The Chip Insider®– Can’t forecast? Here’s why. SMIC's shocker

Summary: Questions, Answers, & Controversies

Can’t forecast? Here’s why:

Having problems with your forecasts going off soon after they’re released? Welcome to the crowd. You’re not the only one. Over the almost 50 years I’ve been forecasting the semiconductor market, one of my most important observations is there are times when it’s easy to forecast and times when it’s difficult. Right now it’s extremely difficult…

In the current environment, there are massive structural changes. This includes several emergent forces: HBM, … Hyperscalers, … Tokenomics, … Re-globalization of the semiconductor industry, which is ... And that’s not to mention the Japanese carry trade and its threat to America’s government debt. Also, there’s the greatest change agent of this century…

It’s the first time I remember that on a regular basis, you can go home on Friday with a solid forecast and have its baseline assumptions upended over the weekend. To all the forecasters out there, fear not the lack of accuracy in the current times. React fast to high frequency changes in the environment. And seek solace in getting direction and the emerging trend right.

SMIC's third-gen 7nm node shocker

I have to say it’s getting silly to see portrayals of Chinese advancements years behind as “shockingly” close to western technology. My YouTube is filled with shocking titles like “Did China just beat Intel?,” China Just Built What ASML Feared Most, “China Just Built What TSMC Feared the MOST.” Or “China’s Chip Giant Just Shocked the Industry with a Major Breakthrough.” It’s like the definitions of ‘beat, feared, shocking’ have been changed to ‘behind, meh, and boring.’ Just remember these definitional changes the next time you’re tempted to click on one of these and waste half an hour of your life just to send ad money their way…

In this SMIC 7nm N+3 case, “significant advancement” claim on a finding of higher density that TSMC’s N6 and a metal pitch that’s less than Intel’s 18A.

Wait a minute (screeching sound), “significant advancement” now means they’ve fallen from the 4 years TechInsights clocked them at in 2023 to now being 6 years behind? Either they got the sign wrong or maybe they’re measuring SMIC’s advancement against itself … and maybe to placate the Chinese? It’s kind of an insult to TSMC, because …

They must really want to insult Intel, or they don’t understand how interconnect architectures are changing… What they’re doing is like comparing a 4-cylinder engine to a 12-cylinder one based on the latter weighing more. Yes, but, the 12 will have more horsepower because its venturi are smaller, increasing the velocity of the fuel-air mixture. So more gets crammed into each cylinder, thereby delivering more horsepower.

“Forecasting is difficult… Especially when it's about the future” — Yogi Berra