing stale. An industry that was once the pinnacle of enabling innovation felt as though it was an afterthought as smartphones, tablets, wearables, and more got all the attention.

However, in the past few years, the semiconductor industry has been undergoing a fascinating renaissance. A sector which one felt like it was a background conversation is now back in the, and I couldn’t be happier.

While there are so many things I can write on, including a dedicated analysis of what is happening in memory, the foundry/manufacturing topic is an important one to understand.

The foundry conversation made headlines earlier in the week when Global Foundries announced they would not be pursuing the 7nm process node anytime soon. The implications for this move mean that only two foundries exist for customers who want to design chips on 7nm and those are TSMC and Samsung. There is a critically important point to understand as the foundation for what is happening and how things may play out from here.

It’s All Apple’s Fault
Fascinatingly, you can draw a line to Apple as the source of some of the turmoil happening in the semiconductor industry. To understand how it is important to look at why Global Foundries is abandoning 7nm and why both TSMC and Samsung can afford to pursue 7nm and beyond.
Global Foundries simply stated there was not enough customer demand for 7nm. This is a partially true statement when the correct way to describe the problem is not enough customers can afford 7nm chips. It is becoming increasingly more expensive for foundries to purse new process node technology meaning that they need customers who can afford the premium pricing 7nm chips will command AND have the appropriate scale to buy in mass. Only one category, and only one customer has both the ability to afford the premiums of latest process node chips and the scale to buy hundreds of millions of chips, and that is Apple.

While this is entirely speculation, I can say with a high degree of confidence that this foundry situation would be different if Apple were not designing its chips. How different is up for debate, but assume for a second that Apple was buying chips from Intel, as was reportedly discussed by Steve Jobs and Intel then CEO Paul Otellini. If Intel were making chips for Apple, it would have given them the financial boost to keep investing in leading-edge process technology because they would have had the margins and the scale to make it worth it. Had Intel worked with Apple on this instead of Apple designing their own, there is an excellent chance Intel would have been able to maintain process technology leadership.

The other area where Apple has impacted the semiconductor market is by dominating demand for high-end chips at scale in the smartphone market. Because Apple is the ASP leader in smartphones, they have mostly sucked the life out of the high-end for everyone else. This carries with it the implication that competitors need to compete in much lower price tiers which means they can’t afford the premium tier of semiconductor chips. This point alone lessens demand for the newest chips at cutting edge process nodes and thus makes it hard for manufacturers to invest the significant RND necessary to keep pursuing next-generation transistor designs.

This ultimately will lead to a theory I laid out a while ago about the last foundry standing. It stands to reason that if it becomes increasingly expensive to get to new process technology and only a small handful of companies can afford to pay the premiums to help justify that investment that fewer fabs will try. If only TSMC and Samsung are making 7nm now, and Intel and Global Foundries are years behind, then a big question mark emerges about who can afford to get to 5nm? Again, I firmly believe this is less a technology problem but more an economic one. Yes, it is becoming increasingly difficult technology-wise to get to 5nm and beyond, but enough R&D is what is needed to do it. Few companies can justify the investment in R&D because few companies can afford to pay the premiums.

While I fully expect Intel to get to 7nm, I think it will become increasingly clear to Intel and their new CEO that economically it may be even harder and possibly not feasible to get to 5nm. Basically, for all the reasons Global Foundries found it not financially viable to get to 7nm, I think it is likely 5nm will cause even more turbulence.

So the question will remain, who is the last fab standing and interestingly, Apple has a lot to say about who wins and loses. For a company not making semiconductors a decade ago to now essentially have such incredible influence in the semiconductor industry is pretty fascinating.

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

One thought on “Semiconductor Foundry Battles”

Leave a Reply

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