TSMC - 3nm FAB to cost $20B

[Phynaz]

By the time we’re through, by the time we’ve built all the necessary capacity, I think we would have spent upwards of $15 billion,” said Chang, who’s calling it quits in June and handing the reins of his $190 billion-plus company to lieutenants Mark Liu and C.C. Wei. “That’s a conservative estimate. Maybe it’s safer to say upwards of $20 billion.

 

[coercitiv]

For the record, I liked the post, not the cost.

 

[Topweasel]

A lot of that is because they are trying to jump start everyone by having this done years before most companies have even a 5nm plan. They are basically going to have to create everything by hand for this. Also sounds like it's going to be big enough to cover all future 3nm demand instead of starting with 1 and then upgrading other fabs as they transition.

 

[Roland00Address]

Real question, is this a thing where you should be penny wise and pound foolish, or will too much risk doom the company. Lets say the final cost is $25 billion and not $20 billion or it is $30 billion and not $25 or $20, well if you are 1 year late, or 3 years, or if you do not have the right amount of capacity you may doom your marketshare of leading node supplies forever.

Yet if you overspend you may just be wasting money, and if this was some other company that did not have all these already made fabs you may be risking bankruptcy. And even with all these other fabs already made you just have to look at the RAM or Flash memory markets to see many companies go bankrupt turning sand into dollars / gold / stock money.

 

[Drazick]

One good month of Apple and they have the fab 🙂.

 

[VirtualLarry]

Will the DRAM / NAND "shortage" be over, then?

 

[whm1974]

Will the DRAM / NAND "shortage" be over, then?

I sure hope so. And BTW I thought silicon was only able to scale down to 5nm?