Moore's Law Eats Its Own

The big news in the semiconductor world this week is Intel's formal announcement of delays in their 7nm process development. CEO Bob Swan indicated 7nm chips would not be ready for volume production until as late as early 2023.

Intel's own Gordon Moore coined "Moore's Law" back in 1965. Semiconductor manufacturers have been trying to keep up ever since. Intel has long been in a race with TSMC to bring the latest processes to market, and patent analytics can tell us that Intel lost this race years ago.

Here's Intel's patent applications by year, since 2010:

We're looking only at patent applications in the H01L21 classification -- these are inventions directed specifically towards, "Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices". This allows us to remove all the other applications and focus only on semiconductor manufacturing patents.

The dropoff in applications in later years is not entirely unexpected -- some of those later patents have not yet published, so that information isn't available. The question is, do we see a similar dropoff in TSMC's applications? Have a look...

Don't focus too much on the raw numbers, TSMC will certainly have more applications since silicon manufacturing is their whole business. Focus on the slope of the trendline. This is a dramatically different profile. This is what we would expect for a company driving innovation in this space. If we normalize each company's data, using 1.0 to represent their number of applications in 2010, the picture looks truly grim for Intel.

The early indications of Intel's declining innovation were visible all the way back in 2015. TSMC was continuing their trend in patent filings, while Intel's had already begun to drop significantly.

Intel is claiming, "We will continue to invest in our future process technology roadmap", but at this point I'm very skeptical. I expect Intel will make a 'face-saving' rollout of some amount of 7nm production in the future, but moving forward, TSMC will be their main source for the smaller nodes. They may end up in a similar situation as GlobalFoundries, who stopped their 7nm development back in 2018.

Failing to innovate is a recipe for falling behind. Moore's Law waits for no one. Not even his own company.