What’s The Other Guy Doing?

Competition is generally a good thing. It improves service, promotes innovation, forces efficiencies and price cuts where necessary, and it ratchets up the pressure to bring products and services to market faster. Those who can’t keep up usually lose market share, and eventually the business sector consolidates until something comes along to disrupt it.

That cycle has been repeated in every major industry for hundreds of years, and the semiconductor industry has pushed it to a level that has garnered attention from every other industry on the planet. But more recently it appears those trends are no longer working. As prices for equipping advanced fabs continues to rise, and as customers continue to weigh their options—28nm FD-SOI, 16/14nm, 10nm, stacked die, and new revs of established node process technology—the normal competitive process has gone awry.

This is a high-stakes game where the ante is measured in billions of dollars, and so placing bets carries enormous risk—so much so, that no one is placing more than initial bets. At advanced nodes, the fear is that process nodes are occurring in too rapid succession, so why invest in 14/16nm when you can jump to 10nm? Of course, at 10nm there are new materials involved to counter the quantum effects, but because 20nm turned out to be something of a bust without the low leakage benefits of finFETs and the single pass of commercially viable EUV lithography, the whole two-year cadence for successive nodes has been interrupted.

To make matters worse, some chipmakers scurried back to 28nm FD-SOI rather than deal with any of this, while others have done only limited forays at 16/14nm. Still others gave up altogether and retreated into developing markets in automotive and RF at 40 and 65nm.

This is confusion at its finest, and after years of advancements in process technology and new packaging approaches at every node, it’s still hard to say which direction the industry will take next. EUV is still being tested, work is well under way at 10nm with work beginning at 7nm, and foundries and their customers continue to waffle while looking over their shoulder to see what the competition is doing. The answer, unfortunately, is not much.

If all of this was confined to one segment, it would be bad enough. The problem is that semiconductor design and manufacturing have evolved into a giant ecosystem. At the most advanced nodes, chips are a collaborative effort between chipmakers, foundries, EDA companies and IP suppliers (although increasingly it’s difficult to tell the difference between IP and EDA companies). This is where the most money historically has been made, and it has been the source of reinvestment at the next nodes and the ones after that.

Any blip in that finely tuned model can have massive repercussions for years to come in multiple segments. There are a few big questions that remain, though:

1. Will it set back progress in design, and if so how far?
2. Will the lack of new capacity at advanced nodes really slow down production, or will existing fab capacity handle the demand.
3. Will the IoT, or at least smart devices in automotive, consumer and health care pick up the slack if there are not enough new designs hitting the market on the mobile end?

These are important questions, and so far there are no answers.