TSMC’s Slam Dunk

 The problem is if you don’t make money from one generation you haven’t got enough money to invest in a fab for the next generation, which is why so many companies have become fabless over time.

Back in the 1980s companies like IDT, Linear Technology, Cypress, Seeq, Xicor,  and others could raise $20 million or so of venture money and build a fab with it.

Gradually, as the cost of fab rose to today’s $20 billion, the number of companies who could sustain the costs of building and operating a leading edge logic fab whittled down till three were left: TSMC, Samsung and Intel. 

At the  2nm process generation, Intel and Samsung took a gamble on getting ahead of TSMC by going for GAA and BSPD at the same time and failed to get their processes to yield.

That means they will lose most, if not all, of the revenues  from the 2nm foundry market. Without those revenues it looks questionable whether they can raise the resources to build the next generation of dedicated foundry fab.

TSMC, which adopted a conservative strategy on GAA and BSPD, looks like taking the vast majority of the revenues from the 2nm foundry market, with Samsung apparently unable to find any foundry customers for its 2nm process, and Intel reported to be about to give up looking for customers for its 2nm  process.

For TSMC it’s a slam dunk.